Blog

My Dated Predictions

rodneybrooks.com/my-dated-predictions/

With all new technologies there are predictions of how good it will be for humankind, or how bad it will be. A common thread that I have observed is how people tend to underestimate how long new technologies will take to be adopted after proof of concept demonstrations. I pointed to this as the seventh of seven deadly sins of predicting the future of AI.

For example, recently the early techno-utopianism of the Internet providing a voice to everyone and thus blocking the ability of individuals to be controlled by governments has turned to depression about how it just did not work out that way. And there has been discussion of how the good future we thought we were promised is taking much longer to be deployed than we had ever imagined. This is precisely a realization of the early optimism about how things would be deployed and used did just not turn out to be.

Over the last few months I have been throwing a little cold water over what I consider to be current hype around Artificial Intelligence (AI) and Machine Learning (ML). However, I do not think that I am a techno-pessimist. Rather, I think of myself as a techno-realist.

In my view having ideas is easy. Turning them into reality is hard. Turning them into being deployed at scale is even harder. And in evaluating the likelihood of success at that I think it is possible to sort technology and technology deployment ideas into a spectrum running from relatively easier to very hard.

But simply spouting off about this is rather easy to do as there is no responsibility for being right or wrong. That applies not just to me, but to pundits ranging from physicists to entrepreneurs to academics, who are making wild predictions about AI and ML.

It is the New Year  and there will be many predictions about what will happen in the coming year. I am going to take this opportunity to make predictions myself, not just about the coming year, but rather the next thirty two years. I am going to write them in this blog with explicit dates attached to them. Hence they are my dated predictions. And they will be here on this blog and copies that live on elsewhere in cyberspace for all to see. I am going to take responsibility for what I say, and make it so that people can hold me to whether I turn out to be right or wrong. If I am unfortunate, some of my predictions will at some point seem rather dated!

I chose thirty two years as I will then be 95 years old, and I suspect I’ll be a little too exhausted by then to carry on arguments about why I was right or wrong on particular points. And 32 is a power of 2, so that’s always a good thing. So the furtherest out date I am going to consider is January 1st, 2050. And that also means that I am only predicting things for exactly the first half of this century (or at least for the first half of the years starting with “20” — there is a whole argument to be had here into which I am not going to get).

I specify dates in three different ways:

NIML meaning “Not In My Lifetime, i.e., not until after January 1st, 2050

NET some date, meaning “No Earlier Than” that date.

BY some date, meaning “By” that date.

Sometimes I will give both a NET and a BY for a single prediction, establishing a window in which I believe it will happen.

My RuleS OF PREDICTION

I am going to try to be very precise about what I am predicting and when. Now in reality precision on defining what I am predicting is almost impossible. Nevertheless I will try.

I had an experience very recently that made me realize just how hard people will try, when challenged, to hold their preconceived notions about technologies and the cornucopia they will provide to humanity. I tweeted out the following (@rodneyabrooks):

When humans next land on the Moon it will be with the help of many, many, Artificial Intelligence and Machine Learning systems.

Last time we got to the Moon and back without AI or ML.

My intent with this tweet was to say that although AI and ML are today very powerful and useful, it does not mean that they are the only way to do things, and it is worth remembering that. They don’t necessarily mean that suddenly everything has changed in the world in some magical way1.

One of the responses to this tweet, which itself was retweeted many, many times, was that Kalman filtering was used to track the spacecraft (completely true), that Kalman filtering uses Bayesian updating (completely true), and that therefore Kalman filtering is an instance of machine learning (complete non sequitur) and that therefore machine learning was used to get to the Moon (a valid inference based on a non-sequitur, and completely wrong).  When anyone says Machine Learning these days (and indeed since the introduction of the term in 1959 by Arthur Samuel (see my post on ML for details)) they mean using examples in some way to induce a representation of some concept that can later be used to select a label or action, based on an input and that saved learned material. Kalman filtering uses multiple data points from a particular process to get a good estimate of what the data is really saying. It does not save anything for later to be used for a similar problem at some future time. So, no, it is not Machine Learning, and no, we did not use Machine Learning to get to the Moon last time, no matter how much you want to believe that Machine Learning is the key to all technological progress.

That is why I am going to try to be very specific about what I mean by my predictions, and why, no doubt, I will need to argue back to many people who will want to claim that the things I predict will not happen before some future time have already happened. I predict that people will be making such claims!

What is Easy and What is Hard?

Building electric cars and reusable rockets is easy. Building flying cars, or a hyperloop system (or a palletized underground car transport network underground) is hard.

What makes the difference?

Cars have been around, and mass produced, for well over a century. If you want to build electric cars rather than gasoline cars, you do not have to invent too much stuff, and figure out how to deploy it at scale.

There has been over a hundred years of engineering and production of windscreen wipers, brakes, wheels, tires, steering systems, windows that can go up and down, car seats, a chassis, and much more. There have even been well over 20 years of large scale production of digitalized drive trains.

To build electric cars at scale, and at a competitive price, and with good range, you may have to be very clever, and well capitalized. But there is an awful lot of the car that you do not need to change. For the majority of the car there are plenty of people around who have worked on those components for decades, and plenty of manufacturing expertise for building the components and assembly.

Although reusable rockets sounds revolutionary there is again prior art and experience. All liquid fuel rockets today owe their major components and capabilities to the V-2 rockets of Wernher von Braun, built for Hitler. It was liquid fueled with high flow turbopumps (580 horsepower!), it used the fuel to cool parts of the engine, and it carried its own liquid oxygen so that it could fly above the atmosphere. It first did so just over 75 years ago. And it was mass produced, with 5,200 of them being built, using slave labor, in just two years.

Since then over 20 different liquid fueled rocket families have been developed around the world, some with over 50 years of operational use, and hundreds of different configurations within those families. Many variations in parameters and trade offs have been examined. Soyuz rockets, a fifty year old family, all lift off with twenty liquid fueled thrust chambers burning. In the Delta family, the Delta IV configuration has a “Heavy” variant, three essentially identical cores in a horizontal line, where the cores are all a first stage of the earlier single core Delta IV.

The technology for soft landing on Earth using jet engine thrusters has been around since 1950s with the Rolls Royce “flying bedstead”, with the later, at large scale, Harrier fighter jet taking off and landing vertically. A rocket engine for vertical landing was used, without atmosphere, for the manned lunar landings on the Moon, starting in 1969.

Today’s Falcon rocket uses grid fins to  steer the first stage when it is returning to the launch site or recovery barge to soft land. These were first developed theoretically in Russia in the 1950’s by Sergey Belotserkovskiy and have been used since the 1970’s for many missiles, both ballistic and others, guided bombs, cruise missiles, and for the emergency escape system for manned Soyuz capsules.

There has been a lot of money spent on developing rockets and this has lead to many useable technologies, lots of know how, and lots of flight experience.

None of this is to say that developing at scale electric cars or reusable rockets is not brave, hard, and incredibly inventive work. It does however build on large bodies of prior work, and therefore it is more likely to succeed. There is experience out there. There are known solutions to many, many, but not all, problems that will arise. Seemingly revolutionary concepts can arise from clusters of hard and brilliantly thought out evolutionary ideas, along with the braveness and determination to undertake them.

We can make estimates about these technologies being technically successful and deployable at scale with some confidence.

For completely new ideas, however, it is much harder to predict with confidence that the technologies will become deployable in any particular amount of time.

There have been sustained projects working on problems of practical nuclear fusion reactors for power generation since the 1950’s. We know that sustained nuclear fusion “works”. That is how our Sun and every other star shines. And humans first produced short time scale nuclear fusion with the first full scale thermonuclear bomb, “Ivy Mike”, being detonated 65 years ago. But we have not yet figured out how to make nuclear fusion practical for anything besides bombs, and I do not think many people would believe any predicted date for at scale practical fusion power generation. It is a really hard problem.

The hyperloop concept has attracted a bunch of start ups and capital for them, though there has been nothing close in concept that has ever been demonstrated, let alone operated at scale. So besides figuring out how to develop ultrastable cylinders that go for hundreds of miles, containing capsules that are accelerated by external air pressure traveling at hundreds of miles per hour while containing living meat of the human variety there are many, many mundane things to be developed.

One of the many challenges is how to seal the capsules and provide entirely self contained life support within for the duration of the journey. Also the capsules must be able to go past stations at which they are not stopping in a stable manner, so stations will need to be optionally sealed off from the tube for a through capsule, while allowing physical ingress and egress for passengers whose capsule has stopped at the station. There will need to be procedures for when a capsule gets stuck a hundred miles from the nearest station. There will need to be communications with the capsule, even though it is in a pretty good Faraday cage. There will need to be the right seats and restraints developed for the safety of the passengers. There will need to user experience elements developed for the sanity of the passengers while they are being whizzed at ultra high speed in windowless capsules. And then there are route rights, earthquake protection, dealing with containing cylinder distortions just because of the centimeter or so of drift induced along the route in the course of year just due to normal smooth deformations of our tectonic plates. And then there are pricing models, and getting insurance, and figuring out how that interacts with individual passenger insurance. Etc., etc.

There will need to be many, many new technologies and new designs developed for every aspect of the hyperloop. None of them will have existed before. None of them have been demonstrated, nor even enumerated as of today. It is going to take a long time to figure all these things out and build a stable system around them, and to do all the engineering needed on all the components. And it is going to be a hard psychological sell for passengers to ride in these windowless high speed systems, so even when all the technology challenges have been knocked down there will still be the challenge of pace of adoption.

So…while there might be some demonstration of some significance in the next 32 years I am confident in saying that there will be no commercial viable passenger carrying systems for hyperloop within that time frame.

I use this framework in trying to predict timing on various technological innovations. If something has not even been demonstrated yet in the lab, even though the physics says that it will be good to go, then I think it is a long, long way off. If it has been demonstrated in prototypes only, then it is still a long way off. If there are versions of it deployed at scale already, and most of what needs doing is evolutionary, then it may happen before too long. But then again, no-one may want to adopt it, so that will slow things down no matter how much enthusiasm there is by the technologists involved in developing it.

ABOUT THAT ADOPTION THING

Adoption of new things in technology takes much longer than one might expect. The original version of the Internet used 32 bit addressing, allowing only 4 billion unique address for all devices on the network, and using a protocol called IPv4, Internet Protocol version 4.  But by the early 1990’s it was recognized that with all the devices that would soon join the network (not just personal devices but so many other things like electricity meters, industrial sensors, traffic sensor and control, TVs, light switches(!), etc., etc.) the world would soon run out of address space.

By 1996 a new protocol,  IPv6, Internet Protocol version 6, had been defined, increasing the address space to 128 bits from 32 bits, allowing for 7.9\times 10^{28} more devices on the network.

Since 1996 there have been various goal dates specified for when all network traffic should use IPv6 rather than IPv4. In 2010 the target date was 2012. In 2014 fully 99% of all network traffic was still using IPv4 with many, many clever edge systems to cram much more than 4 billion devices into a 4 billion device address space. By the end of 2017 various categories of network traffic running on IPv6 ranged from under 2% to just over 20%. It is still a long way from full adoption of IPv6.

There were no technical things stopping the adoption of IPv6, in fact quite the opposite. As the number of devices that wanted to connect to the Internet grew there had to be many very clever innovations and work arounds in order to limp along with IPv4 rather than adopt IPv6.

Using my heuristics (rate of replacement of equipment, maturity of technical solutions, real need for what it provides, etc.) that I use to make my predictions in this post, I would have thought that IPv6 would have been universal by 2010 or so. I would have been wildly over optimistic about it.

AND ABOUT THAT “IT ALWAYS TAKES LONGER THAN YOU THINK” THINg

SpaceX first announced their Falcon Heavy rocket in April 2011, broke ground on their Vandenberg AFB, California, launch pad for it in June 2011, and expected a maiden flight in 2013. The rocket was first moved to a launch pad on December 28, 2017, at pad 39A at the Kennedy Space Center in Florida. It is now expected to fly in 2018. Development time has stretched from two years to seven years. So far.

It always takes longer than you think. It just does.

PREDICTIONS ABOUT SELF DRIVING CARS

The first three entries in the table below are about flying cars. I am pretty sure that practical flying cars will need to be largely self driving while flying, so they sort of fit the category. By flying cars I mean vehicles that can be driven anywhere a car can be driven. Otherwise it is not a car. And I mean that a person who does not have a pilots license, but does perhaps have a few hours of special training, can get into wearing normal clothing that would be appropriate to wear at an office, and is able to travel 100 miles, say, with much of the journey in the air. It should require no previous arrangement for the journey, no special filing of plans, nothing beyond using a maps like app on a smartphone in order to know the route to get to the destination. In other words, apart from a little extra training it should be just like an average person today using a conventional automobile to travel 100 miles.

Now let’s talk about self driving cars, or driverless cars. I wrote two blog posts early in 2017 about driverless cars. The first talked about unexpected consequences of driverless cars, in that pedestrians and other drivers will interact with them in different ways than they do with cars with drivers in them, and how the cars may bring out anti-social behavior in humans outside of them. It also pointed out  that owners of individual driverless cars may use them in new ways that they could never use a regular car, sometime succumbing to anti-social behavior themselves. The second post was about edge cases in urban environments where there are temporary signs that drivers must read, where on a regular basis it is impossible to drive according to the letter of the law, where mobility as a service will need to figure out how much control a passenger is allowed to have, and where police and tow truck drivers must interact with these cars, and the normal human to human interaction with drivers will no longer be present nor subjugatable by a position of authority.

For me it seems clear that driverless cars are not going to simply be the same sorts of cars as normal cars, but simply without human drivers. They are going to be fundamentally different beasts with different use modes, and different ways of fitting into the world.

Horseless carriages did not simply one for one replace horse drawn carriages. Instead they demanded a whole new infrastructure of paved roads, a completely new ownership model, a different utilization model, completely different fueling and maintenance procedures, a different rate of death for occupants, a different level of convenience, and ultimately they lead to a very different structure for cities as they enabled suburbia.

I think the popular interpretation is that driverless cars will simply replace cars with human drivers, one for one. I do not think that is going to happen at all. Instead our cities will be changed with special lanes for driverless cars, geo-fencing of where they can be and where cars driven by humans can be, a change in the norm for pick up and drop off location flexibility, changes to parking regulations, and in general all sorts of small incremental modifications to our cities.

But first let’s talk about the rate of adoption of driverless cars.

As I pointed out in my seven deadly sins post, in 1987 Ernst Dickmanns and his team at the Bundeswehr University in Munich had their autonomous van drive at 90 kilometers per hour (56mph) for 20 kilometers (12 miles) on a public freeway. Of course there were people inside the van but they had their hands off the controls. For the last 30 years researchers have been improving the ability of cars to drive on public roads, but it has mostly been about the driving, with very little about the interaction, the pick up and drop off of people, the interface with other services and restrictions, and with non-driving passengers inside the cars. All of these will be important.

From one point of view it has been slow, slow, slow incremental progress over the last thirty years, even though the work has been focused on only a small part of the problem. Just about a year ago I saw a tweet which I loved, which said something like “The customers knew that they had gotten a driverless Uber as there were two people in the front seat instead of just one.”. It is only just in the last few weeks that have started seeing actual unoccupied cars on public roads, from Waymo in Phoenix, Arizona. A tweet about this story referred to them as being the first “driverless driverless cars”…

But adoption is still a ways off. The price of the sensors still needs to come way down, and all the operational things about how the cars will be used and interface with passengers still needs to be worked out, let alone all the actual regulatory and liability environment under which they will operate needs to be put in place. Within some constraints, all these things will eventually be solved. But it is going to be much slower than many expect.

The true test of the viability of driverless cars will be when they are not just in testing or in demonstration, but when the owners of driverless taxis or ride sharing services or parking garages for end consumer self driving cars are actually making money at it. This will happen only gradually and in restricted geographies and markets to start with. My milestone predictions below are not about demonstrations, but about viable sustainable businesses. Without them the deployment of driverless cars will never really take off.

I think the under discussed reality of how driverless cars will get adopted is through geo fencing of where certain activities of those cars can take place, without any human driven cars in that vicinity. Furthermore applications of driverless cars will initially be restricted to certain cities and even areas within those cities, and perhaps even certain times of day and in certain weather conditions. It may be that for quite a while the cars for the first mobility as a service driverless cars (e.g., for Uber and Lyft like services) will only operate in a driverless mode some of the time, and at other times will need to have hired human drivers.

Prediction
[Self Driving Cars]
DateComments
A flying car can be purchased by any US resident if they have enough money.NET 2036There is a real possibility that this will not happen at all by 2050.
Flying cars reach 0.01% of US total cars.NET 2042That would be about 26,000 flying cars given today's total.
Flying cars reach 0.1% of US total cars.NIML
First dedicated lane where only cars in truly driverless mode are allowed on a public freeway.
NET 2021This is a bit like current day HOV lanes. My bet is the left most lane on 101 between SF and Silicon Valley (currently largely the domain of speeding Teslas in any case). People will have to have their hands on the wheel until the car is in the dedicated lane.
Such a dedicated lane where the cars communicate and drive with reduced spacing at higher speed than people are allowed to driveNET 2024
First driverless "taxi" service in a major US city, with dedicated pick up and drop off points, and restrictions on weather and time of day.NET 2022The pick up and drop off points will not be parking spots, but like bus stops they will be marked and restricted for that purpose only.
Such "taxi" services where the cars are also used with drivers at other times and with extended geography, in 10 major US citiesNET 2025A key predictor here is when the sensors get cheap enough that using the car with a driver and not using those sensors still makes economic sense.
Such "taxi" service as above in 50 of the 100 biggest US cities.NET 2028It will be a very slow start and roll out. The designated pick up and drop off points may be used by multiple vendors, with communication between them in order to schedule cars in and out.
Dedicated driverless package delivery vehicles in very restricted geographies of a major US city.NET 2023The geographies will have to be where the roads are wide enough for other drivers to get around stopped vehicles.
A (profitable) parking garage where certain brands of cars can be left and picked up at the entrance and they will go park themselves in a human free environment.NET 2023The economic incentive is much higher parking density, and it will require communication between the cars and the garage infrastructure.
A driverless "taxi" service in a major US city with arbitrary pick and drop off locations, even in a restricted geographical area.
NET 2032This is what Uber, Lyft, and conventional taxi services can do today.
Driverless taxi services operating on all streets in Cambridgeport, MA, on Greenwich Village, NY, NET 2035Unless parking and human drivers are banned from those areas before then.
A major city bans parking and cars with drivers from a non-trivial portion of a city so that driverless cars have free reign in that area.NET 2027
BY 2031
This will be the starting point for a turning of the tide towards driverless cars.
The majority of US cities have the majority of their downtown under such rules.NET 2045
Electric cars hit 30% of US car sales.NET 2027
Electric car sales in the US make up essentially 100% of the sales.NET 2038
Individually owned cars can go underground onto a pallet and be whisked underground to another location in a city at more than 100mph.NIMLThere might be some small demonstration projects, but they will be just that, not real, viable mass market services.
First time that a car equipped with some version of a solution for the trolley problem is involved in an accident where it is practically invoked.NIMLRecall that a variation of this was a key plot aspect in the movie "I, Robot", where a robot had rescued the Will Smith character after a car accident at the expense of letting a young girl die.
Predictions about ROBOTICS, AI and ML

Those of you who have been reading my series of blog posts on the future of Robotics and Artificial Intelligence know that I am more sanguine about how fast things will deploy at scale in the real world than many cheerleaders and fear mongers might believe. My predictions here are tempered by that sanguinity.

Some of these predictions are about the public perception of AI (that has been the single biggest thing that has changed in the field in the last three years), some are about technical ideas, and some are about deployments.

Prediction
[AI and ML]
DateComment
Academic rumblings about the limits of Deep LearningBY 2017Oh, this is already happening... the pace will pick up.
The technical press starts reporting about limits of Deep Learning, and limits of reinforcement learning of game play.BY 2018
The popular press starts having stories that the era of Deep Learning is over.BY 2020
VCs figure out that for an investment to pay off there needs to be something more than "X + Deep Learning".NET 2021I am being a little cynical here, and of course there will be no way to know when things change exactly.
Emergence of the generally agreed upon "next big thing" in AI beyond deep learning.NET 2023
BY 2027
Whatever this turns out to be, it will be something that someone is already working on, and there are already published papers about it. There will be many claims on this title earlier than 2023, but none of them will pan out.
The press, and researchers, generally mature beyond the so-called "Turing Test" and Asimov's three laws as valid measures of progress in AI and ML.NET 2022I wish, I really wish.
Dexterous robot hands generally available.NET 2030
BY 2040 (I hope!)
Despite some impressive lab demonstrations we have not actually seen any improvement in widely deployed robotic hands or end effectors in the last 40 years.
A robot that can navigate around just about any US home, with its steps, its clutter, its narrow pathways between furniture, etc.Lab demo: NET 2026
Expensive product: NET 2030
Affordable product: NET 2035
What is easy for humans is still very, very hard for robots.
A robot that can provide physical assistance to the elderly over multiple tasks (e.g., getting into and out of bed, washing, using the toilet, etc.) rather than just a point solution.NET 2028There may be point solution robots before that. But soon the houses of the elderly will be cluttered with too many robots.
A robot that can carry out the last 10 yards of delivery, getting from a vehicle into a house and putting the package inside the front door.Lab demo: NET 2025
Deployed systems: NET 2028
A conversational agent that both carries long term context, and does not easily fall into recognizable and repeated patterns.Lab demo: NET 2023
Deployed systems: 2025
Deployment platforms already exist (e.g., Google Home and Amazon Echo) so it will be a fast track from lab demo to wide spread deployment.
An AI system with an ongoing existence (no day is the repeat of another day as it currently is for all AI systems) at the level of a mouse.NET 2030I will need a whole new blog post to explain this...
A robot that seems as intelligent, as attentive, and as faithful, as a dog.NET 2048This is so much harder than most people imagine it to be--many think we are already there; I say we are not at all there.
A robot that has any real idea about its own existence, or the existence of humans in the way that a six year old understands humans.NIML

These predictions may seem a little random and disjointed. And they are. But that is the way progress is going to be made in Robotics, AI, and ML. There is not going to be a general intelligence that can suddenly do all sorts of things that humans (or chimpanzees) can do. It is going to be point solutions for a long, long time to come.

Building human level intelligence and human level physical capability is really, really hard. There has been a little tiny burst of progress over the last five years, and too many people think it is all done. In reality we are less than 1% of the way there, with no real intellectual ideas yet on how to get to 5%. And yes, I made up those percentages and can not really justify them. I may well have inflated them by a factor of 10 or more, and for that I apologize.

PREDICTIONS ABOUT SPACE TRAVEL

I have been a fan of spaceflight since my childhood, when every week my father would fly from Adelaide to Woomera, South Australia, to work on the first stage engines of a European satellite launch initiative know as Europa. Every couple of months I would go with him on a Friday evening to meetings of a club of enthusiasts where they would have the latest film footage from NASA which would be projected and discussed.

I decided back then that my life goal was to eventually live on another planet. So far my major progress towards that goal is to have not died on Earth before leaving. In my realistic moments I realize now that I may eventually fail at my goal.

So here are my predictions about space travel. Not as optimistic as wish I could be. But, realistic, I think.

Prediction
[Space]
DateComments
Next launch of people (test pilots/engineers) on a sub-orbital flight by a private company.BY 2018
A few handfuls of customers, paying for those flights.NET 2020
A regular sub weekly cadence of such flights.NET 2022
BY 2026
Regular paying customer orbital flights.NET 2027Russia offered paid flights to the ISS, but there were only 8 such flights (7 different tourists). They are now suspended indefinitely.
Next launch of people into orbit on a US booster.NET 2019
BY 2021
BY 2022 (2 different companies)
Current schedule says 2018.
Two paying customers go on a loop around the Moon, launch on Falcon Heavy.NET 2020The most recent prediction has been 4th quarter 2018. That is not going to happen.
Land cargo on Mars for humans to use at a later date
NET 2026SpaceX has said by 2022. I think 2026 is optimistic but it might be pushed to happen as a statement that it can be done, rather than for an pressing practical reason.
Humans on Mars make use of cargo previously landed there.NET 2032Sorry, it is just going to take longer than every one expects.
First "permanent" human colony on Mars.NET 2036It will be magical for the human race if this happens by then. It will truly inspire us all.
Point to point transport on Earth in an hour or so (using a BF rocket).NIMLThis will not happen without some major new breakthrough of which we currently have no inkling.
Regular service of Hyperloop between two cities.NIMLI can't help but be reminded of when Chuck Yeager described the Mercury program as "Spam in a can".


1AI and ML have been around for a long time already. I have been in pursuit of their magic for a long time. I have worked in both Artificial Intelligence and Machine Learning for over forty years. My 1977 Master’s thesis used Markov chains to prove the convergence of a particular machine learning algorithm. It was an abysmally terrible thesis.

AI/ML Is Not Uniquely Powerful Enough To Need Controlling

rodneybrooks.com/aiml-is-not-uniquely-powerful-enough-to-need-controlling/

Note: This short post is intended as a counterpoint to some claims that are being made about the need to control AI research. I don’t directly refer to those claims. You can figure it out. 

When humans next land on the Moon it will be with the help of many, many, Artificial Intelligence and Machine Learning systems.

Last time we got to the Moon and back without AI or ML.

I think this highlights the fact that current versions of AI and ML are just technologies. Different technologies can get to the same goal.

Some AI/ML researchers are making a bug fuss about how their work needs to be regulated as it is uniquely powerful. I disagree that it is uniquely powerful. Current day AI and ML is nothing like the intelligence or learning possessed by biological systems. They are both very narrow slices of the whole system. They are not particularly powerful.

Modern day Prometheuses rely on all sorts of technologies. Neither AI nor ML given them a particular leg up despite how exciting they might seem to current practitioners. It is the goal of a Prometheus that is important, not the particular technological tools that are used to achieve that goal.

Point 1: Swarms of killer drones could just as well be developed without any “AI”, using other technologies. We both got to the Moon, and had precise cruise missiles without any technologies that we would today call AI or ML1. We can develop “slaughterbots” without using anything that practitioners today would call AI or ML. So banning AI or ML in weapons systems will not change outcomes. It is futile. If you don’t like the sorts of things those weapons systems do, then work to ban the things they do, not the particular and very fungible technologies that are just one of many ways to produce that behavior.

Earlier this week, on December 18th, twitter user @ewschaetzle sent out a quote from H. P. Lovecraft from 1928, saying it “seems to capture the (misguided) fear that some have expressed toward .”:

The most merciful thing in the world, I think, is the inability of the human mind to correlate all its contents. We live on a placid island of ignorance in the midst of black sees of infinity. and it was not meant we should voyage far. The sciences, each straining in its own direction, have hitherto harmed us little; but some day the piecing together of dissociated knowledge will open up such terrifying vistas of reality, and of our frightful position therein, that we shall either go mad from the revelation or flee from the deadly light into the peace and darkness of a new dark age.

I have not found the full quote elsewhere but here is a partial version of it.

I like this quote a lot.

Three months ago in a long essay blog post (and in a better edited version in Technology Review) I pointed out seven common mistakes that people are making in predicting the future of AI, and by implication, the future of ML. In general they are vastly overestimating both its current power and how quickly it will develop.

Lovecraft’s words give a rationale for why this overestimation leads many other sensible, and even brilliant, entrepreneurs, physicists, and others to say that AI in general is incredibly dangerous and we must control its development. It is complex and they get scared.

Point 2: If one wants to legislate control of “AI research or development” in some way, then one must believe that those rules or laws will change at least one person’s behavior in some way. Without some change in behavior there is no point to legislation or rules, beyond smug self satisfaction the such laws or rules have been enacted. My question to those who say we should have these rules is: Show me one explicit change of behavior that you would like to see. Tell me who would have to do what differently than they currently are doing, and how that would impact the future. Tell me how it would make the world safer from AI.

So far I have not seen anyone suggest any explicit law or rule. All I have heard is “we must control it”. How? Let alone why?



1 Someone on twitter disagreed with my claim that we got to the Moon without ML by saying that Kalman filters, which were developed for navigation in the Apollo missions use Bayesian statistics, so therefore we did use ML to get to the Moon. That is a silly argument. ML today, and what ML refers to is much, much more than Kalman filters which were developed as state estimators, not as anything to do with learning from datasets. There is no pre-learned anything in using Kalman filters.

What If There Were Men On The Moon Today?

rodneybrooks.com/what-if-there-were-men-on-the-moon-today/

I was sitting on the beach looking at the full moon above. I looked through my binoculars to see more detail. And then it occurred to me that I could see on that surface every single location that humans had landed on the surface of a body in space that was not Earth. Six times, stretching from 48 years ago to 45 years ago.

That brought to mind this iconic photograph taken by Michael Collins during the Apollo 11 mission. He was alone in the Command Module, and visible in the foreground is the Lunar Lander with Buzz Aldrin and Neil Armstrong in it. In the background is Earth. This makes Michael Collins the only person who has ever lived who was not inside the frame of this photo. The only person not included.

This got me to thinking. What if, horribly, one of the six Lunar Modules that landed on the Moon with two astronauts on board, had failed to take off or failed to reach orbit for docking with the Command Module. If that had happened, every time we look at the Moon today we would see a grave site. It would be the most visible grave site in the world, visible from every place on the surface of the Earth.

How would that have changed the way we viewed mankind’s place in the Universe? Would we have seen exploration as a failure and something we should not do any more? Or would we have been inspired to try harder and not let 45 years pass without a return to space faring?

When we return to the Moon, or go to Mars, we will do it with more intent to stay longer than the six times we visited the Moon. Then we only landed for between 24 and 72 hours. When we spend a longer time, and with more people, on the surface of another body eventually there will be deaths. We saw that with the Shuttle program, and the Soviet Union had their own deaths in their programs. But if the exploration, and indeed settling, is permanent then I don’t think we will have the same empty feeling looking up at the Moon and Mars, as we would today if there had been an Apollo tragedy on the surface of the Moon.

When we go to the Moon, or Mars, the next time, let’s make it for real.

 

[FoR&AI] The Seven Deadly Sins of Predicting the Future of AI

rodneybrooks.com/the-seven-deadly-sins-of-predicting-the-future-of-ai/

[An essay in my series on the Future of Robotics and Artificial Intelligence.]

We are surrounded by hysteria about the future of Artificial Intelligence and Robotics. There is hysteria about how powerful they will become how quickly, and there is hysteria about what they will do to jobs.

As I write these words on September 2nd, 2017, I note just two news stories from the last 48 hours.

Yesterday, in the New York Times, Oren Etzioni, chief executive of the Allen Institute for Artificial Intelligence, wrote an opinion piece titled How to Regulate Artificial Intelligence where he does a good job of arguing against the hysteria that Artificial Intelligence is an existential threat to humanity. He proposes rather sensible ways of thinking about regulations for Artificial Intelligence deployment, rather than the chicken little “the sky is falling” calls for regulation of research and knowledge that we have seen from people who really, really, should know a little better.

Today, there is a story in Market Watch that robots will take half of today’s jobs in 10 to 20 years. It even has a graphic to prove the numbers.

The claims are ludicrous. [I try to maintain professional language, but sometimes…] For instance, it appears to say that we will go from 1 million grounds and maintenance workers in the US to only 50,000 in 10 to 20 years, because robots will take over those jobs. How many robots are currently operational in those jobs? ZERO. How many realistic demonstrations have  there been of robots working in this arena? ZERO. Similar stories apply to all the other job categories in this diagram where it is suggested that there will be massive disruptions of 90%, and even as much as 97%, in jobs that currently require physical presence at some particular job site.

Mistaken predictions lead to fear of things that are not going to happen. Why are people making mistakes in predictions about Artificial Intelligence and robotics, so that Oren Etzioni, I, and others, need to spend time pushing back on them?

Below I outline seven ways of thinking that lead to mistaken predictions about robotics and Artificial Intelligence. We find instances of these ways of thinking in many of the predictions about our AI future. I am going to first list the four such general topic areas of predictions that I notice, along with a brief assessment of where I think they currently stand.

A. Artificial General Intelligence. Research on AGI is an attempt to distinguish a thinking entity from current day AI technology such as Machine Learning. Here the idea is that we will build autonomous agents that operate much like beings in the world. This has always been my own motivation for working in robotics and AI, but the recent successes of AI are not at all like this.

Some people think that all AI is an instance of AGI, but as the word “general” would imply, AGI aspires to be much more general than current AI. Interpreting current AI as an instance of AGI makes it seem much more advanced and all encompassing that it really is.

Modern day AGI research is not doing at all well on being either general or getting to an independent entity with an ongoing existence. It mostly seems stuck on the same issues in reasoning and common sense that AI has had problems with for at least fifty years. Alternate areas such as Artificial Life, and Simulation of Adaptive Behavior did make some progress in getting full creatures in the eighties and nineties (these two areas and communities were where I spent my time during those years), but they have stalled.

My own opinion is that of course this is possible in principle. I would never have started working on Artificial Intelligence if I did not believe that. However perhaps we humans are just not smart enough to figure out how to do this–see my remarks on humility in my post on the current state of Artificial Intelligence suitable for deployment in robotics. Even if it is possible I  personally think we are far, far further away from understanding how to build AGI than many other pundits might say.

[Some people refer to “an AI”, as though all AI is about being an autonomous agent. I think that is confusing, and just as the natives of San Francisco do not refer to their city as “Frisco”, no serious researchers in AI refer to “an AI”.]

B. The Singularity. This refers to the idea that eventually an AI based intelligent entity, with goals and purposes, will be better at AI research than us humans are. Then, with an unending Moore’s law mixed in making computers faster and faster, Artificial Intelligence will take off by itself, and, as in speculative physics going through the singularity of a black hole, we have no idea what things will be like on the other side.

People who “believe” in the Singularity are happy to give post-Singularity AI incredible power, as what will happen afterwards is quite unpredictable. I put the word believe in scare quotes as belief in the singularity can often seem like a religious belief. For some it comes with an additional benefit of being able to upload their minds to an intelligent computer, and so get eternal life without the inconvenience of having to believe in a standard sort of supernatural God. The ever powerful technologically based AI is the new God for them. Techno religion!

Some people have very specific ideas about when the day of salvation will come–followers of one particular Singularity prophet believe that it will happen in the year 2029, as it has been written.

This particular error of prediction is very much driven by exponentialism, and I will address that as one of the seven common mistakes that people make.

Even if there is a lot of computer power around it does not mean we are close to having programs that can do research in Artificial Intelligence, and rewrite their own code to get better and better.

Here is where we are on programs that can understand computer code. We currently have no programs that can understand a one page program as well as a new student in computer science can understand such a program after just one month of taking their very first class in programming. That is a long way from AI systems being better at writing AI systems than humans are.

Here is where we are on simulating brains at the neural level, the other methodology that Singularity worshipers often refer to. For about thirty years we have known the full “wiring diagram” of the 302 neurons in the worm C. elegans, along with the 7,000 connections between them.  This has been incredibly useful for understanding how behavior and neurons are linked. But it has been a thirty years study with hundreds of people involved, all trying to understand just 302 neurons. And according to the OpenWorm project trying to simulate C. elegans bottom up, they are not yet half way there.  To simulate a human brain with 100 billion neurons and a vast number of connections is quite a way off. So if you are going to rely on the Singularity to upload yourself to a brain simulation I would try to hold off on dying for another couple of centuries.

Just in case I have not made my own position on the Singularity clear, I refer you to my comments in a regularly scheduled look at the event by the magazine IEEE Spectrum. Here is the the 2008 version, and in particular a chart of where the players stand and what they say. Here is the 2017 version, and in particular a set of boxes of where the players stand and what they say. And yes, I do admit to being a little snarky in 2017…

C. Misaligned Values. The third case is that the Artificial Intelligence based machines get really good at execution of tasks, so much so that they are super human at getting things done in a complex world. And they do not share human values and this leads to all sorts of problems.

I think there could be versions of this that are true–if I have recently bought an airline ticket to some city, suddenly all the web pages I browse that rely on advertisements for revenue start displaying ads for airline tickets to the same city. This is clearly dumb, but I don’t think it is a sign of super capable intelligence, rather it is a case of poorly designed evaluation functions in the algorithms that place advertisements.

But here is a quote from one of the proponents of this view (I will let him remain anonymous, as an act of generosity):

The well-known example of paper clips is a case in point: if the machine’s only goal is maximizing the number of paper clips, it may invent incredible technologies as it sets about converting all available mass in the reachable universe into paper clips; but its decisions are still just plain dumb.

Well, no. We would never get to a situation in any version of the real world where such a program could exist. One smart enough that it would be able to invent ways to subvert human society to achieve goals set for it by humans, without understanding the ways in which it was causing problems for those same humans. Thinking that technology might evolve this way is just plain dumb (nice turn of phrase…), and relies on making multiple errors among the seven that I discuss below.

This same author repeatedly (including in the piece from which I took this quote, but also at the big International Joint Conference on Artificial Intelligence (IJCAI) that was held just a couple of weeks ago in Melbourne, Australia) argues that we need research to come up with ways to mathematically prove that Artificial Intelligence systems have their goals aligned with humans.

I think this case C comes from researchers seeing an intellectually interesting research problem, and then throwing their well known voices promoting it as an urgent research question. Then AI hangers-on take it, run with it, and turn it into an existential problem for mankind.

By the way, I think mathematical provability is a vain hope. With multi-year large team efforts we can not prove that a 1,000 line program can not be breached by external hackers, so we certainly won’t be able to prove very much at all about large AI systems. The good news is that us humans were able to successfully co-exist with, and even use for our own purposes, horses, themselves autonomous agents with on going existences, desires, and super-human physical strength, for thousands of years. And we had not a single theorem about horses. Still don’t!

D. Really evil horrible nasty human-destroying Artificially Intelligent entities. The last category is like case C, but here the supposed Artificial Intelligence powered machines will take an active dislike to humans and decide to destroy them and get them out of the way.

This has been a popular fantasy in Hollywood since at least the late 1960’s with movies like 2001: A Space Odyssey (1968, but set in 2001), where the machine-wreaked havoc was confined to a single space ship, and Colossus: The Forbin Project (1970, and set in those times) where the havoc was at a planetary scale. The theme has continued over the years, and more recently with I, Robot (2004, set in 2035) where the evil AI computer VIKI takes over the world through the instrument of the new NS-5 humanoid robots. [By the way, that movie continues the bizarre convention from other science fiction movies that large complex machines are built with spaces that have multi hundred feet heights around them so that there can be great physical peril for the human heroes as they fight the good fight against the machines gone bad…]

This is even wronger than case C. I think it must make people feel tingly thinking about these terrible, terrible dangers…

In this blog, I am not going to address the issue of military killer robots–this often gets confused in the press with issue D above, and worse it often gets mashed together by people busy fear mongering about issue D. They are very separate issues. Furthermore I think that many of the arguments about such military robots are misguided. But it is a very different issue and will have to wait for another blog post.

Now, the seven mistakes I think people are making. All seven of them influence the assessments about timescales for and likelihood of each of scenarios A, B, C, and D, coming about. But some are more important I believe in the mis-estimations than others. I have labeled in the section headers for each of these seven errors where I think they do the most damage. The first one does some damage everywhere!

1. [A,B,C,D] Over and under estimating

Roy Amara was a futurist and the co-founder and President of the Institute For The Future in Palo Alto, home of Stanford University, countless venture capitalists, and the intellectual heart of Silicon Valley. He is best known for his adage, now referred to as Amara’s law:

We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.

There is actually a lot wrapped up in these 21 words which can easily fit into a tweet and allow room for attribution. An optimist can read it one way, and a pessimist can read it another. It should make the optimist somewhat pessimistic, and the pessimist somewhat optimistic, for a while at least, before each reverting to their norm.

A great example⁠1 of the two sides of Amara’s law that we have seen unfold over the last thirty years concerns the US Global Positioning System. Starting in 1978 a constellation of 24 satellites (30 including spares) were placed in orbit. A ground station that can see 4 of them at once can compute the the latitude, longitude, and height above a version of sea level. An operations center at Schriever Air Force Base in Colorado constantly monitors the precise orbits of the satellites and the accuracy of their onboard atomic clocks and uploads minor and continuous adjustments to them. If those updates were to stop GPS would fail to have you on the correct road as you drive around town after only a week or two, and would have you in the wrong town after a couple of months.

The goal of GPS was to allow precise placement of bombs by the US military. That was the expectation for it. The first operational use in that regard was in 1991 during Desert Storm, and it was promising. But during the nineties there was still much distrust of GPS as it was not delivering on its early promise, and it was not until the early 2000’s that its utility was generally accepted in the US military. It had a hard time delivering on its early expectations and the whole program was nearly cancelled again and again.

Today GPS is in the long term, and the ways it is used were unimagined when it was first placed in orbit. My Series 2 Apple Watch uses GPS while I am out running to record my location accurately enough to see which side of the street I ran along. The tiny size and tiny price of the receiver would have been incomprehensible to the early GPS engineers. GPS is now used for so many things that the designers never considered. It synchronizes physics experiments across the globe and is now an intimate component of synchronizing the US electrical grid and keeping it running, and it even allows the high frequency traders who really control the stock market to mostly not fall into disastrous timing errors. It is used by all our airplanes, large and small to navigate, it is used to track people out of jail on parole, and it determines  which seed variant will be planted in which part of many fields across the globe. It tracks our fleets of trucks and reports on driver performance, and the bouncing signals on the ground are used to determine how much moisture there is in the ground, and so determine irrigation schedules.

GPS started out with one goal but it was a hard slog to get it working as well as was originally expected. Now it has seeped into so many aspects of our lives that we would not just be lost if it went away, but we would be cold, hungry, and quite possibly dead.

We see a similar pattern with other technologies over the last thirty years. A big promise up front, disappointment, and then slowly growing confidence, beyond where the original expectations were aimed. This is true of the blockchain (Bitcoin was the first application), sequencing individual human genomes, solar power, wind power, and even home delivery of groceries.

Perhaps the most blatant example is that of computation itself. When the first commercial computers were deployed in the 1950’s there was widespread fear that they would take over all jobs (see the movie Desk Set from 1957). But for the next 30 years computers were something that had little direct impact on people’s lives and even in 1987 there were hardly any microprocessors in consumer devices. That has all changed in the second wave over the subsequent 30 years and now we all walk around with our bodies adorned with computers, our cars full of them, and they are all over our houses.

To see how the long term influence of computers has consistently been underestimated one need just go back and look at portrayals of them in old science fiction movies or TV shows about the future. The three hundred year hence space ship computer in the 1966 Star Trek (TOS) was laughable just thirty years later, let alone three centuries later.  And in Star Trek The Next Generation, and Star Trek Deep Space Nine, whose production spanned 1986 to 1999, large files still needed to be carried by hand around the far future space ship or space station as they could not be sent over the network (like an AOL network of the time). And the databases available for people to search were completely anemic with their future interfaces which were pre-Web in design.

Most technologies are overestimated in the short term. They are the shiny new thing. Artificial Intelligence has the distinction of having been the shiny new thing and being overestimated again and again, in the 1960’s, in the 1980’s, and I believe again now. (Some of the marketing messages from large companies on their AI offerings are truly delusional, and may have very bad blowback for them in the not too distant future.)

Not all technologies get underestimated in the long term, but that is most likely the case for AI. The question is how long is the long term. The next six errors that I talk about help explain how the timing for the long term is being grossly underestimated for the future of AI.

2. [B,C,D] Imagining Magic

When I was a teenager, Arthur C. Clarke was one of the “big three” science fiction writers along with Robert Heinlein and Isaac Asimov. But Clarke was more than just a science fiction writer. He was also an inventor, a science writer, and a futurist.

In February 1945 he wrote a letter⁠2 to Wireless World about the idea of geostationary satellites for research, and in October of that year he published a paper⁠3 outlining how they could be used to provide world-wide radio coverage. In 1948 he wrote a short story The Sentinel which provided the kernel idea for Stanley Kubrick’s epic AI movie 2001: A Space Odyssey, with Clarke authoring a book of the same name as the film was being made, explaining much that had left the movie audience somewhat lost.

In the period from 1962 to 1973 Clarke formulated three adages, which have come to be known as Clarke’s three laws (he said that Newton only had three, so three were enough for him too):

  1. When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.
  2. The only way of discovering the limits of the possible is to venture a little way past them into the impossible.
  3. Any sufficiently advanced technology is indistinguishable from magic.

Personally I should probably be wary of the second sentence in his first law, as I am much more conservative than some others about how quickly AI will be ascendant. But for now I want to expound on Clarke’s third law.

Imagine we had a time machine (powerful magic in itself…) and we could transport Issac Newton from the late 17th century to Trinity College Chapel in Cambridge University. That chapel was already 100 years old when he was there so perhaps it would not be too much of an immediate shock to find himself in it, not realizing the current date.

Now show Newton an Apple. Pull out an iPhone from your pocket, and turn it on so that the screen is glowing and full of icons and hand it to him. The person who revealed how white light is made from components of different colored light by pulling apart sunlight with a prism and then putting it back together again would no doubt be surprised at such a small object producing such vivid colors in the darkness of the chapel. Now play a movie of an English country scene, perhaps with some animals with which he would be familiar–nothing indicating the future in the content. Then play some church music with which he would be familiar. And then show him a web page with the 500 plus pages of his personally annotated copy of his masterpiece Principia, teaching him how to use the pinch gestures to zoom in on details.

Could Newton begin to explain how this small device did all that? Although he invented calculus and explained both optics and gravity, Newton was never able to sort out chemistry and alchemy. So I think he would be flummoxed, and unable to come up with even the barest coherent outline of what this device was. It would be no different to him than an embodiment of the occult–something which was of great interest to him when he was alive. For him it would be indistinguishable from magic. And remember, Newton was a really smart dude.

If something is magic it is hard to know the limitations it has. Suppose we further show Newton how it can illuminate the dark, how it can take photos and movies and record sound, how it can be used as a magnifying glass, and as a mirror. Then we show him how it can be used to carry out arithmetical computations at incredible speed and to many decimal places. And we show it counting his steps has he carries it.

What else might Newton conjecture that the device in front of him could do? Would he conjecture that he could use it to talk to people anywhere in the world immediately from right there in the chapel? Prisms work forever. Would he conjecture that the iPhone would work forever just as it is, neglecting to understand that it needed to be recharged (and recall that we nabbed him from a time 100 years before the birth of Michael Faraday, so the concept of electricity was not quite around)? If it can be a source of light without fire could it perhaps also transmute lead to gold?

This is a problem we all have with imagined future technology. If it is far enough away from the technology we have and understand today, then we do not know its limitations. It becomes indistinguishable from magic.

When a technology passes that magic line anything one says about it is no longer falsifiable, because it is magic.

This is a problem I regularly encounter when trying to debate with people about whether we should fear just plain AGI, let alone cases C or D from above. I am told that I do not understand how powerful it will be. That is not an argument. We have no idea whether it can even exist. All the evidence that I see says that we have no real idea yet how to build one. So its properties are completely unknown, so rhetorically it quickly becomes magical and super powerful. Without limit.

Nothing in the Universe is without limit. Not even magical future AI.

Watch out for arguments about future technology which is magical. It can never be refuted. It is a faith-based argument, not a scientific argument.

3. [A,B,C] Performance versus competence

One of the social skills that we all develop is an ability to estimate the capabilities of individual people with whom we interact. It is true that sometimes “out of tribe” issues tend to overwhelm and confuse our estimates, and such is the root of the perfidy of racism, sexism, classism, etc. In general, however, we use cues from how a person performs some particular task to estimate how well they might perform some different task. We are able to generalize from observing performance at one task to a guess at competence over a much bigger set of tasks. We understand intuitively how to generalize from the performance level of the person to their competence in related areas.

When in a foreign city we ask a stranger on the street for directions and they reply in the language we spoke to them with confidence and with directions that seem to make sense, we think it worth pushing our luck and asking them about what is the local system for paying when you want to take a bus somewhere in that city.

If our teenage child is able to configure their new game machine to talk to the household wifi we suspect that if sufficiently motivated they will be able to help us get our new tablet computer on to the same network.

If we notice that someone is able to drive a manual transmission car, we will be pretty confident that they will be able to drive one with an automatic transmission too. Though if the person is North American we might not expect it to work for the converse case.

If we ask an employee in a large hardware store where to find a particular item, a home electrical fitting say, that we are looking for and they send us to an aisle of garden tools, we will probably not go back and ask that very same person where to find a particular bathroom fixture. We will estimate that not only do they not know where the electrical fittings are, but that they really do not know the layout of goods within the store, and we will look for a different person to ask with our second question.

Now consider a case that is closer to some performances we see for some of today’s AI systems.

Suppose a person tells us that a particular photo is of people playing Frisbee in the park, then we naturally assume that they can answer questions like “what is the shape of a Frisbee?”, “roughly how far can a person throw a Frisbee?”, “can a person eat a Frisbee?”, “roughly how many people play Frisbee at once?”, “can a 3 month old person play Frisbee?”, “is today’s weather suitable for playing Frisbee?”; in contrast we would not expect a person from another culture who says they have no idea what is happening in the picture to be able to answer all those questions.  Today’s image labelling systems that routinely give correct labels, like “people playing Frisbee in a park” to online photos, have no chance of answering those questions.  Besides the fact that all they can do is label more images and can not answer questions at all, they have no idea what a person is, that parks are usually outside, that people have ages, that weather is anything more than how it makes a photo look, etc., etc.

This does not mean that these systems are useless however. They are of great value to search engine companies. Simply labelling images well lets the search engine bridge the gap from search for words to searching for images. Note too that search engines usually provide multiple answers to any query and let the person using the engine review the top few and decide which ones are actually relevant. Search engine companies strive to get the performance of their systems to get the best possible answer as one of the top five or so. But they rely on the cognitive abilities of the human user so that they do not have to get the best answer first, every time. If they only gave one answer, whether to a search for “great hotels in Paris”, or at an e-commerce site only gave one image selection for a “funky neck tie”, they would not be as useful as they are.

Here is what goes wrong. People hear that some robot or some AI system has performed some task. They then take the generalization from that performance to a general competence that a person performing that same task could be expected to have. And they apply that generalization to the robot or AI system.

Today’s robots and AI systems are incredibly narrow in what they can do. Human style generalizations just do not apply. People who do make these generalizations get things very, very wrong.

4. [A,B] Suitcase words

I spoke briefly about suitcase words (Marvin Minsky’s term4) in my post explaining how machine learning works. There I was discussing how the word learning can mean so many different types of learning when applied to humans. And as I said there, surely there are different mechanisms that humans use for different sorts of learning. Learning to use chopsticks is a very different experience from learning the tune of a new song. And learning to write code is a very different experience from learning your way around a particular city.

When people hear that Machine Learning is making great strides and they think about a machine learning in some new domain, they tend to use as a mental model the way in which a person would learn that new domain. However, Machine Learning is very brittle, and it requires lots of human preparation by researchers or engineers, special purpose coding for processing input data, special purpose sets of training data, and a custom learning structure for each new problem domain. Today’s Machine Learning by computers is not at all the sponge like learning that humans engage in, making rapid progress in a new domain without having to be surgically altered or purpose built.

Likewise when people hear that computers can now beat the world chess champion (in 1997) or the world Go champion (in 2016) they tend to think that it is “playing” the game just like a human would. Of course in reality those programs had no idea what a game actually was (again, see my post on machine learning), nor that they are playing. And as pointed out in this article in The Atlantic during the recent Go challenge the human player, Lee Sedol, was supported by 12 ounces of coffee, whereas the AI program, AlphaGo, was running on a whole bevy of machines as a distributed application, and was supported by a team of more than 100 scientists.

When a human plays a game a small change in rules does not throw them off–a good player can adapt. Not so for AlphaGo or Deep Blue, the program that beat Garry Kasparov back in 1997.

Suitcase words lead people astray in understanding how well machines are doing at tasks that people can do. AI researchers, on the other hand, and worse their institutional press offices, are eager to claim progress in their research in being an instance of what a suitcase word applies to for humans. The important phrase here is “an instance”. No matter how careful the researchers are, and unfortunately not all of them are so very careful, as soon as word of the research result gets to the press office and then out into the unwashed press, that detail soon gets lost. Headlines trumpet the suitcase word, and mis-set the general understanding of where AI is, and how close it is to accomplishing more.

And, we haven’t even gotten to saying many of Minsky’s suitcase words about AI systems; consciousness, experience, or thinking. For us humans it is hard to think about playing chess without being conscious, or having the experience of playing, or thinking about a move. So far, none of our AI systems have risen to an even elementary level where one of the many ways in which we use those words about humans apply. When we do, and I tend to think that we will, get to a point where we will start using some of those words about particular AI systems, the press, and most people, will over generalize again.

Even with a very narrow single aspect demonstration of one slice of these words I am afraid people will over generalize and think that machines are on the very door step of human-like capabilities in these aspects of being intelligent.

Words matter, but whenever we use a word to describe something about an AI system, where that can also be applied to humans, we find people overestimating what it means. So far most words that apply to humans when used for machines, are only a microscopically narrow conceit of what the word means when applied to humans.

Here are some of the verbs that have been applied to machines, and for which machines are totally unlike humans in their capabilities:

anticipate, beat, classify, describe, estimate, explain, hallucinate, hear, imagine, intend, learn, model, plan, play, recognize, read, reason, reflect, see, understand, walk, write

For all these words there have been research papers describing a narrow sliver of the rich meanings that these words imply when applied to humans. Unfortunately the use of these words suggests that there is much more there there than is there.

This leads people to misinterpret and then overestimate the capabilities of today’s Artificial Intelligence.

5. [A,B,B,B,…] Exponentials

Many people are suffering from a severe case of “exponentialism”.

Everyone has some idea about Moore’s Law, at least as much to sort of know that computers get better and better on a clockwork like schedule.

What Gordon Moore actually said was that the number of components that could fit on a microchip would double every year.  I published a blog post in February about this and how it is finally coming to an end after a solid fifty year run. Moore had made his predictions in 1965 with only four data points using this graph:

He only extrapolated for 10 years, but instead it has lasted 50 years, although the time constant for doubling has gradually lengthened from one year to over two years, and now it is finally coming to an end.

Double the components on a chip has lead to computers that keep getting twice as fast. And it has lead to memory chips that have four times as much memory every two years. It has also led to digital cameras which have had more and more resolution, and LCD screens with exponentially more pixels.

The reason Moore’s law worked is that it applied to a digital abstraction of true or false.  Was there an electrical charge or voltage there or not? And the answer to that yes/no question is the same when the number of electrons is halved, and halved again, and halved again. The answer remains consistent through all those halvings until we get down to so few electrons that quantum effects start to dominate, and that is where we are now with our silicon based chip technology.

Moore’s law, and exponential laws like Moore’s law can fail for three different reasons:

  1. It gets down to a physical limit where the process of halving/doubling no longer works.
  2. The market demand pull gets saturated so there is no longer an economic driver for the law to continue.
  3. It may not have been an exponential process in the first place.

When people are suffering from exponentialism they may gloss over any of these three reasons and think that the exponentials that they use to justify an argument are going to continue apace.

Moore’s Law is now faltering under case (a), but it has been the presence of Moore’s Law for fifty years that has powered the relentless innovation of the technology industry and the rise of Silicon Valley, and Venture Capital, and the ride of the geeks to be amongst the richest people in the world, that has led too many people to think that everything in technology, including AI, is exponential.

It is well understood that many cases of exponential processes are really part of an “S-curve”, where at some point the hyper growth flattens out. Exponential growth of the number of users of a social platform such as Facebook or Twitter must turn into an S-curve eventually as there are only a finite number of humans alive to be new users, and so exponential growth can not continue forever. This is an example of case (b) above.

But there is more to this. Sometimes just the demand from individual users can look like an exponential pull for a while, but then it gets saturated.

Back in the first part of this century when I was running a very large laboratory at M.I.T. (CSAIL) and needed to help raise research money for over 90 different research groups, I tried to show sponsors how things were continuing to change very rapidly through the memory increase on iPods. Unlike Gordon Moore I had five data points! The data was how much storage one got for one’s music in an iPod for about $400.  I noted the dates of new models and for five years in a row, somewhere in the June to September time frame a new model would appear.  Here are the data:

Year GigaBytes
2003 10
2004 20
2005 40
2006 80
2007 160

The data came out perfectly (Gregor Mendel would have been proud…) as an exponential. Then I would extrapolate a few years out and ask what we would do with all that memory in our pockets.

Extrapolating through to today we would expect a $400 iPod to have 160,000 GigaBytes of memory (or 160 TeraBytes). But the top of the line iPhone of today (which costs more than $400) only has 256 GigaBytes of memory, less than double the 2007 iPod, while the top of the line iPod (touch) has only 128 GigaBytes which ten years later is a decrease from the 2007 model.

This particular exponential collapsed very suddenly once the amount of memory got to the point where it was big enough to hold any reasonable person’s complete music library, in their hand. Exponentials can stop when the customers stop demanding.

Moving on, we have seen a sudden increase in performance of AI systems due to the success of deep learning, a form of Machine Learning. Many people seem to think that means that we will continue to have increases in AI performance of equal multiplying effect on a regular basis. But the deep learning success was thirty years in the making, and no one was able to predict it, nor saw it coming. It was an isolated event.

That does not mean that there will not be more isolated events, where backwaters of AI research suddenly fuel a rapid step increase in performance of many AI applications. But there is no “law” that says how often they will happen. There is no physical process, like halving the mass of material as in Moore’s Law, fueling the process of AI innovation. This is an example of case (c) from above.

So when you see exponential arguments as justification for what will happen with AI remember that not all so called exponentials are really exponentials in the first place, and those that are can collapse suddenly when a physical limit is hit, or there is no more economic impact to continue them.

6. [C,D] Hollywood scenarios

The plot for many Hollywood science fiction movies is that the world is just as it is today, except for one new twist. Certainly that is true for movies about aliens arriving on Earth. Everything is going along as usual, but then one day the aliens unexpectedly show up.

That sort of single change to the world makes logical sense for aliens but what about for a new technology? In real life lots of new technologies are always happening at the same time, more or less.

Sometimes there is a rational, within Hollywood reality, explanation for why there is a singular disruption of the fabric of humanity’s technological universe. The Terminator movies, for instance, had the super technology come from the future via time travel, so there was no need to have a build up to the super robot played by Arnold Schwarzenegger.

But in other movies it can seem a little silly.

In Bicentennial Man, Richard Martin, played by Sam Neill, sits down to breakfast being waited upon by a walking talking humanoid robot, played by Robin Williams. He picks up a newspaper to read over breakfast. A newspaper! Printed on paper. Not a tablet computer, not a podcast coming from an Amazon Echo like device, not a direct neural connection to the Internet.

In Blade Runner, as Tim Harford recently pointed out, detective Rick Deckard, played by Harrison Ford, wants to contact the robot Rachael, played by Sean Young. In the plot Rachael is essentially indistinguishable from a human being. How does Deckard connect to her? With a pay phone. With coins that you feed in to it. A technology that many of the readers of this blog may never had seen. (By the way, in that same post, Harford remarks: “Forecasting the future of technology has always been an entertaining but fruitless game.” A sensible insight.)

So there are two examples of Hollywood movies where the writers, directors, and producers, imagine a humanoid robot, able to see, hear, converse, and act in the world as a human–pretty much an AGI (Artificial General Intelligence). Never mind all the marvelous materials and mechanisms involved. But those creative people lack the imagination, or will, to consider how else the world may have changed as that amazing package of technology has been developed.

It turns out that many AI researchers and AI pundits, especially those pessimists who indulge in predictions C and D, are similarly imagination challenged.

Apart from the time scale for many C and D predictions being wrong, they ignore the fact that if we are able to eventually build such smart devices the world will have changed significantly from where we are. We will not suddenly be surprised by the existence of such super intelligences. They will evolve technologically over time, and our world will be different, populated by many other intelligences, and we will have lots of experience already.

For instance, in the case of D (evil super intelligences who want to get rid of us) long before we see such machines arising there will be the somewhat less intelligent and belligerent machines. Before that there will be the really grumpy machines. Before that the quite annoying machines. And before them the arrogant unpleasant machines.

We will change our world along the way, adjusting both the environment for new technologies and the new technologies themselves. I am not saying there may not be challenges. I am saying that they will not be as suddenly unexpected as many people think. Free running imagination about shock situations are not helpful–they will never be right, or even close.

“Hollywood scenarios” are a great rhetorical device for arguments, but they usually do not have any connection to future reality.

7. [B,C,D] SPEED OF Deployment

As the world has turned to software the deployment frequency of new versions has become very high in some industries. New features for platforms like Facebook are deployed almost hourly. For many new features, as long as they have passed integration testing, there is very little economic downside if a problem shows up in the field and the version needs to be pulled back–often I find that features I use on such platforms suddenly fail to work for an hour or so (this morning it was the pull down menu for Facebook notifications that was failing) and I assume these are deployment fails. For revenue producing components, like advertisement placement, more care is taken and changes may happen only on the scale of weeks.

This is a tempo that Silicon Valley and Web software developers have gotten used to. It works because the marginal cost of newly deploying code is very very close to zero.

Hardware on the other hand has significant marginal cost to deploy. We know that from our own lives. Many of the cars we are buying today, which are not self driving, and mostly are not software enabled, will likely still be on the road in the year 2040. This puts an inherent limit on how soon all our cars will be self driving. If we build a new home today, we can expect that it might be around for over 100 years. The building I live in was built in 1904 and it is not nearly the oldest building in my neighborhood.

Capital costs keep physical hardware around for a long time, even when there are high tech aspects to it, and even when it has an existential mission to play.

The US Air Force still flies the B-52H variant of the B-52 bomber. This version was introduced in 1961, making it 56 years old. The last one was built in 1963, a mere 54 years ago. Currently these planes are expected to keep flying until at least 2040, and perhaps longer–there is talk of extending their life out to 100 years. (cf. The Millennium Falcon!)

The US land-based Intercontinental Ballistic Missile (ICBM) force is all Minuteman-III variants, introduced in 1970. There are 450 of them. The launch system relies on eight inch floppy disk drives, and some of the digital communication for the launch procedure is carried out over analog wired phone lines.

I regularly see decades old equipment in factories around the world. I even see PCs running Windows 3.0 in factories–a software version released in 1990. The thinking is that “if it ain’t broke, don’t fix it”. Those PCs and their software have been running the same application doing the same task reliably for over two decades.

The principal control mechanisms in factories, including brand new ones in the US, Europe, Japan, Korea, and China, is based on Programmable Logic Controllers, or PLCs. These were introduced in 1968 to replace electromechanical relays. The “coil” is still the principal abstraction unit used today, and the way PLCs are programmed is as though they were a network of 24 volt electromechanical relays. Still.  Some of the direct wires have been replaced by Ethernet cables. They emulate older networks (themselves a big step up) based on the RS485 eight bit serial character protocol, which themselves carry information emulating 24 volt DC current switching.  And the Ethernet cables are not part of an open network, but instead individual cables are run point to point physically embodying the control flow in these brand new ancient automation controllers. When you want to change information flow, or control flow, in most factories around the world it takes weeks of consultants figuring out what is there, designing new reconfigurations, and then teams of tradespeople rewiring and reconfiguring hardware. One of the major manufacturers of this equipment recently told me that they aim for three software upgrades every twenty years.

In principle it could be done differently. In practice it is not. And I am not talking about just in technological backwaters. I just this minute looked on a jobs list and even today, this very day, Tesla is trying to hire full time PLC technicians at their Fremont factory. Electromagnetic relay emulation to automate the production of the most AI-software advanced automobile that exists.

A lot of AI researchers and pundits imagine that the world is already digital, and that simply introducing new AI systems will immediately trickle down to operational changes in the field, in the supply chain, on the factory floor, in the design of products.

Nothing could be further from the truth.

The impedance to reconfiguration in automation is shockingly mind-blowingly impervious to flexibility.

You can not give away a good idea in this field. It is really slow to change. The example of the AI system making paper clips deciding to co-opt all sorts of resources to manufacture more and more paper clips at the cost of other human needs is indeed a nutty fantasy. There will be people in the loop worrying about physical wiring for decades to come.

Almost all innovations in Robotics and AI take far, far, longer to get to be really widely deployed than people in the field and outside the field imagine. Self driving cars are an example. Suddenly everyone is aware of them and thinks they will soon be deployed. But it takes longer than imagined. It takes decades, not years. And if you think that is pessimistic you should be aware that it is already provably three decades from first on road demonstrations and we have no deployment. In 1987 Ernst Dickmanns and his team at the Bundeswehr University in Munich had their autonomous van drive at 90 kilometers per hour (56mph) for 20 kilometers (12 miles) on a public freeway. In July 1995 the first no hands on steering wheel, no feet on pedals, minivan from CMU’s team lead by Chuck Thorpe and Takeo Kanade drove coast to coast across the United States on public roads.  Google/Waymo has been working on self driving cars for eight years and there is still no path identified for large scale deployment. It could well be four or five or six decades from 1987 before we have real deployment of self driving cars.

New ideas in robotics and AI take a long long time to become real and deployed.

Epilog

When you see pundits warn about the forthcoming wonders or terrors of robotics and Artificial Intelligence I recommend carefully evaluating their arguments against these seven pitfalls. In my experience one can always find two or three or four of these problems with their arguments.

Predicting the future is really hard, especially ahead of time.



Pinpoint: How GPS is Changing Technology, Culture, and Our Minds, Greg Milner, W. W. Norton, 2016.

2 “V2s for Ionosphere Research?”, A. C. Clarke, Wireless World, p. 45, February, 1945.

3 “Extra-Terrestrial Relays: Can Rocket Stations Give World-wide Radio Coverage”, Arthur C. Clarke, Wireless World, 305–308, October, 1945.

The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind, Marvin Minsky, Simon and Schuster, 2006.

[FoR&AI] Machine Learning Explained

rodneybrooks.com/forai-machine-learning-explained/

[An essay in my series on the Future of Robotics and Artificial Intelligence.]

Much of the recent enthusiasm about Artificial Intelligence is based on the spectacular recent successes of machine learning, itself often capitalized as Machine Learning, and often referred to as ML. It has become common in the technology world that the presence of ML in a company, in a development process, or in a product is viewed as a certification of technical superiority, something that will outstrip all competition.

Machine Learning is what has enabled the new assistants in our houses such as the Amazon Echo (Alexa) and Google Home by allowing them to reliably understand as we speak to them. Machine Learning is how Google chooses what advertisements to place, how it saves enormous amounts of electricity at its data centers, and how it labels images so that we can search for them with key words. Machine learning is how DeepMind (a Google company) was able to build a program called Alpha Go which beat the world Go champion. Machine Learning is how Amazon knows what recommendations to make to you whenever you are at its web site. Machine Learning is how PayPal detects fraudulent transactions. Machine Learning is how Facebook is able to translate between languages. And the list goes on!

While ML has started to have an impact on many aspects of our life, and will more and more so over the coming decades, some sobriety is not out of place. Machine Learning⁠1 is not magic. Neither AI programs, nor robots, wander around in the world ready to learn about whatever there is around them.

Every successful application of ML is hard won by researchers or engineers carefully analyzing the problem that is at hand. They select one or many different ML algorithms, and custom design how to connect them together and to the data. In some cases there is an extensive period of training on very large sets of data before the algorithm can be run on the problem that is being solved. In that case there may be months of work to do in collecting the right sort of data from which ML will actually learn. In other cases the learning algorithm will be integrated in to the application and will learn while doing the task that is desired–it might require some training wheels in the early stages, and they too must be designed. In any case there is always a big design project about how, when the ultimate system is operational, the data that comes in will be organized, processed and mapped before it reaches the ML component of the system.

When we are tending plants we pour water on them and perhaps give them some fertilizer and they grow. I think many people in the press, in management, and in the non-technical world have been dazzled by the success of Machine Learning, and have come to think of it a little like water or fertilizer for hard problems. They often mistakenly believe that a generic version will work on any and all problems. But while ML can sometimes have miraculous results it needs to be carefully customized after the DNA of the problem has beed analyzed.  And even then it might not be what is needed–to extend the metaphor, perhaps it is the climate that needs to be adjusted and no amount of fertilizer or ML will do the job.

How does Machine Learning work, and is it the same as when a child or adult learns something new? The examples above certainly seem to cover some of the same sort of territory, learning how to understand a human speaking, learning how to play a game, learning to name objects based on their appearance.

Machine Learning started with games

In the early 1940’s as war was being waged world wide there were only a handful of electronic digital computers in existence. They had been built, using the technology of vacuum tubes, to calculate gunnery tables and to decrypt coded military communications of the enemy.  Even then, however, people were starting to think about how these computers might be used to carry out intelligent activities, fifteen years before the term Artificial Intelligence was first floated by John McCarthy.

Alan Turing, who in 1936 had written the seminal paper that established the foundations of modern computation, and Donald Michie, a classics student from Oxford (later he would earn a doctorate in genetics), worked together at Bletchley Park, the famous UK code breaking establishment that Churchill credited with subtracting years from the war. Turing contributed to the design of the Colossus computer there, and through a key programming breakthrough that Michie made, the design of the second version of the Colossus was changed to accommodate his ideas even better. Meanwhile at the local pub the pair had a weekly chess game together and discussed how to program a computer to play chess, but they were only able to get as far as simulations with pen and paper.

In the United States right after the war, Arthur Samuel3, an expert on vacuum tubes was the leader of an effort to built the ILLIAC computer at the University of Illinois at Urbana-Champaign. While the computer was still being built he planned out how to program it to play checkers (or draughts in British English), but left in 1949 to join IBM before the University computer was completed. At IBM he worked on both vacuum tubes and transistors to bring IBM’s first commercial general purpose digital computers to market. On the side he was able to implement a program that by 1952 could play checkers against a human opponent. This was one of the first non-arithmetical programs to run on general purpose digital computers, and has been called the first AI program to run in the United States.

Samuel continued to improve the program over time and in 1956 it was first demonstrated to the public. But Samuel wondered whether the improvements he was making to the program by hand could be made by the machine itself. In 1959 he published a paper titled “Some Studies in Machine Learning Using the Game of Checkers”⁠2, the first time the phrase “Machine Learning” was used–earlier there had been models of learning machines, but this was a more general concept.

The first sentence in his paper was: “The studies reported here have been concerned with programming of a digital computer to behave in a way which, if done by human beings or animals, would be described as involving the process of learning.” Right there is his justification for using the term learning, and while I would not quibble with it, I think that it may have had some unintended consequences which we will explore towards the end of this post.

What Samuel had realized, demonstrated, and exploited, was that digital computers were by 1959 fast enough to take over some of the fine tuning that a person might do for a program, as he had been doing since the first version of his program in 1952, and ultimately eliminate the need for much of that effort by human programmers by letting the computer do some Machine Learning on appropriate parts of the problem. This is exactly what has lead, almost 60 years later to the great influence that ML is now having on the world.

One of the two learning techniques Samuel described was something he called rote learning, and today would be labelled as a well known programming technique called memoization4, and sped up the program. The other learning technique that he investigated involved adjusting numerical weights on how much the program should believe each of over thirty measures of how good or bad a particular board position was for the program or its human opponent. This is closer in spirit to techniques in modern ML. By improving this measure the program could get better and better at playing. By 1961 his program had beat the Connecticut state checker champion. Another first for AI, and enabled by the first ML program.

Arthur Samuel built his AI and ML systems not as an academic researcher but as a scholar working on his own time apart from his day job. However he had an incredible advantage over all the AI academic researchers. Whereas access to computers was rare and precious for them, Samuel’s day job was as a key participant building the first mass produced digital computers, and each one needed to be run for many hours to catch early life defects before it could be shipped. He had a surfeit of free computer time. Just about no one else in the world had such a luxurious computational environment.

Sometimes the less lucky academics had to resort to desperate measures. And so it was for Donald Michie, colleague of Alan Turing back at Bletchley Park. By 1960 he was a Senior Lecturer in Surgical Science at the University of Edinburgh, but his real interests lay in Artificial Intelligence, though he always preferred the term Machine Intelligence.

In 1960 Surgical Science did not have much pull in getting access to a digital computer. So Donald Michie himself built a machine that could learn to play the game of tic-tac-toe (Noughts and Crosses in British English) from 304 matchboxes, small rectangular boxes which were the containers for matches, and which had an outer cover and a sliding inner box to hold the matches. He put a label on one end of each of these sliding boxes, and carefully filled them with precise numbers of colored beads. With the help of a human operator, mindlessly following some simple rules, he had a machine that could not only play tic-tac-toe but could learn to get better at it.

He called his machine MENACE, for Matchbox Educable Noughts And Crosses Engine, and published⁠5 a report on it in 1961. In 1962 Martin Gardner⁠6 reported on it in his regular Mathematical Games column in Scientific American, but illustrated it with a slightly simpler version to play hexapawn, three chess pawns against three chess pawns on a three by three chessboard. This was a way to explain Machine Learning and provide an experimental vehicle to the scientifically interested lay population, who certainly would not have had access to a digital computer at that time. Gardner suggested that people try building a matchbox computer to play simplified checkers with two pieces for each player on a four by four board. But he felt that even the simplest version of chess that he could come up with, on a five by five board would require too many matchboxes to be practical.

I first read about the matchbox computer in 1967 in a book⁠7 published the previous year which was written by a group of teachers at a British high school. They neither attributed the idea to Michie, nor the game they described it learning, hexapawn, to Gardner. As a barely teenager who had to hand build every machine for every experiment I wanted to do in AI, I must admit I thought that the matchbox computer was too simple a project and so did not pursue it. Now, however, I have come to realize that it is the perfect way of introducing how Machine Learning works, as everything is there to see. Even though MENACE is over fifty years old many of the problems that it faced are still relevant to machine learning algorithms today, and it shares many characteristics with almost all of today’s machine learning. Due to its simplicity it can be described in complete detail and no mathematics is needed to get a strong intuitive understanding of how it works.

Today people generally recognize three different classes of Machine Learning, supervised, unsupervised, and reinforcement learning, all three very actively researched, and all being used for real applications.  Donald Michie’s MENACE introduced the idea of ML reinforcement learning, and he explicitly refers to reinforcement as a key concept in how it works.

How a collection of matchboxes plays & learns

I am going to take the details I give here from a retelling of how MENACE worked from a more accessible 1963 paper⁠8. In a very often republished picture from that paper Donald Michie (or at least his hands) can be seen both playing tic-tac-toe against the machine, and operating the machine.

On the sheet of paper in from of him you can just see a large tic-tac-toe diagram. There are stacks of matchboxes toward the rear of the table, must likely glued together so that they stay in place. Some of the matchboxes have their drawers partially pulled out, and he is holding one of the drawers in his left hand. As we will see this image captures most of what is going on as MENACE learns to play better and better tic-tac-toe.

To make the description more clear I am going to introduce a second person; let’s call him Alan. Alan will operate the matchbox machine according to fixed rules, and will not have to make any decisions that are not determined completely by those rules. Donald, the human player will not touch the machine, but instead will write out the moves on a standard three by three grid, accepting the moves of MENACE as delivered to him by Alan, playing his own moves, and being the adjudicator of when he or MENACE has won by getting three in a row.

Michie had MENACE always play first, with a ‘O’ and so we will do that here also. Below is what a game might look like, starting with an empty board, MENACE playing first in the middle of the top row, Donald replying, with an ‘X’ in the top right corner as his first play to be made, and back and forth. Notice that I am using a period, or ‘.’, for a blank, so that I don’t have to draw the customary horizontal and vertical lines to divide the squares. I have put a little indicator under each board position where it is MENACE’s turn to play. At MENACE’s third move it blocks an immediate win where Donald would be able to complete the diagonal from the upper right to the lower left, but Donald replies with a move to the bottom right corner, threatening now two possible three in a rows on his next turn, and MENACE is able to block only one of them so MENACE loses to Donald.

...   .O.   .OX   .OX   .OX   .OX   .OX   OOX   OOX   
...   ...   ...   ...   .X.   .X.   .X.   .X.   .XX   
...   ...   ...   .O.   .O.   OO.   OOX   OOX   OOX 
 ^           ^           ^           ^

The way that MENACE plays is that there is a matchbox for every possible board configuration that could arise in the course of the game when it is MENACE’s turn. There is a box for every configuration where it is MENACE’s first, second, third, or fourth turn, but not for its fifth turn as it has no choice to make there as there will only be one empty square.

The configurations are drawn on a small label pasted to the front of each individual drawer. When it is MENACE’s turn, Alan finds the matchbox with a label that matches the current state of play on the piece of paper on which Donald is keeping track of the game. He opens the drawer which has some number of colored beads in it. Without looking he randomly picks one of the beads. Importantly he leaves the drawer open and after showing the bead to Donald he puts it on the table in front of the open drawer from which it came. The boxes are arranged left to right corresponding to less moves played and then more moves played so it is easy to keep tack of which bead came from which drawer. There are nine colors of bead, and each color corresponds to one of the nine squares in tic-tac-toe. After seeing the bead Donald writes down an ‘O’ at the appropriate square on his piece of paper, and then writes his own ‘X’ as his next move, and the cycle repeats.

Although the actual colors of the beads do not really matter, here are the colors and their correspondence to the squares that were used in the original experiments (this time I have put in the horizontal vertical lines to divide the color words).

 white | lilac | silver
-----------------------
 black | gold  | green 
-----------------------
 amber |  red  | pink  

How many beads are there in each box?  For all the first move boxes, of which there is one, corresponding to the empty board, there are four beads for each possible move, so 36 in total.  For all possible second moves for MENACE there are only seven possibilities, and each of those empty squares has three beads.  For the third move there are 2 beads for each of the five possibilities, and for each fourth move there is one beed for each of the three possibilities.

To start out MENACE is playing each move completely at random. But suppose MENACE loses a game. Then Alan discards the beads below each open drawer and closes them all. This is negative reinforcement as MENACE lost, and so made moves it should not make in the future. In particular its fourth move, with only one bead, led to a loss that was at that point completely out of its control. So removing that bead means that MENACE will never play that bad move again. Its third move was perhaps a little suspect so that goes down to only one bead instead of two and it is less likely to try that again, but if it does it will not be tricked in exactly the same way again.

If MENACE draws the game then it gets positive reinforcement as each bead that was picked from each drawer is put back in, along with an extra bonus bead of the same color. If it won the game then it gets three additional beads same colored beads along with the one played at each turn. In this way MENACE tends to do the things that worked in the past, but if the opponent (in this case Donald) finds a new way to win against what used to work then MENACE will gradually adapt to that and avoid that losing line of play.

That is it. With Alan following this simple set of rules MENACE learns to play better and better over time. But there is one point of practicality.

Human structuring of the learning problem

As I described it above there are a lot of matchboxes needed for MENACE. There is 1 for MENACE’s first move, 72 for its second move, 756 for its third, and 1372 for its fourth move, for a total of 2201 matchboxes.

But let’s look at another possible game, where every position is different from the previous game we looked at.

...   ...   ...   ...   ...   O..   O..   O.O   O.O   
...   ..O   ..O   O.O   OXO   OXO   OXO   OXO   OXO   
...   ...   ..X   ..X   ..X   ..X   X.X   X.X   XXX
 ^           ^           ^           ^

But wait, this is really the same game as before just rotated ninety degrees clockwise. It is going to take a lot longer to learn how to play tic-tac-toe if MENACE has to independently learn that its second move in this game is just as bad as its second move in the previous game. To make MENACE learn faster, and to reduce the number of matchboxes down to a more manageable level, Donald Michie took into account that up to eight different patterns of Noughts and Crosses might really be essentially the same.

Here is an example where an original board positions is rotated clockwise by a quarter, half, and three-quarter turn, and where a reflection about the vertical axis, the horizontal axis, and the two diagonal axes all give different board positions Nonetheless these eight positions are essentially the same as far as the rules of the game of tic-tac-toe are concerned.

OX.   ..O   ...   ...   .XO   ...   O..   ...
.O.   .OX   .O.   XO.   .O.   .O.   XO.   .OX
...   ...   .XO   O..   ...   OX.   ...   ..O

Some board positions may not result in so many different looking positions when rotated or reflected. For instance, a single play in the center of the board is not changed at all by these spatial transformations.

In any case, by considering all the rotations and reflections of a board position as a single position, and therefore only assigning one matchbox to all of them combined, the requirements for MENACE are reduced to 1 matchbox, as before, for MENACE’s first move, 12 for its second, 108 for its third, and 183 for its fourth move, bringing the total⁠9 to 304. Furthermore by looking at the symmetries in what move is played there are often less essentially different moves than there are empty squares. For instance in both these cases MENACE is about to play an ‘O’:

...   .O.
...   .O.
...   X.X

In each case there are only three essentially different moves that can be played, so MENACE’s matchboxes for these move need only start out with three different colored beads rather than nine or five respectively.

By taking into account these symmetries the MENACE machine can be much smaller, and the speed of learning will be much faster, as lessons from one symmetric position will be automatically learned at another. The cost is that Alan is going to have to do quite a bit more work to operate MENACE. Now he will have to look at the position that Donald shows on the piece of paper where Donald is playing and not just look for an identical label on the front of a matchbox, but look for one that might be a rotation or reflection of the state of the game. Then, when Alan randomly selects a bead which indicates a particular move on the label on the matchbox from which it game, he will have to figure out which square that corresponds to on Donald’s sheet of paper through the inverse rotation or reflection that he used to select the matchbox. Fortunately this extra work is all quite deterministic and Alan is still following a strict set of rules with no room for judgement to creep in. We will come back to this a little later and mechanize Alan’s tasks though a few sheets of very simple instructions that will do all this extra work.

How well do matchboxes learn?

MENACE is learning what move it should choose in one of 304 essentially different board configurations for its first four moves in a game of tic-tac-toe. Since Alan randomly picks out one bead from the matchbox corresponding to one of those configurations it is making a random move from a small number of moves but the probability of a particular move goes up when there are more beads of a particular color from positive reinforcements from previous games, and the probability of a move which leads to a loss goes down relative to the other possible moves as its beads are removed.

Look back at the  two examples just above for a first move and a third move for MENACE. The empty board starts out with 12 beads, for each of three different colors representing placing an ‘O’ on a corner, in the middle of an edge, or in the middle of the board. The board waiting for MENACE’s third ‘O’ to be played starts out with just six beads, of three different colors, corresponding to playing the blocking move between the ‘X’s, one of the corners, or one of the other two middle edges. We will refer to number of beads of the same color in a single box as a parameter. By mapping all symmetric situations to a common matchbox and restricting the different moves to essentially different moves, there are 1087 parameters that MENACE adjusts over time through the removal or addition of beads. When MENACE starts off it has a total of 1720 beads representing those 1087 parameters in 304 different matchboxes.

When MENACE starts out it is playing uniformly randomly over all essentially different moves. If two uniformly random players play against each other, the first to play wins 59% of the time, draws 13%, and loses 28%. This shows the inherent bias in the game for the first player, which makes learning a little easier for MENACE.

In his original paper Michie reported that MENACE became quite a good player after only 220 games and was winning most games, but neither I nor others who report simulating MENACE (you can find many with a web search) saw MENACE doing that well at all. In fact since a perfect player never loses at tic-tac-toe then two perfect players always draw the game 100% of the time. It seems likely that Michie was carefully training MENACE with deliberately chosen games, and then playing against it in a fairly random way. He alludes to this when he later converted to a computer simulation of MENACE and mentions that playing against random moves results in much slower learning than playing against a deliberate policy.

To explore this I made a computer simulation of MENACE and three different simulated strategies of Donald playing against MENACE. I let learning proceed for 4,000 games, and did this multiple times against each of the three simulated players. Since there is randomness in picking a bead from a matchbox, the random number generator used by the computer to simulate this ensures that different trials of 4,000 games will lead to different actual games being played. Every so often I turned off the learning and tested how well⁠10 MENACE was currently playing against the three simulated players, including the two it was currently not learning from.

The three simulated players were as follows. Player A played completely randomly at all times. Player B played optimally and was unbeatable. Player C was optimal except that 25% of its moves were random instead of the optimal play. These are the three different versions of Donald that I used in my simulations.

In the table below the first row shows the performance of MENACE before it has learned at all, against each of the three simulated players. Each triple of numbers is the percentage of wins, draws, and losses (these are rounded percentages from a very large number of test games so don’t necessarily add to exactly 100%). As expected it never wins against Player B which plays optimally and can not be beaten.  Player C which makes mistakes 25% of the time can be beaten, but only about a quarter of the time.  In each row below that, MENACE was trained from scratch playing 4,000 games against a different one of these players. In each column we show, with MENACE stopped from further learning and adjusting its parameters, how it typically did against each of the three players once trained. We say “typically” as there is some variation in the resulting percentages between different trials with the same conditions, but only by a few points, and not in all cases.

\   compete |             |             |             |
 \  against | Player A    | Player B    | Player C    |
  \------\  |             |             |             |
trained   \ |             |             |             |
against    \|             |             |             |
------------|-------------|-------------|-------------|
no training |  59/ 13/ 28 |   0/ 24/ 76 |  27/ 19/ 53 |
============|=============|=============|=============|
Player A    |  86/  8/  6 |   0/ 28/ 72 |  50/ 20/ 30 |
------------|-------------|-------------|-------------|
Player B    |  71/ 15/ 14 |   0/100/  0 |  38/ 48/ 14 |
------------|-------------|-------------|-------------|
Player C    |  90/  8/  2 |   0/ 99/  1 |  56/ 42/  2 |
------------|-------------|-------------|-------------|

The first thing to notice is that how MENACE plays really does depend on what player it was trained against. When it is trained against Player B, which always plays optimally, it very quickly, usually after only about 200 games, learns to always play to a draw. But with that training (look in the same row) it is really not very good at playing against Player C which plays optimally with a 25% error rate. That is probably because in its training it never got to win at all against Player B, so it has not learned any winning moves to use against Player C.

When MENACE is trained against Player A (look in the row labelled Player A), which plays completely randomly it does learn to play against it quite well, and it also does reasonably well against Player C, probably because it has accidentally won enough times during training to have boosted some winning moves when they are available. It does dismally against the optimal Player B however. This particular box in the table has the highest variance of all in the table. Sometimes after 4,000 games it is doing less than half as well against Player B than when it started out learning.

When MENACE trains against Player C it does the best overall. It sees enough losses early on that after about 400 games it is starting to get good at avoiding losses, though it is still slowly, slowly getting better at that aspect of its game even after 4,000 games of learning. It usually doesn’t get quite as good as Player B, and very occasionally still loses to it, but it is really good at winning when there are opportunities for it to do so. We can see that against Player C it as learned to take advantage of its mistakes to drive home a win.

While not as good as a person, MENACE does get better against different types of players. It does however end up tuning its game to the type of player it is playing against.

There is also something surprising about the number of beads. MENACE starts off with 1,720 beads, but depending on which of Players A, B, or C, it is learning from it has from 2,300 to 3,200 beads after just 200 games, and always there is at least one parameter with over one hundred beads representing it by that time.  By 4,000 games it may have more than 35,000 beads representing just 1,087 parameters, with as many as 6,500 beads for one of the parameters.  This seems unnecessary, and perhaps the impact of rewarding all the moves with three beads on a win. However when I changed my simulation to never add more beads to a parameter that already had at least one hundred beads, a practical limit perhaps for a MENACE machine built from physical matchboxes, it tended to slow down learning in most cases represented in the table above, and even had small drops in typical levels of play even after 4,000 games of experience when playing against Players A and C.

Note that besides eliminating ever taking the very last bead away from a matchbox after a loss, this is the only place where I deviated from Michie’s description of his MENACE. Since he chose his plays carefully to instruct MENACE, and since he only played 220 games by hand, he perhaps did not come across the phenomenon of large numbers of beads.

Mechanizing MENACE a little more

In preparation for comparing how MENACE learns to how a person learns I want to make the role of Alan, the human operator of MENACE, a little clearer. In the description derived from Michie’s original paper, Michie himself played the role of both Donald and Alan. In my description above I talked about Alan matching the image of the paper on which Donald was playing to the labels on the matchboxes, possibly having to rotate or reflect the game board. And after randomly selecting a bead from that box, Alan would need to figure out which square that applied to on Donald’s piece of paper.

That sounds a little fuzzy, and perhaps requiring some reasoning on Alan’s part, so now we’ll make explicit a very rule driven approach that we could enforce, to ensure that Alan’s role is completely rote, with no judgement at all required.

We will make the communication between Donald and Alan very simple. Donald will hand Alan a string of nine characters drawn from ‘.’, ‘O’, and ‘X’, representing the board position after his play, and Alan will hand back a string where one of the periods has been replaced by an ‘O’. To enable this we will number the nine positions on the tic-tac-toe board as follows.

123
456
789

The string representing the board is just the contents of these squares in numerical order. So, for instance, if Donald has just played his ‘X’ to make the following board position, then he should give Alan the string printed to the right.

...
.OX       ....OX...
...

We will get rid of the labels, images of tic-tac-toe board positions from the front of the match boxes, and replace them with the numbers 1 through 304, so that each matchbox has a unique numerical label.  We will label the matchbox corresponding to the empty board with 1, as that will be how Alan starts a game, by drawing a bead from there, and he will look up what square that color means in a “Transform #1” on a sheet of paper with eight different transforms listed. Here they are:

Transform #1:
  white =1  lilac =2  silver=3  
  black =4  gold  =5  green =6  
  amber =7  red   =8  pink  =9  

Transform #2:
  white =3  lilac =6  silver=9  
  black =2  gold  =5  green =8  
  amber =1  red   =4  pink  =7  

Transform #3:
  white =9  lilac =8  silver=7  
  black =6  gold  =5  green =4  
  amber =3  red   =2  pink  =1  

Transform #4:
  white =7  lilac =4  silver=1  
  black =8  gold  =5  green =2  
  amber =9  red   =6  pink  =3  

Transform #5:
  white =3  lilac =2  silver=1  
  black =6  gold  =5  green =4  
  amber =9  red   =8  pink  =7  

Transform #6:
  white =7  lilac =8  silver=9  
  black =4  gold  =5  green =6  
  amber =1  red   =2  pink  =3  

Transform #7:
  white =1  lilac =4  silver=7  
  black =2  gold  =5  green =8  
  amber =3  red   =6  pink  =9  

Transform #8:
  white =9  lilac =6  silver=3  
  black =8  gold  =5  green =2  
  amber =7  red   =4  pink  =1  

The eight transforms correspond to four rotations (of zero, one, two and three quarters clockwise), and four reflections.

The remaining 303 matchboxes correspond to the essentially different board positions for MENACE’s second, third, and fourth moves. Although there are 72 different board positions for MENACEs second move there are only twelve that essentially distinct, and here they all are, numbered 2 through 13 as the next twelve matchboxes after the one for the first move.

#1    #2    #3    #4    #5    #6    #7    #8    #9    #10   #11   #12   #13
 |     |     |     |     |     |     |     |     |     |     |     |     |
...   .O.   .O.   .O.   .O.   X..   .X.   OX.   O.X   O..   O..   O..   XO.   
...   X..   .X.   ...   ...   .O.   .O.   ...   ...   .X.   ..X   ...   ...   
...   ...   ...   X..   .X.   ...   ...   ...   ...   ...   ...   ..X   ...   

When Alan is given a string by Donald (there is only one possible string for the first move, the empty board, but there are 72 possibilities for the MENACE’s second move, 756 for the third, and 1372 for the fourth move) Alan just mindlessly looks it up in a big table that is printed on a few sheets of paper. Each line has a string representing a board position, a box number, and a transform number. For instance, for the second move for MENACE we talked about above with string ....OX... Alan would find it, simply by matching character for character, in the following part of the table (for the first and second moves by MENACE):

.........  Box: #  1, Transform #1

.......OX  Box: # 13, Transform #3
.......XO  Box: #  8, Transform #3
......O.X  Box: #  9, Transform #6
......OX.  Box: #  8, Transform #6
......X.O  Box: #  9, Transform #3
......XO.  Box: # 13, Transform #6
.....O..X  Box: # 13, Transform #8
.....O.X.  Box: #  2, Transform #8
.....OX..  Box: #  4, Transform #8
.....X..O  Box: #  8, Transform #8
.....X.O.  Box: #  2, Transform #3
.....XO..  Box: # 11, Transform #6
....O...X  Box: #  6, Transform #3
....O..X.  Box: #  7, Transform #3
....O.X..  Box: #  6, Transform #4
....OX...  Box: #  7, Transform #2      <== this one
....X...O  Box: # 10, Transform #3
....X..O.  Box: #  3, Transform #3
....X.O..  Box: # 10, Transform #4
....XO...  Box: #  3, Transform #2
...O....X  Box: #  4, Transform #4
...O...X.  Box: #  2, Transform #4
...O..X..  Box: # 13, Transform #4
...O.X...  Box: #  5, Transform #4
...OX....  Box: #  3, Transform #4
...X....O  Box: # 11, Transform #3
...X...O.  Box: #  2, Transform #6
...X..O..  Box: #  8, Transform #4
...X.O...  Box: #  5, Transform #2
...XO....  Box: #  7, Transform #4
..O.....X  Box: #  9, Transform #2
..O....X.  Box: # 11, Transform #2
..O...X..  Box: # 12, Transform #2
..O..X...  Box: #  8, Transform #2
..O.X....  Box: # 10, Transform #2
..OX.....  Box: # 11, Transform #5
..X.....O  Box: #  9, Transform #8
..X....O.  Box: #  4, Transform #3
..X...O..  Box: # 12, Transform #4
..X..O...  Box: # 13, Transform #2
..X.O....  Box: #  6, Transform #2
..XO.....  Box: #  4, Transform #7
.O......X  Box: #  4, Transform #5
.O.....X.  Box: #  5, Transform #1
.O....X..  Box: #  4, Transform #1
.O...X...  Box: #  2, Transform #5
.O..X....  Box: #  3, Transform #1
.O.X.....  Box: #  2, Transform #1
.OX......  Box: # 13, Transform #5
.X......O  Box: # 11, Transform #8
.X.....O.  Box: #  5, Transform #3
.X....O..  Box: # 11, Transform #4
.X...O...  Box: #  2, Transform #2
.X..O....  Box: #  7, Transform #1
.X.O.....  Box: #  2, Transform #7
.XO......  Box: #  8, Transform #5
O.......X  Box: # 12, Transform #1
O......X.  Box: # 11, Transform #7
O.....X..  Box: #  9, Transform #7
O....X...  Box: # 11, Transform #1
O...X....  Box: # 10, Transform #1
O..X.....  Box: #  8, Transform #7
O.X......  Box: #  9, Transform #1
OX.......  Box: #  8, Transform #1
X.......O  Box: # 12, Transform #3
X......O.  Box: #  4, Transform #6
X.....O..  Box: #  9, Transform #4
X....O...  Box: #  4, Transform #2
X...O....  Box: #  6, Transform #1
X..O.....  Box: # 13, Transform #7
X.O......  Box: #  9, Transform #5
XO.......  Box: # 13, Transform #1

This tells Alan that the move given to him by Donald is to be played by matchbox #7, and then he is to use Transform #2, which we saw above, to interpret the color of the drawn move as to which square is meant.  We can see what position box #7 corresponds to above, though Alan does not know that. He simply reaches into box #7 and pulls out a random bead. As it happens, in my simulation of MENACE where it never tries to play two essentially the same moves, the only beads in #7 are colored white, black, amber, and red, corresponding to essentially different moves down the left column and in the bottom at the middle using the original MENACE bead color interpretations.  Under Transform #2 we see that those colors correspond to squares 3, 2, 1, and 4, respectively, which are across the top row and the left middle square for the way Donald is playing. So whichever one of those colors is removed from the box, Alan simply goes to that position in the string that was given to him by Donald, and changes the blank to an ‘O’. So suppose that Alan pulls out a black bead. In that case he changes the second element to an ‘O’, and gives it back to Donald who then interprets the string to mean the following new board position:

                .O.
.O..OX...       .OX
                ...

The only remaining thing is the reinforcement signal. Donald, the human player, is the one who is responsible for deciding when the game is over and at that point needs to communicate one of just three options to Alan; L, for loss, meaning forfeit all the beads out of boxes, D, for draw, meaning put the beads back with an extra one of the same color for each, or W, for win, meaning put them back with three extra ones of each.

Summary of What Alan Must Do

With these modifications we have made the job of Alan both incredibly simple and incredibly regimented.

    1. When Donald gives Alan a string of nine characters Alan looks it up in a table, noting the matchbox number and transform number.
      1. He opens the numbered matchbox, randomly picks a bead from it and leaves it on the table in front of the open matchbox.
      2. He looks up the color of the bead in the numbered transform, to get a number between one and nine.
      3. He replaces that numbered character in the string with an ‘O’, and hands the string back to Donald.
    2. When Donald gives Alan a sign for one of L, D, or W, Alan does the following:
      1. For L he removes the beads on the table and closes the open matchboxes.
      2. For D he adds one more bead of the same color to each one on the table, and puts the pairs in the matchboxes behind them, and closes the matchboxes.
      3. For W he adds three more beads of the same color to each one on the table, and puts the sets of four in the matchboxes behind them, and closes the matchboxes.

That is all there is. Alan looks up things on a few sheets of paper, acts on matchboxes, and changes a character in a string.

One could say that Alan is a Turing machine.

The thing that learns how to play tic-tac-toe is a combination of Alan following these completely strict rules, and the contents of the matchboxes, the colored beads, whose number varies over time.

Is this how a person would learn?

For anyone who has played tic-tac-toe the most striking thing about about the way that MENACE learns is that it has no concept of “three in a row”. When we teach a child how to play the game that is the most important thing to explain, showing how rows, columns, and diagonals can all give rise to three O’s or X’s in a row. We explain to the child that getting three in a row is the goal of the game. So the first rule for playing tic-tac-toe is to complete three in a row on your move if that option is available. MENACE does not know this at all.

The next thing, or second rule, we might show our tic-tac-toe pupil is that assuming they have no winning move, the next best thing is to block the opponent if they have two of three in a row already with an empty spot to play and complete it. This does not guarantee an eventual win, as there are seventeen essentially different situations where the ‘X’ player may have two three-in-a-row’s ready to play, and the ‘O’ player can only block one of them.  Here are two examples of that.

.O.   XOO
OOX   O..
.XX   .XX

However just these two rules are a marked improvement over random play.  If we play tic-tac-toe with the preference of rule 1 if it is applicable, then rule 2 if that is applicable, and if neither is applicable then make a random move, we actually get a pretty good player. Here is the same sort of table as above, with an identical first row showing how well random untrained play succeeds against Players A, B, and C, then in the second row how well the addition of rules 1 and 2 improve a random player

            | Player A    | Player B    | Player C    |
------------|-------------|-------------|-------------|
random play |  59/ 13/ 28 |   0/ 24/ 76 |  27/ 19/ 53 |
============|=============|=============|=============|
+ rules 1&2 |  86/  10/ 4 |   0/ 82/ 18 |  51/ 37/ 13 |
------------|-------------|-------------|-------------|

Just the addition of those two rules gets to a level of play against a random player (Player A) that MENACE only gets to after about 4,000 games learning from Player A. Against Player B, the optimal player, it does not get as good as it does when it is trained for 200 games by Player B, but it is better against either of the other two players than when it has been been trained for 4,000 games against Player B. And against Player C, the player with 25% error rate from optimal play, it is almost as good as it ever gets even being trained by Player C.

Clearly these rules are very powerful. But they are also rather easy for a child to learn as they don’t require thinking ahead beyond the very next move. All the information is right there in the board layout, and there is no need to think ahead about what the opponent might do next once the current move is made. What is it that the child has that MENACE does not?

One answer might be geometrical representations. MENACE does not have any way to represent “three in a row” as a concept that can be applied to different situations. Each matchbox is a kingdom unto itself about one particular essentially unique board configuration. If one particular matchbox learns, through reinforcement, that it is good to place a third ‘O’ to make a diagonal, there is no way to transfer that insight, were MENACE able to have it, to other essentially different situations where there is also a diagonal that can be filled in. And certainly not to a situation about completing a horizontal or vertical row.

As we saw, Michie did incorporate some geometric “knowledge” into MENACE by mapping all rotations and reflections of the tic-tac-toe board to a common matchbox. But the machine itself has no insight into this–it was all done ahead of time by Michie (whose preparation was extended slightly by me so that Alan could be very explicitly machine-like in his tasks) by producing the dictionary of positions that mapped to matchboxes number 1 through 304, and which of the eight inversion lookup tables that mapped from color of bead to numbered square on the board should be used. That manual design process handled some mappings between different aspects of three in a row but not all. In general a researcher or engineer using Machine Learning to solve a problem does something very similar, in reducing the space of inputs. The art of it is to reduce the input space so that learning can happen more quickly, but not over reduce the space so that subtle differences in situations are obliterated by the pre-processing. By mapping from all the general board positions to precisely those that are essentially different, Donald Michie, the Machine Learning engineer in this case, managed so satisfy both those goals.

A child knows something about geometry in a way that MENACE does not. A child can talk about things being in a row independently of learning tic-tac-toe. A child has learned that in-a-row-ness is independent of orientation of the line the defines the row. By a certain age a child comes to know that the left-to-rightness of some ordering depends on the point of view of the observer, so they are able to see that two in a row with an empty third one is an important generalization that applies equally to the horizontal and vertical rows around the edges, thinking about them in both directions, and also applies to the horizontal and vertical rows that go through the middle square, and to the two diagonals that also go through that square. The child may or may not generalize that to two at each end of a row with the middle to be filled in–perhaps that might be a different concept for young children. But the rowness of things is something they have a lot of experience with, and are able to apply to tic-tac-toe. In computer science we would talk about rowness being a first class object for a child–something that can be manipulated by other programs, or in a child by many cognitive systems. In MENACE rowness is hidden in the pre-analysis of the problem that Donald Michie did in order to map tic-tac-toe to collection of numbered matchboxes with beads in them.

The learning that MENACE does somehow feels different to the learning that human does when playing tic-tac-toe. That is not to say that all learning that a human does is necessarily completely different from what MENACE does. Perhaps things that humans learn in an unconscious fashion (e.g., how to adjust their stance to stay balanced–negative and positive reinforcement signals based on whether they hit the ground or not), where we have no way to access what is happening inside us, nor an ability to talk about it, is more like MENACE learning.

Not all learning is necessarily the same sort of learning.

Is this how a person would play?

A more fundamental question, perhaps, is whether MENACE plays tic-tac-toe like a person does, and I think the answer is a clear no. The MENACE system consisting of the matchboxes and Alan strictly following rules only fills in part of the role of a normal player. The rest of what is usually a social interaction between two people is all taken on by Donald.

There is no representation inside the MENACE (where we include in the definition of MENACE the sheets of papers that Alan consults, and the rules that we have instructed Alan to strictly follow) of tic-tac-toe being a game that is played. MENACE does not know what a game is, or even that it is playing a game. All that happens inside MENACE is that one at a time, either three or four times sequentially, one of its matchbox drawers is opened and a bead is randomly removed, and then either the beads are taken away, or they are put back in the boxes from where they came with either one or three additional beads of the same color, and the boxes are closed.

All the gameness of tic-tac-toe is handled by the human Donald. It is he who initiates the game by handing Alan a string of nine periods. It is he who manages the consistency of subsequent turns by annotating his hand drawn tic-tac-toe board with the moves. It is he who decides when the game has been won, drawn, or lost, and communicates to Alan the reinforcement signal that is to be applied to the open matchboxes. It is he, Donald, who decides whether and when to initiate a new game.

MENACE does not know, nor does it learn, what a game is. The designer of MENACE abstracted that away from the situation, so that MENACE could be a pure learning machine.

That today is both the strength and weakness of modern Machine Learning. Really smart people, researchers or engineers, come up with an abstraction for the problem in the real world that they want to apply ML to. Those same smart people figure out how data should flow to and fro between the learning system and the world to which it is to be applied. They set up machinery which times and gates that information flow from the application. They set up a signaling system on when the learning system is supposed to respond to an input (in MENACE’s case a string of nine characters drawn from ‘.’, ‘X’, and ‘O’) and produce an output. And those same people set up a system which tells the learning system when to learn, to adjust the numbers inside it, in response to a reinforcement signal, or in some other forms of ML a very different, but still similarly abstracted signal–we will see that in the next chapter.

tic-tac-toe machine resonates with modern ML

Although MENACE is well over fifty years old, it it shares many properties with modern Machine Learning systems, though of course it is much smaller and simpler than the systems that people use today–one must expect something from 50+ years of hard intellectual work. But the essential problems that MENACE and today’s ML algorithms have are very instructive as they can give some intuition about some of the limits we might expect for modern AI and ML.

Parameters. After the design work was done on MENACE, all that could change during learning as the value of the 1087 parameters, the numbers of various colored beads in various matchboxes. Those numbers impact the probability of randomly picking a bead of a particular color from a matchbox. If the number of red beads goes down and the number of amber beads goes up over time in a single matchbox, then it is more likely that Alan will pick an amber bead at random.  In this way MENACE has learned that for the particular situation on a tic-tac-toe board corresponding to that matchbox the square corresponding to the amber bead is a better square to play than the one corresponding to a red bead. All MEANCE is doing is juggling these numbers up and down. It does not learn any new structure to the problem while it learns. The structure was designed by a researcher or engineer, in this case Donald Michie.

This is completely consistent with most modern Machine Learning systems. The researchers or engineers structure the system and all that can change during learning is a fixed quantity of numbers or parameters, pushing them up or down, but not changing the structure of the system at all. 1087 may seem like a lot of parameters for playing tic-tac-toe, but really that is the price of eliminating the geometry of the board from the MENACE machine.  In modern applications of Machine Learning there are often many millions of parameters. Sometimes they take on integer values as do the number of beads in MENACE, but more usually these days the parameters are represented as floating point numbers in computers, things that can take on values like 5.37, -201.65, 894.78253, etc.

Notice how simply changing a big bunch of numbers and not changing the underlying abstraction that connected the external problem (playing tic-tac-toe) to a geometry-free internal representation (the numbers of different colored beads in matchboxes) is very different from how we have become familiar with using computers. When we manage our mail box folders, creating special folders for particular categories (e.g., “upcoming trips”, “kids”, etc.) and then sub folders (e.g., “Chicago May 5”, “soccer”, etc.) and then filing emails in those subfolders, we are changing the structure of our representation of the important things in our life which are covered by emails. Machine Learning, as in the case of MENACE, usually has an engineering phase were the problem is converted to a large number of parameters, and after that there is no dynamic updating of structures.

In contrast, I think all our intuitions tell us that our own learning often has our internal mental models tweak and sometimes even radically change how we are categorizing aspects of the skill or capability that we are learning.

Large Parameters. My computer simulations of MENACE soon had the numbers of beads of a particular color in particular boxes ranging from none or one up to many thousand. This intuitively seem strange but is not uncommon in today’s Machine Learning systems. Sometimes there will be parameters that are between zero and one, were just a change of one ten thousandth in value will have drastic effects on the capabilities that the system is learning, while at the same time there will be parameters that are up in the millions. There is nothing wrong with this, but it does feel a little different from our own introspections of how we might weigh things relatively in our own minds.

Many Examples Needed. If we taught tic-tac-toe to an adult we would think that just a few examples would let them get the hang of the game. MENACE on the other hand, even when carefully tutored by Donald Michie took a couple of hundred examples to get moderately good. My simulation is still making relatively big progress after three thousand games and is often still slowly getting even a little better at four thousand games. In modern Machine Learning systems there may be tens of millions of different examples that are needed to train a particular system to get to adequate performance. But the system does not just get exposed to each of these training examples once. Often each of those millions of examples needs to be shown to the system hundreds of thousands of times. Just being exposed to the examples once leaves way to much bias from the most recently processed examples. Instead by having them re-exposed over and over, after the ML system has already seen all of them many times, the recentness bias gets washed away into more equal influence from all the examples.

Training examples are really important. Learning to play against just one of Player A, B, or C, always lead to very different performance levels against each of these different players with learning turned off in my computer simulation of MENACE.  This too is a huge issue for modern Machine Learning systems. With millions of examples needed there is a often a scale issue of how to collect enough training data.  In the last couple of years companies have sprung up which specialize in generating training data sets and can be hired for specific projects.  But getting a good data set which does not have unexpected biases in it can often be a problem.

When MENACE is trained against Player B, the optimal player that can not be beaten, MENACE does not learn how to win, as it never has an experience of winning so it never receives reinforcement for winning. It does learn how to not be defeated, and so playing against Players A or C its win rate does go up a little as they each sometimes screw up, but MENACE’s winning rate does not go up as much as it does when it trains against those two players. In our example with MENACE my simulations worked best overall when trained against Player C, as that had a mixture of examples that  were tough to win against (when it got through a game without making a random bad choice), and because of its occasional random choices examples which more fully spanned all of the possible playing styles MENACE might meet. In the parlance of Machine Learning we would say that when MENACE was trained only against Player B, the optimal player, it overfit its playing style to the relatively small number of games that it saw (no wins, and few losses) so was not capable when playing against more diverse players.

In general, the more complex the problem for which Machine Learning is to be used, the more training data that will be needed.  In general, training data sets are a big resource consideration in building a Machine Learning system to solve a problem.

Credit assignment. The particular form of learning that MENACE both first introduced and demonstrates is reinforcement learning, where the system is given feedback only once it has completed a task. If many actions were taken in a row, as is the case with MENACE, either three of four moves of its own before it gets any feedback, then there is the issue of how far back the feedback should be used.

In the original MENACE all three forms of reinforcement, for a win, a draw, or a loss, were equally applied to all the moves. Certainly it makes sense to apply the reinforcement to the last move, as it directly did lead to that win, or a loss. In the case of a draw however, it could in some circumstances not be the best move as perhaps choosing another move would have given a direct win. As we move backward, credit for whether earlier moves were best, worst, or indifferent is a little less certain. In the case of Player A or C as the opponent it may have simply made a bad move in reply to a bad move by MENACE early on, so giving the earlier move three beads for a win may be encouraging something that Player  B, the optimal player, will be able to crush. A natural modification would be three beads for the last move in a winning game, two beads for the next to last, and one bead for the third to last move.  Of course people have tried all these variations and under different circumstances much more complex schemes would be the best. We will discuss this more, a little later.

In modern reinforcement learning systems a big part of the design is how credit is assigned. In fact now it is often the case that the credit assignment itself is also something that is learned by a parallel learning algorithm, trying to optimize the policy based on the particulars of the environment in which the reinforcement learner finds itself.

Getting front end processing right. In MENACE Michie developed what might be called “front end processing” to map all board positions to only those that were essentially distinct. This simultaneously drastically cut down the number of parameters that had to be learned, let the learning system automatically transfer learning across different cases in the full world (i.e., across symmetries in the tic-tac-toe board), and introduced zero entanglements that could confuse the learning process.

Up until a few years ago Machine Learning systems applied to understanding human speech usually had as their front end programs that had been written by people to determine the fundamental units of speech that were in sound being listened to.

Those fundamental units of speech are called phonemes, and they can be very different for different human languages. Different units of speech lead to different words being heard. For instance, the four English words pad, pat, bad, and bat all have three phonemes with the same middle phoneme corresponding to the vowel sound (in English the same letters may be used represent to different phonemes, so the word paper, while having the same letter ‘a’ for the second phoneme (of four in this word) has a very different sound associated with it, and is therefore a different phoneme), the four different phonemes p, b, d, and t, lead to four different words being heard as p and b are varied at the start, and d and t are varied at the end.

In earlier speech understanding systems the specially built front end phoneme detector programs relied on some numerical estimators of certain frequency characteristics of the sounds and produced phoneme labels as their output that were fed into the Machine Learning system to recognize the speech. It turned out that those detectors were limiting the performance of the speech understanding systems no matter how well they learned. Relatively recently those programs were replaced by other machine learning system, that didn’t necessarily output conventional phoneme representations, and this lead to a remarkable overall increase in reliability of speech understanding systems. This is why today, but only in the last few years, many people now have devices in their homes, such as Amazon’s Echo or Google’s Home, that they can easily interact with via voice.

Getting the front end processing right for an ML problem is a major design exercise. Getting it wrong can lead to much larger learning systems than necessary, making learning slower, perhaps impossibly slower, or it can make the learning problem impossible if it destroys vital information from the real domain. Unfortunately, since in general it is not known whether a particular problem will be amenable to a particular Machine Learning technique, it is often hard to debug where things have gone wrong when an ML system does not perform well.  Perhaps inherently the technique being used will not be able to learn what is desired, or perhaps the front end processing is getting in the way of success.

Geometry is hard. Just as MENACE knew no geometry and so tackled tic-tac-toe in a fundamentally different way than how a human would approach it, most Machine Learning systems are not very good at preserving geometry nor therefore are they good at exploiting it. Geometry does not play a role in speech processing, but for many other sorts of tasks there is some inherent value to the geometry of the input data. The engineers or researchers building the front end processing for the system need to find a way to accommodate the poor geometric performance of the ML system being used.

The issue of geometry and the limitations of representing it in a set of numeric parameters arranged in some fixed system, as was the case in MENACE, has long been recognized. It was the major negative result of the book Perceptrons⁠11 written by Marvin Minsky and Seymour Papert in 1969. While people have attributed all sorts of motivations to the authors I think that their insights on this front, formally proved in the limited cases they consider, still ring true today.

Fixed structure stymies generalization. MENACE’s fixed structure meant that anything it implicitly learned about filling or blocking three in a row on a diagonal could not be transferred to filling or blocking a vertical or horizontal row. The fixed structures spanning thousands or millions of variable numerical parameters of most Machine Learning systems likewise stymies generalization. We will see some surprising consequences of this when we look at some of the most recent exciting results in Machine Learning in a later blog post–programs that learn to play a video game but then fail completely and revert to zero capability on exactly the same game when the colors of pixels are mapped to different colorations, or if each individual pixel is replaced by a square of four identical pixels.

Furthermore, any sort of meta-learning is usually impossible too. Since MENACE doesn’t know that it is playing a game, and since there is nothing besides the play and reward mechanism that can access the matchboxes, there is no way that observations of the flow of a game can be ruminated upon. A child might learn a valuable meta-lesson in playing tic-tac-toe, that when you have an opportunity to win take it immediately as it might go away if the other player gets to take a turn. That would correspond to learning rule 1 in our comparison between MENACE and how a person might learn.

Machine Learning engineers and researchers must, at this point in the history of AI, form an optimized and fixed description of the problem and let ML adjust parameters. All possibility of reflective learning is removed from these very impressive learning systems. This greatly restricts how much power of intelligence and AI system with current day Machine Learning systems can tease out of their learning exploits. Humans are generally much much smarter than this.

A Few Developments in Reinforcement Learning

The description of reinforcement learning comes from 1961, and is the first use of the term reinforcement learning when applied to a machine process that I can find. There have been some developments in reinforcement learning since 1961, but only in details as this section shows. The fundamental ideas were all there in Donald Michie’s matchboxes.

Reinforcement learning is still an active field of research and application today. It is commonly used in robotics applications, and for playing games. It was part of the system that beat the world Go champion in 2016, but we will come back to that in a little bit.

After Michie’s first paper, reinforcement learning was formalized over the next twenty years. Without resorting to the mathematical formulation, today reinforcement learning is used where there are a finite number of states that the world can be in.  For MENACE those states correspond to the 304 matchboxes of essentially different tic-tac-toe board positions where it is O’s turn to play. For each state there are a number of possible actions (the different colored beads in each matchbox corresponding to the possible moves). The policy that the system currently has is the probability of each action in each state, which for MENACE corresponds to the number of beads of a particular color in a matchbox divided by the total number of beads in that same matchbox. Reinforcement learning tries to learn a good policy.

The structure of states and actions for MENACE, and indeed for reinforcement learning for many games, is a special case, in that the system can never return to a state once it has left it. That would not be the case for chess or Go where it is possible to get back to exactly the same board position that has already been seen.

For many systems of reinforcement learning real numbers are used rather than integers as in MENACE. In some cases they are probabilities, and for a given state they must sum to exactly one. For many large reinforcement learning problems, rather than represent the policy explicitly for each state, it is represented as a function approximated by some other sort of learning system such as a neural network, or a deep learning network. The steps in the reinforcement process are the same, but rather than changing values in a big table of states and actions, the 1087 parameters of MENACE, a learning update is given to another learning system.

MENACE, and many other game playing systems, including chess and Go this time, are a special case of reinforcement learning in another way. The learning system can see the state of the world exactly. In many robotics problems where reinforcement learning is used that is not the case. There the robot may have sensors which can not distinguish all the nuances in the world (e.g., for a flying robot it may not know the exact current wind speed and direction ten meters away from it in the direction of travel). For these sorts of reinforcement learning problems the world is referred to as partially observable.

in MENACE any rewards, be they positive or negative were spread equally over all moves leading up the win, loss, or draw. But in reality it could be that an early move was good, and just a dumb move at the end was bad. To handle this problem Christoper Watkins came up with a method that became known as Q-learning for his Ph.D. thesis12, titled “Learning from Delayed Rewards”, at Cambridge University in 1989. The Q function that he learns is an estimate of what the ultimate reward will be by taking a particular action in a particular state. Three years later he and Peter Dayan published a paper that proved that under some reasonable assumptions his algorithm always eventually converged on the correct answer as to how the reward should be distributed.

This method, which is at its heart the reinforcement learning of Donald Michie’s MENACE from 1961, is what is powering some of today’s headlines. The London company DeepMind, which was bought by Google, uses reinforcement learning (as they explain here) with the Q-learning implemented in something called deep learning (another popular headline topic). This is how they built their Alpha Go program which recently beat both the human Korean and Chinese Go champions.

As a side note, when I visited DeepMind in June this year I asked how well their program would have done if on the day of the tournament the board size had been changed from 19 by 19 to 29 by 29. I estimated that the human champions would have been able to adapt and still play well. My DeepMind hosts laughed and said that even changing to an 18 by 18 board would have completely wiped out their program…this is rather consistent with what we have observed about MENACE. Alpha Go plays Go in a way that is very different from how humans apparatently play Go.

Overloaded words

In English, at least, ships do not swim. Ships cruise or sail, whereas fish and humans swim. However in English planes fly, as do birds. By extension people often fly when they go on vacation or on a business trip. Birds move from one place to another by traveling through the air. These days, so too can people.

But really people do not fly at all like birds fly. Our technology lets us “fly” a quarter of the way around the world, non-stop, in less than a day. Birds who can fly that far non-stop (and there are some) certainly take a lot longer than a day to do that.

If humans could fly like birds we would think nothing of chatting to a friend on the street on a sunny day, and as they walk away, flying up into a nearby tree, landing on a branch, and being completely out of the sun. If I could fly like a bird then when on my morning run I would not have to wait for a bridge to get across the Charles River to get back home, but could choose to just fly across it at any point in its meander.

We do fly. We do not fly like birds. Human flying is very different in scope, in method, and in auxiliary equipment beyond our own bodies.

Arthur Samuel introduced the term Machine Learning for two sorts of things his computer program was doing as it got better and better over time at and through the experience of playing checkers. A person who got better and better over time at and through the experience of playing checkers would certainly be said to be learning to be a better player. With only eight to ten hours experience Samuel’s program (he was so early at this he did not give a name to his program–that innovation had to away the early 1960’s) got better at playing checkers than Samuel himself. Thus, in his first sentence of his paper, again, does Samuel justify the term learning: “The studies reported here have been concerned with programming of a digital computer to behave in a way which, if done by human beings or animals, would be described as involving the process of learning.”

What I have tried to do in this post is to show how Machine Learning works, and to provide an argument that it works in a way that feels very different to how human learning of similar tasks proceeds. Thus, taking an understanding of what it is like for a human to learn something and applying that knowledge to an AI system that is doing Machine Learning may lead to very incorrect conclusions about the capabilities of that AI system.

Minsky13 labels as suitcase words terms like consciousness, experience, and thinking. These are words that have so many different meanings that people can understand different things by them. I think that learning is also a suitcase word. Even for humans it surely refers to many different sorts of phenomena. Learning to ride a bicycle is a very different experience from learning ancient Latin. And there seems to be very little in common in the experience of learning algebra and learning to play tennis. So, too, is Machine Learning very different from any sort of the myriad of different learning capabilities of a person.

The word “learn” can lead to misleading conclusions.

Postscript

I am going to indulge myself a little by pontificating here. Be warned.

In 1991 I wrote a long (I have been pontificating since I was relatively young) paper14  on the history of Artificial Intelligence and how it had been shaped by certain key ideas. In the final paragraphs of that paper I lamented that there was a bandwagon effect in Artificial Intelligence Research, and said that “[m]any lines of research have become goals of pursuit in their own right, with little recall of the reasons for pursuing those lines”.

I think we are in that same position today in regard to Machine Learning. The papers in conferences fall into two categories. One is mathematical results showing that yet another slight variation of a technique is optimal under some carefully constrained definition of optimality. A second type of paper takes a well know learning algorithm, and some new problem area, designs the mapping from the problem to a data representation (e.g., the mapping from tic-tac-toe board positions to the numbers 1 through 304 for the three hundred and four matchboxes that comprise MENACE), and show the results of how well that problem area can be learned.

This would all be admirable if our Machine Learning ecosystem covered even a tiny portion of the capabilities of human learning. It does not. And, I see no alternate evidence of admirability.

Instead I see a bandwagon today, where vast numbers of new recruits to AI/ML have jumped aboard after recent successes of Machine Learning, and are running with particular versions of it as fast as they can. They have neither any understanding of how their tiny little narrow technical field fits into a bigger picture of intelligent systems, nor do they care. They think that the current little hype niche is all that matters, are blind to its limitations, and are uninterested in deeper questions.

I recommend reading Christopher Watkins Ph.D. thesis12 for an example of something that is admirable. It revitalized reinforcement learning by introducing Q-learning, and that is still having impact today, thirty years later. But more importantly most of the thesis is not about the particular algorithm or proofs about how well it works under some newly defined metric. Instead, most of the thesis is an illuminating discussion about animal and human learning, and attempting to get lessons from there about how to design a new learning algorithm. And then he does it.



1 Machine Learning: A Probabilistic Perspective, Kevin P. Murphy, MIT Press, 2012.

2 “Some Studies in Machine Learning Using the Game of Checkers”, Arthur L. Samuel, IBM Journal of Research and Development, 3(3):210–229, 1959.

3 When I first joined the Stanford Artificial Intelligence Laboratory (SAIL) in 1977 I got to meet Arthur Samuel. Born in 1901 he was certainly the oldest person in the lab at that time. After retiring from IBM in 1966 he had come to SAIL as a researcher. Arthur was a delightful and generous person, and besides his research he worked on systems programming in assembler language for the Lab’s time shared computer. He was the principal author of the full screen editor (a rarity at that time) that we had, called Edit TV, or ET at the command level. He was still programming at age 85, and last logged in to the computer system when he was 88, a few months before he passed away.

4 Perhaps I am wrong about exactly what Samuel was referring to. In his Ph.D. thesis12, which I talk about later in the post, Christopher Watkins allows that perhaps Samuel means what I interpret him to mean, though perhaps there is a smarter version of it that was implemented that involved recomputing the saved computations when more of the game tree had been searched. Watkins was unable to tell exactly from reading the paper.

5 “Trial and Error”, Donald Michie, Penguin Science Survey, vol 2, 1961.

6 “How to build a game-learning machine and then teach it to play, and to win”, Martin Gardner, Scientific American, 206(3):138–153, March 1962.

7 We Built Our Own Computers, A. B. Bolt, J. C. Harcourt, J. Hunter, C. T. S. Mayes, A. P. Milne, R. H Surcombe, and D. A. Hobbs, Cambridge University Press, 1966.

8 “Experiments on the Mechanization of Game-Learning Part I. Characterization of the Mode and its parameters”, Donald Michie, Computer Journal, 6(3):232–236, 1963.

9 Michie reports only 287 essentially different situations so his version of MENACE had only 287 matchboxes (though in a 1986 paper he refers to there being 288 matchboxes). Many people have since built copies of MENACE both physically and in computer simulations, and all the ones that I have found on the web report 304 matchboxes, virtual or otherwise. This matches how I counted them in my simulation of MENACE as a program.

10 In all the test results I give I froze the learning and ran 100,000 games–I found that about that number were necessary to give 2 digits, i.e., a percentage, that was stable for different such trials. Note that in total there are 301,248 different legal ways to play out a game of tic-tac-toe. If we consider only essentially different situations by eliminating rotational and reflective symmetries then that number drops to 31,698.

11 Perceptrons: An introduction to Computational Geometry, Marvin Minsky and Seymour Papert, MIT Press, 1968.

12 Learning from Delayed Rewards, Christopher J. C. H. Watkins, Ph.D. thesis, King’s College, Cambridge University, May 1989.

13 The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind, Marvin Minsky, Simon and Schuster, 2006.

14 “Intelligence Without Reason, Proceedings of 12th International Joint Conference on Artificial Intelligence, Sydney, Australia, August 1991, 569–595.

[FoR&AI] Domo Arigato Mr. Roboto

rodneybrooks.com/forai-domo-arigato-mr-roboto/

[An essay in my series on the Future of Robotics and Artificial Intelligence.]

Friday March 11th, 2011, was a bad day for Japan. At 2:46pm local time a magnitude 9.1 earthquake occurred 72 kilometers offshore, east of the Oshika Peninsula which is in the Tohoku region of Japan. A great tsunami was triggered with maximum wave height believed to be 42.5 meters (133 feet) and a few minutes after the earthquake it hit the town of Miyako, 432 kilometers (300 miles) north of Tokyo. Hundreds of kilometers of the coastal region was devastated with almost 16,000 deaths, over 2,500 people missing, and three quarters of a million buildings either collapsed, partially collapsed, or were severely damaged.

The following week things got worse. Japan has been forever changed by what happened in March and April of that year.

A little before 8am on Friday April 25th, 2014, I met up with a small number of robotics researchers from the United States in the Ueno train station in Tokyo. It was a somber rendezvous, but I did not yet realize the sobering emotions I would feel later in the day.

As a technologist I have had more than my fair share of what I think of as “science fiction” days, most of them quite uplifting and exciting. Science fiction days for me are days where I get to experience for real something that heretofore most people have only ever experienced by watching a movie. For instance on July 4, 1997, I was at JPL (the Jet Propulsion Laboratory in Pasadena, California) watching live images come in from the surface of Mars soon after the soft landing of the Pathfinder mission. A little later in the afternoon, to hearty cheers, the Sojourner robot rover deployed onto the surface of Mars, the first mobile ambassador from Earth. Dan Goldin, the administrator of NASA, congratulated all the JPL technologists on the first “faster, cheaper, better” mission. That phrase was a cleaned up version of a title of a paper⁠1 I had written in 1989 with Anita Flynn: “Fast, Cheap, and Out of Control: A Robot Invasion of the Solar System”, where we had proposed the idea of small rovers to explore planets, and explicitly Mars, rather than large ones that were under development at that time. The rover that landed in 1997 was descended from a project at JPL that Colin Angle, then a brand new graduate of M.I.T., and I had helped get started that same year, 1989. The day of the landing was a great science fiction day, and it was related to the one I was about to experience almost seventeen years later.

Really though, April 25th, 2014 was for me two science fiction days rolled into one. Both of them were dystopian.

The group that formed up in Ueno station was lead by Gill Pratt. Gill had been a faculty member in the M.I.T. Artificial Intelligence Laboratory when I had been its director in the late 1990s. He had lead the “leg laboratory”, within the AI Lab, working on making robots that could walk and run. Now he was a program manager at DARPA, the Defense Advanced Research Projects Agency, part of the US Defense Department, leading the DARPA Robot Challenge, a competition whose final was to be held the next year to push forward how robots could help in disaster situations. We robotics researchers were in Japan that week to take part in a joint US/Japan robotics workshop that was held as a satellite event for a summit in Tokyo between Prime Minister Abe and President Obama.

On that Friday morning we took an express train to Iwaki, and from there a fifty minute minibus ride to the “J-village”.  Now things started to get a little surreal. J.League is the Japan Professional Football League, and the J-village was, until the earthquake and tsunami, the central training facility for that league, with multiple soccer pitches, living quarters, a gym, swimming pool, and large administrative buildings. Now three of the pitches were covered in cars, commuter lots for clean up crews. Trucks and minibuses coming from the north were getting scanned for radiation, while all northbound traffic had to go through security gates from the soccer facility. The J-village was now the headquarters of the operation to deal with the radiation released from the Fukushima Daiichi nuclear power plant, when the tsunami had hit it, ultimately leading to three of its six reactors melting down. The J-village was right on the border of a 20 kilometer radius exclusion zone established around that plant, and was being operated by TEPCO, the Tokyo Electric Power Company which owned Fukushima Daiichi, along with Fukushima Daini, also in the exclusion zone, whose four reactors were able to be shut down safely without significant damage.

Inside the main building the walls signaled professional soccer, decorated with three meter high images of Japanese stars of the game. But everything else looked makeshift and temporary. We were met by executives from TEPCO and received our first apology from them for their failures at Daiichi right after the tsunami. We would receive more apologies during the day. This was clearly a ritual for all visitors as none of use felt we were owed any sort of apology. As had happened the day before in a meeting with a government minister, and again rather embarrassingly, I was singled out for special thanks.

After Colin Angle and I had helped get the small rover program at JPL going, where it was led by David Miller and Rajiv Desai, we got impatient about getting robots to other places in the solar system. So, joined by Helen Greiner a friend of Colin’s and for whom I had been graduate counsellor at M.I.T., we started a space exploration robot company originally called IS Robotics. In a 2002 book⁠2 I told the story of our early adventures with that company and how our micro-rovers being tested at Edwards Air Force Base as a potential passenger on a Ballistic Missile Defense Organization (BMDO, popularly known as “Star Wars”) mission to the Moon forced NASA’s hand into adding the Sojourner rover to the Pathfinder mission. By 2001 our company had been renamed to be iRobot, and on the morning of September 11 of that year we got a call to send robots to ground zero in New York City. Those robots scoured nearby evacuated buildings for any injured survivors that might still be trapped inside. That led the way for our Packbot robots to be deployed in the thousands in Afghanistan and Iraq searching for nuclear materials in radioactive environments, and dealing with road side bombs by the tens of thousands. By 2011 we had almost ten years of operational experience with thousands of robots in harsh war time conditions.

A week after the tsunami, on March 18th 2011, when I was still on the board of iRobot, we got word that perhaps our robots could be helpful at Fukushima. We rushed six robots to Japan, donating them, and not worrying about ever getting reimbursed–we knew the robots were on a one way trip. Once they were sent into the reactor buildings they would be too contaminated to ever come back to us.  We sent people from iRobot to train TEPCO staff on how to use the robots, and they were soon deployed even before the reactors had all been shut down.

The oldest of the reactors had been operating for 40 years, and the others shared the same design. None of them had digital monitoring installed, so as they overheated and explosions occurred and they released high levels of radiation there was no way to know what was going on inside the reactor buildings.  The four smaller robots that iRobot sent, the Packbot 510, wieighing 18kg (40 pounds) each with a long arm, were able to open access doors, enter, and send back images. Sometimes they needed to work in pairs so that the one furtherest away from the human operators could send back signals via an intermediate robot acting as a wifi relay. The robots were able to send images of analog dials so that the operators could read pressures in certain systems, they were able to send images of pipes to show which ones were still intact, and they were able to send back radiation levels. Satoshi Tadokoro, who sent in some of his robots later in the year to climb over steep rubble piles and up steep stairs that Packbot could not negotiate, said⁠3 “[I]f they did not have Packbot, the cool shutdown of the plant would have [been] delayed considerably”. The two bigger brothers, both were the 710 model, weighing 157kg (346 pounds) with a lifting capacity of 100kg (220 pounds) where used to operate an industrial vacuum cleaner, move debris, and cut through fences so that other specialized robots could access particular work sites.

Japan has been consistently grateful for that help; we were glad that our technology could be helpful in such a dire situation.

In 2014, at the J-village, after a briefing on what was to come for us visitors, we were issued with dosimeters and we put on disposable outer garments to catch any radioactive particles. We then entered a highly instrumented minibus, sealed from unfiltered external air circulation, and headed north on the Rikuzenhama Highway.  The first few villages we saw were deserted but looked well kept up. That was because by that time the owners of the houses were allowed to come into the exclusion zone for a few hours each day to tend to their  properties. Further in to the zone everything started to look abandoned. After we passed the Fukushima Daini plant which we could see in the distance, we got off the highway and headed down into the town of Tomioka. The train station, quite close to the coast had been washed away, with just the platform remaining, and a single toilet sitting by itself still attached to the plumbing below. Most of the houses had damage to their first floors and from our minibus driving by we could see people’s belongings still inside. At one point we had to go around a car upside down on its roof in the middle of the road. Although it was three years after the event, Tomioka was frozen in time, just as it had been left by the tsunami. This was the first science fiction experience of the day. For all the world it looked like the set of a post-apocalyptic Hollywood movie. But this was a real post-apocalyptic location.

Back on the highway we continued north to the Fukushima Daiichi plant for science fiction experience number two. There are about six thousand people who work at the site cleaning up the damage to the power plant from the tsunami. Only a much smaller number are there on any given day, as continued exposure to the radiation levels is not allowed. We entered a number of buildings higher up the hill than the four reactors that were directly hit by the tsunami. All of them had makeshift piping for air, a look of temporary emergency setups, and all inside were wearing light protective garments as were we. Those who were outside had much more substantial protective clothing, including filtered breathing masks. Most times as we transitioned into and out of buildings we had to go through elaborate security gates where we entered machines that scanned us for radiation in case we had gotten some radioactive debris attached to us. Eventually we got to a control center, really just a few tables with lap tops on them, where the iRobot robots were still being operated from. We watched remotely as one was inside one of the reactor buildings measuring radiation levels–in some cases levels so high that a person could only spend just a few minutes per year in such an area.

Outside we drove around in our sealed bus. We saw where the undamaged fuel rods that had been inside the reactor buildings, but not inside the reactors, were being brought for temporary storage. That task was expected to be completed by the end of this decade. We saw almost (at that time) 1,000 storage tanks, each with about 1,000 tons of contaminated ground water that came down the hill during rainfall and then would be contaminated as it seeped through the ground around the reactor buildings. We saw where they were trying to freeze the ground down to many meters in depth to stop water flowing underground from the hill to the reactor buildings. We saw where along the ocean side of the reactor buildings workers had installed a steel wall from interlocking pylons driven into the seabed, holding back the ocean but more importantly stopping any ground water from leaking into the ocean. Everywhere were people in white protective suits with breathing equipment, working for short periods of time and then being cycled out so that their radiation exposure levels were not unsafe. Eventually we drove down to right near reactor number four, and saw the multi-hundred ton superstructure that had been installed over the building by remotely operated cranes so that the undamaged fuel rods could be lifted out of the damaged water pools where they were normally stored. We wanted to stay a little longer but the radiation level was creeping up, so soon it was decided that we should get out of there. And finally we received a briefing about the research plans on how to develop new robots that starting around the year 2020 would be able to begin the decades long clean up the three melted down reactors.

That really was a science fiction experience.

Robots

Robots were essential to the shutdown of Fukushima Daiichi, and will be for the next thirty or more years as the cleanup continues. The robots that iRobot sent were controlled by operators who looked at images sent back to decide where they should go, whether they should try to climb a pile of debris or not, and give the robots detailed instructions on how to handle unique door handles. In the sequence of three images below a pair of Packbot 510’s first open a door using a large rotary handle, push it open, and then proceed through.

[These are photographs of the operators’ console and in some cases you might just be able to make out the reflection of the operators in protective suits wearing breathing equipment.] Below we see a 510 model confronted by relatively light debris that it will be able to get over fairly safely.

In the image below a 710 model is set up to go and vacuum up radioactive material.

But the robots we sent to Fukushima were not just remote control machines. They had an Artificial Intelligence (AI) based operating system, known as Aware 2.0, that allowed the robots to build maps, plan optimal paths, right themselves should they tumble down a slope, and to retrace their path when they lost contact with their human operators. This does not sound much like sexy advanced AI, and indeed it is not so advanced compared to what clever videos from corporate research labs appear to show, or painstakingly crafted edge-of-just-possible demonstrations from academic research labs are able to do when things all work as planned. But simple and un-sexy is the nature of the sort of AI we can currently put on robots in real, messy, operational environments.

But wait! What about all those wonderful robots we have seen over the years, in the press, the ones that look like Albert Einstein, or the humanoids that have been brought into science museums around the United States for shows, or the ones we see brought out whenever a US President visits Japan4. You have seen them. Like the humanoid ones that walk on two legs, though with bended knees, which does look a little weird, turning to the audience and talking from behind a dark glass visor, sometimes seeming to interact with people, taking things from them, handing things to them, chatting, etc. What about them? They are all fake! Fake in the sense that though they are presented as autonomous they are not. They are operated by a team of usually six people, off stage. And everything on stage has been placed with precision, down to the millimeter. I have appeared on stage before those robots many times and been warned not to walk near or touch or any of the props, for example staircases, as that will make the robot fail, and when it does fail it is not aware that it has.

Corporate marketers had oversold a lot of robots, and confused many people about current robots’ true capabilities. Corporate marketing robots had no chance at all of helping in Fukushima.

Those robots are not real5.

Reality is hard.

Reality

Robotics, including self driving cars, is where Artificial Intelligence (AI) collides with the un-sanitized natural world. Up until now the natural world has been winning, and will probably continue to do so most of the time for quite some time.

We have come to expect our technology to be 100% reliable. We expect our car to start every morning and for the wheels to drive it forward when we push down on the gas pedal. We expect the plane that we board to both take off and land safely, even if, through experience, we tolerate it being late. We expect the internet to provide the web pages we go to on our smart phones. We expect our refrigerators and microwave ovens to work every day so that we can eat and survive.

AI has gotten a pass on providing 100% reliability as so many of its practical applications are mediated by a functioning cognitive human who naturally fills in the missing pieces for the AI system. Us humans do this all the time for small children and for the very elderly. We are wired to be accommodating to other intelligences that we think of as less than us.  Most of our AI technology is very much less than us, so we accommodate.

The demands of having robots interact with the un-sanitized natural world cancel that free pass. The natural world usually does not care that it is a robot rather than a person, and so the natural world is not accommodating. In my opinion there is a mismatch between what is popularly believed about AI and robotics, and what the reality is for the next few decades.

I have spent the last forty years as part of the Artificial Intelligence (and more general computer science) research groups at either Stanford or M.I.T. as a student, post-doc, faculty (both places), or more recently, emeritus professor. Through companies that I have co-founded, iRobot and Rethink Robotics (and yes, I was also once a co-founder and consultant for eight years to a silicon valley AI software company–it eventually failed, and then there was the robotics VC firm I co-founded, and then, etc., etc.), I have been involved in putting a lot of robots to work in five different domains–less than a handful of robots on other planets, tens of millions of robots vacuuming people’s floors, thousands of robots in the military for forward reconnaissance and for handling improvised explosive devices, robots in thousands of factories around the world working side by side with people, and many hundreds of robots in research labs all over the world, used for experiments in manipulation. I think it is fair to say that companies that I have cofounded have put more AI into more robots, and in more domains of application, than anyone else, ever.

All of these robots have had some level of AI, but none come remotely close to what the popular press seems to believe about robots and what many prognosticators warn about, sometimes as imminent dangers from AI and robots.

There seems to me to be a misconnect here.

Right now I believe we are in an AI bubble. And I believe that should it not burst, it will certainly deflate before too long. The existence of this bubble makes it hard for all sorts of people to know what to believe about the near term for AI and robotics. For some people the questions surrounding AI and robotics are simply an intellectual curiosity. For some it is a very real question about what their job prospects might look like in just a few years. For executives at companies and for those in leadership positions in governments and the military, it is a fraught time understanding the true promise and dangers of AI. Most of what we read in the headlines, and from misguided well meaning academics, including from physicists and cosmologists, about AI is, I believe, completely off the mark.

In debates about how quickly we will realize in practice the AI and robotics of Hollywood I like to see myself as the voice of reason. I fear that I am often seen as the old fuddy-duddy who does not quite get how powerful AI is, and how quickly we are marching towards super intelligence, whatever that may be. I am even skeptical about how soon we will see self driving6 cars on our roads, and have been repeatedly told that I “just don’t understand”.

Concomitantly I am critical of what I see as scare mongering about how powerful AI will soon be, especially when it is claimed that we as humans must start taking precautions against it now. Given the sorts of things I saw in and around Fukushima Daiichi, I understand a general fear of technology that is not easily understood by non-experts, but I do want us humans to be realistic about what is scary, and what is only imagined to be scary.

I am even more critical about some of the arguments about ethical decisions that will robots will face in just the next few years. I do believe that there are plenty of ethical decisions facing us humans as we deploy robots in terms of the algorithms they should be running, but it is far, far, from having robots making ethical decisions on the fly. There will be neither benevolent AI nor malevolent AI in the next few decades where AI systems have any internal understanding of those terms. This is a research dream, and I do not criticize people thinking about it as research. I do criticize them thinking that this research is going to turn into reality any time soon, and talking about “regulations” on what sort of AI we can build or even research as a remedy for imagined pitfalls. Instead we will need to worry about people using technology, including AI, for malevolent purposes, and we should encourage the use of technology, including AI, for benevolent purposes. AI is nowhere near ready enough to make any fundamental difference in this regard.

Why am I seen as an outlier in my opinions of where AI is going?  In my dealings with individual researchers in academia and industrial research laboratories, and in my discussions with C-level executives at some of the best known companies that use AI, I find much common ground, and general agreement with my positions. I am encouraged by that. I want to share with readers of this blog what the basis is for my estimations of where we are in deploying AI and robotics, and why it is still hard; I will expand on these arguments in the next few essays that I  post after this one.

Perhaps after reading my forthcoming essays you will conclude that I am an old fuddy-duddy. Or, perhaps, I am a realist. I’ll be posting a lot of long form essays about these topics over the next few months. People who read them will get to decide for themselves where I fall in the firmament of AI.

Now, at the same time, we have only been working on Artificial Intelligence and robotics for just a few decades. Already AI has started to have real impact on our lives. There will be much more that comes from AI and robotics. For those who are able to see through what is hype and what is real there are going to be great opportunities.

There are great opportunities for researchers who concentrate on the critical problems that remain, and are able to prioritize what research will have the greatest impact.

For those who want to start companies in AI and robotics, understanding what is practical and matches what the market will eagerly accept, there is again great opportunity. We should expect to see many large and successful companies rise up in this space over the next couple of decades.

For those are willing to dare greatly, and who want to make scientific contributions for the ages there is so much at such a deep level that we do not yet understand that there is plenty of room for a few more Ada Lovelaces, Alan Turings, Albert Einsteins, and Marie Curies to make their marks.

Hubris and humility

For four years starting in 1988 I co-taught the M.I.T. introductory Artificial Intelligence (AI) class, numbered then and still now 6.034, with Professor Patrick Henry Winston. He still teaches that class which has only gotten better, and it is available online⁠7 for all to experience.

Back then Patrick used to start the first class with a telling anecdote. Growing up in Peoria, Illinois, Patrick had at one time had a pet raccoon. Like others of its species Patrick’s raccoon was very dexterous, and so hard to keep in a cage as it was usually clever enough to find a way to open the door unless it was really locked tight. Patrick would regale the class with how intelligent his raccoon had been. Then he would dead pan “but I never expected it to be smart enough to build a copy of itself”.

This was caution for humility. Then, as now, there was incredible promise, and I would say hype around Artificial Intelligence, and so this was Patrick’s cautionary note. We might think we were just around the corner from building machines that were just as smart, by whatever measure, as people, but perhaps we were really no more than dexterous raccoons with computers.

At that time AI was not a term that ever appeared in the popular press, IBM went out of its way to say that computers could not think, only people could think, and AI was not thought of appropriate stature to be part of many computer science departments. Patrick’s remarks led me to wonder out loud whether we were overwhelming ourselves with our own hubris. I liked to extend Patrick’s thinking and wondered about super-intelligent aliens (biological or otherwise) observing us from high orbit or further afield. I imagined them looking down at us, like we might look at zoo animals, and being amused by our cleverness, but being clear about our limitations. “Look at those ones in their little AI Lab at M.I.T.! They think they are going to be able to build things as smart as themselves, but they have no idea of the complexities involved and their little brains, not even with help from their computers (oh, they have so got those wrong!), are just never going to get there. Should we tell them, or would that be unkind to dash their little hopes? Those humans are never going to develop nor understand how intelligence works.”

Still, today, Patrick Winston’s admonition is is a timely caution for us humans. A little humility about the the possible limits of our capabilities is in order. Humility about the future of AI and also Machine Learning (ML) is in desperate short supply. Hubris, from some AI researchers, from venture capitalists (VCs), and from some captains of technology industries is dripping thick and fast. Often the press manages to amplify the hubris as that is what makes a good attention grabbing story.



1 “Fast, Cheap and Out of Control: A Robot Invasion of the Solar System”, Rodney A. Brooks and Anita M. Flynn, Journal of the British Interplanetary Society, 42(10)478–485, October 1989.

2 Flesh and Machines, Rodney A. Brooks, Pantheon, New York, 2002.

3 “The Day After Fukushima”, Danielle DeLatte, Space Safety Magazine, (7)7–9, Spring 2013.

4 During the 2014 Abe/Obama summit, as usual, a Japanese humanoid robot built by a Japanese auto maker was brought out to interact with the US President. The robot kicked a soccer ball towards President Obama and he kicked it back, to great applause. Then President Obama turned to his hosts and asked, not so innocently, so is that robot autonomous or tele-operated from behind the scenes? That is an educated and intellectually curious President.

5 Partially in response to the Fukushima disaster the US Defense Advanced Research Projects Agency (DARPA) set up a challenge competition for robots to operate in disaster areas. Japanese teams were entered in this competition; the first time there had been significant interaction between Japanese roboticists and DARPA–they were a very strong and welcome addition. The competition ran from late 2011 to June 5th and 6th of 2015 when the final competition was held. The robots were semi-autonomous with communications from human operators over a deliberately unreliable and degraded communications link. This short video focusses on the second place team but also shows some of the other teams, and gives a good overview of the state of the art in 2015. For a selection of greatest failures at the competition see this link. I was there and watched all this unfold in real time–it was something akin to watching paint dry as there were regularly 10 to 20 minute intervals when absolutely nothing happened and a robot just stood there frozen. This is the reality of what our robots can currently do in unstructured environments, even with a team of researchers communicating with them when the can.

6 I have written two previous blog posts on self driving cars. In the first I talked about the ways such cars will need to interact with pedestrians in urban areas, and many of the problems that will need to be solved.  In the second I talked about all the uncommon cases that we as drivers in urban situations need to face, such as blocked roads, temporarily banned parking, interacting with police, etc. In a third post I plan to talk about how self driving cars will change the natures of our cities. I do not view self driving cars as doomed by any of these problems, in fact I am sure they will become the default way for cars to operate in the lifetimes of many people who are alive today. I do, however, think that the optimistic forecasts that we have seen from academics, pundits, and companies are wildly off the mark. In fact that reality is starting to set in. Earlier this month the brand new CEO of Ford said that the previous goal of commercial self driving cars by 2021 was not going to happen. No new date was announced.

You can find the 24 lectures of 6.034 online. I particularly recommend lectures 12a and 12b on neural networks and deep neural networks to all those who want to understand the basics of how deep learning works–the only prerequisite is a little multi-variable differential calculus. Earlier in 2017 I posted a just slightly longer introduction than this paragraph.

[FoR&AI] Future of Robotics and Artificial Intelligence

rodneybrooks.com/forai-future-of-robotics-and-artificial-intelligence/

I plan on publishing  a set of essays on the future of robotics and Artificial Intelligence in the late summer and fall of 2017, perhaps extending in to 2018. I’ll list them all here as they come out. They are designed to be read as stand alone essays, and in any order, but I’ll order them here in my guess at the optimal order in which to read them.

The origins of “Artificial Intelligence” published on April 26, 2018.

Domo Arigato Mr. Roboto published on August 28, 2017.

The Seven Deadly Sins of Predicting the Future of AI published on September 7, 2017.

Machine Learning Explained published on August 28, 2017.

Steps Toward Super Intelligence I, How We Got Here, published July 15, 2018.

Steps Toward Super Intelligence II, Beyond the Turing Test, published July 15, 2018.

Steps Toward Super Intelligence III, Hard Things Today, published July 15, 2018.

Steps Toward Super Intelligence IV, Things to Work on Now, published July 15, 2018.

Experiments In Automobile UI/UX

rodneybrooks.com/experiments-in-automobile-uiux/

Our automobiles are getting three makeovers simultaneously, with promises of a fourth. There is more action in reinventing of automobiles all at once than there has been since the first ones crawled into existence.

First, our cars are turning electric, and the UK recently said that no new gas or diesel automobiles will be allowed on the road starting in 2040. Don’t be surprised to see more countries and states (e.g., California) follow suit.

Second, our cars are getting more driver assist features with lane change and backup audible warnings, automatic parking and lane changing, new forms of smart cruise control, new bumper-to-bumper traffic control options, etc. These are level 1 and level 2 autonomy (see my blog post from earlier in the year), with level 3 starting to show up just a little, and overly enthusiastic predictions of  levels 4 and 5 (again, you can see some of my thoughts on how I think that is further off than expected).

And third, we are getting new user interfaces in our cars, and that is the subject of this short post.

I drive a lot of different rental cars during any given year, so even though my own car is now eleven years old I get exposed to a lot of the new interfaces that are appearing in very standard level compact cars.

I’ve been driving the same rental car for two weeks as of today, and there are some things I really like about it. Most cars I have rented over the last two years have had a backup camera that has a live image on the LCD screen in the center of the dashboard. This is great, as it is usually a much better view than can be had by scanning all three rear view mirrors. The better versions of this feature show you an overlay of exactly where the car will go as you backup with the current setting of the steering wheel.  Here is a view from the screen in my car this morning:

This is so much better than a plain old rear view mirror. It shows me exactly what I might hit as I back up, and for me that is especially useful in a parking garage as I am just so great at hitting things while going backward…

But it is not all champagne and roses. After having this car for two weeks, twice this morning, while I was driving along the street, the following window popped up on that same screen, which is a touch screen, by the way:

How can this be a good idea, or a good User Interface feature? It could lead to a very bad User Experience! A window pops up to tell me I shouldn’t take my eyes of the road to deal with the interface. That is a good message. But not while I’m driving! Both times it made me take my eyes off the road to read what this warning was about, and then I needed to reach out and servo my finger to the “OK” virtual button to dismiss it. It was a real temptation to do it while driving. Exactly the thing it is warning against!

Bad UI. Potentially disastrous UX.

This reminded me of an interchange I saw on Facebook right after the newest Tesla came out. Someone complained about the lack of dials and knobs, and so much of the UI being put on the very big LCD that is in the middle of a Tesla dashboard. Another person chided that person, saying essentially, “get over it, we now live in the world of the iPhone, not the Blackberry”.

I thought that latter comment completely missed a real issue. The knobs and levers with their fixed positions and fixed meanings within any particular car, along with the tactile feedback that they give us allows us to do a lot of control operations without taking our eyes off the road at all. That is a very good thing until the task of driving is completely taken over by the car itself. We need our attention out on the road. Moving control functions, that are needed while in motion and while controlling the car, to a touch screen, is probably not a good idea. Being able to do things without using our eyes is a safety feature while driving (and while walking with an iPhone…).

[By the way, why do our turn indicator blinkers make a clicking sound? Because the original ones that were introduced in the 1950’s operated by running a current through and heating up a bi-metallic strip. As it heated up it bent until it hit a contact, hence the click, which then drained the current to the indicator lights on the left or right side of the car, allowing the bi-metatallic strip to cool down and repeat. Now cars simulate that same old clicking sound so that we know when the indicators are blinking.]

I suspect that we are all going to be guinea pigs over the next few years with auto-makers bringing out some really great new UI features, along with some real failures.

Be careful!!

 

 

Edge Cases For Self Driving Cars

rodneybrooks.com/edge-cases-for-self-driving-cars/

Perhaps through this essay I will get the bee out of my bonnet^{\big 1} that fully driverless cars are a lot further off than many techies, much of the press, and even many auto executives seem to think. They will get here and human driving will probably disappear in the lifetimes of many people reading this, but it is not going to all happen in the blink of an eye as many expect. There are lots of details to be worked out.

In my very first post on this blog I talked about the unexpected consequences of having self driving cars. In this post I want to talk about about a number of edge cases, which I think will cause it to be a very long time before we have level 4 or level 5 self driving cars wandering our streets, especially without a human in them, and even then there are going to be lots of problems.

First though, we need to re-familiarize ourselves with the generally accepted levels of autonomy that every one is excited about for our cars.

Here are the levels from the autonomous car entry in Wikipedia which attributes this particular set to the SAE (Society of Automotive Engineers):

  • Level 0: Automated system has no vehicle control, but may issue warnings.
  • Level 1: Driver must be ready to take control at any time. Automated system may include features such as Adaptive Cruise Control (ACC), Parking Assistance with automated steering, and Lane Keeping Assistance (LKA) Type II in any combination.
  • Level 2: The driver is obliged to detect objects and events and respond if the automated system fails to respond properly. The automated system executes accelerating, braking, and steering. The automated system can deactivate immediately upon takeover by the driver.
  • Level 3: Within known, limited environments (such as freeways), the driver can safely turn their attention away from driving tasks, but must still be prepared to take control when needed.
  • Level 4: The automated system can control the vehicle in all but a few environments such as severe weather. The driver must enable the automated system only when it is safe to do so. When enabled, driver attention is not required.
  • Level 5: Other than setting the destination and starting the system, no human intervention is required. The automatic system can drive to any location where it is legal to drive and make its own decision.

There are many issues with level 2 and level 3 autonomy, which might make them further off in the future than people are predicting, or perhaps even  forever impractical due to limitations on how quickly humans can go from not paying attention to taking control in difficult situations. Indeed as outlined in this Wired story many companies have decided to skip level 3 and concentrate on levels 4 and 5. The iconic Waymo (formerly Google) car has no steering wheel or other conventional automobile controls–it is born to be a level 4 or level 5 car. [This image is from Wikipedia.]

So here I am going to talk only about level 4 and level 5 autonomy, and not really make a distinction between them.  When I refer to an “autonomous car” I’ll be talking about ones with level 4 or level 5 autonomy.

I will make distinctions between cars with conventional controls so that they are capable of being driven by a human in the normal way, and cars like the Waymo one pictured above with no such controls, and I will refer to that as an unconventional car. I’ll use those two adjectives, conventional, and unconventional, for cars, and then distinguish what is necessary to make them practical in some edge case circumstances.

I will also refer to gasoline powered driverless cars versus all electric driverless cars, i.e., gasoline vs. electric.

Ride-sharing companies like Uber are putting a lot of resources into autonomous cars. This makes sense given their business model as they want to eliminate the need for drivers at all, thus saving their major remaining labor cost. They envision empty cars being summoned by a customer, driving to wherever that customer wants to be picked up, with absolutely no one in the car. Without that, having the autonomy technology doesn’t make sense to this growing segment of the transportation industry. I’ll refer to such an automobile, with no-one in it as a Carempty. In contrast, an autonomous car which has a conscious person in it, whether it is an unconventional car and they can’t actually drive it in the normal way, or whether it is a conventional car but they are not at all involved in the driving, perhaps sitting in the back seat, as Careless, as presumably that person shouldn’t have to care less about the driving other than indicating where they want to go.

So we have both an unconventional and a conventional Carempty and Careless, and perhaps they are gasoline or electric.

Many of the edge cases I will talk about here are based on the neighborhood in which I live, Cambridgeport in Cambridge, Massachusetts. It is a neighborhood of narrow one way streets, packed with parked cars on both sides of the road so that it is impossible to pass by if a car or truck stopped in the road. A few larger streets are two way, and some of them have two lanes, one in each direction, but at least one nearby two way street only has one lane–one car needs to pull over, somehow, if two cars are traveling on the opposite direction (the southern end of Hamilton Street in the block where the “The Good News Garage” of the well known NPR radio brothers “Click and Clack” is located).

HOW MUCH DRIVING CAN A NON-DRIVER DO?

In a conventional Careless a licensed human can take over the driving when necessary, unless say it is a ride sharing car, and in that case humans might be locked out of using the controls directly. For an unconventional Careless, like one of the Waymo cars pictured above, the human can not take over directly either. So a passenger in a conventional ride-sharing car, or in an unconventional car are in the same boat. But how much driving can that human do?

In both cases the human passenger needs to be able to specify the destination. For a ride-sharing service that may have been done on a smart phone app when calling for the service. But once in the car the person may want to change their mind, or demand that the car take a particular route–I certainly often do that with less experienced drivers who are clearly going a horrible way, often at the suggestion of their automated route planners. Should all this interaction be via an app? I am guessing, given the rapid improvements in voice systems, such as we see in the Amazon Echo, or the Google Home, we will all expect to be able to converse by voice with any autonomous car that we find ourselves in.

We’ll ignore for the moment a whole bunch of teenagers each yelling instructions and pranking the car. Let’s just think about a lone sensible mature person in the car trying to get somewhere.

Will all they be able to do is give the destination and some optional route advice, or will they be able to give more detailed instructions when the car is clearly screwing up, or missing some perceptual clue that the occupant can clearly recognize? The next few sections give lots of examples from my neighborhood that are going to be quite challenging for autonomous cars for many years to come, and so such advice will come in handy.

In some cases the human might be called upon to, or just wish to, give quite detailed advice to the car. What if they don’t have a driver’s license? Will the be guilty of illegally driving a car in that case? How much advice should they be allowed to give (spoiler alert, the car might need a lot in some circumstances)? And when should the car take the advice of the human? Does it need to know if the person in the car talking to it has a driver’s license?

Read on.

WHAT TO DO ABOUT A BLOCKED ROAD

In my local one-way streets the only thing to do if a car or other vehicle is stopped in the travel lane is to wait for it to move on. There is no way to get past it while it stays where it is.

The question is whether to toot the horn or not at a stopped vehicle.

Why would it be stopped? It could be a Lyft or an Uber waiting for a person to come out of their house or condominium. A little soft toot will often get cooperation and they will try to find a place a bit further up the street to pull over.  A loud toot, however, might cause some ire and they will just sit there. And if it is a regular taxi service then no amount of gentleness or harshness will do any good at all. “Screw you” is the default position.

Sometimes a car is stopped because the driver is busy texting, most usually when they are at an intersection, had to wait for some one to the cross in front of them, their attention wandered, they started reading a text, and now they are texting and have forgotten that they are in charge of an automobile. From behind one can often tell what they are up to by noticing their head inclination, even from inside the car behind. A very gentle toot will usually get them to move; they will be slightly embarrassed at their own illegal (in Massachusetts) behavior.

And sometimes it is a car stopped outside an eldercare residence building with someone helping a very frail person into or out of the car.  Any sort of toot from a stopped car behind is really quite rude in these circumstances, distressing for the elderly person being helped, and rightfully ire raising for the person taking care of that older person.

Another common case of road blockage is that a garbage truck stopped to pick up garbage. There are actual two varieties, one for trash, and one for recyclables. It is best to stop back a bit further from these trucks than from other things blocking the road, as people will be running around to the back of the truck and hoisting heavy bins into it. And there is no way to get these trucks to move faster than they already are. Unlike other trucks, they will continue to stop every few yards. So the best strategy is to follow, stop and go, until the first side street and take that, even if, as it is most likely a one-way street, it sends you off in a really inconvenient direction.

Yet a third case is a delivery truck. It might be a US Postal Service truck, or a UPS or Fedex truck, or sometimes even an Amazon branded truck. Again tooting these trucks makes absolutely no difference–often the driver is getting a signature at a house, or may be in the lobby of a large condominium complex. It is easy for a human driver to figure out that it is one of these sorts of trucks. And then the human knows that it is not so likely to stop again really soon, so staying behind this truck once it moves rather than taking the first side street is probably the right decision.

If on the other hand it is a truck from a plumbing service, say, it is worth blasting it with your horn. These guys can be shamed into moving on and finding some sort of legal parking space. If you just sit there however it could be many minutes before they will move.

A Careless automobile could ask its human occupant whether it should toot. But should it make a value judgement if the human is spontaneously demanding that it toot its horn loudly?

A Carempty automobile could just never toot, though the driver in a car behind it might start tooting it, loudly. Not tooting is going to slow down Carempties quite a bit, and texting drivers just might not care at all if they realize it is a Carempty that is showing even a little impatience. And should an autonomous car be listening for toots from a car behind it, and change its behavior based on what it hears? We expect humans to do so. But are the near future autonomous cars going to be so perfect already that they should take no external advice?

Now if Carempties get toot happy, at least in my neighborhood that will annoy the residents having tooting cars outside their houses at a much higher level than at the moment, and they might start to annoy the human drivers in the neighborhood.

The point here is that there is a whole lot of perceptual situations that a an autonomous vehicle will need to recognize if it is to be anything more than a clumsy moronic driver (an evaluation us locals often make of each other in my neighborhood…). As a class, autonomous vehicles will not want to get such a reputation, as the humans will soon discriminate against them in ways subtle and not so subtle. 

Maps DOn’T Tell the Whole Story

Recently I pulled out of the my garage and turned right onto the one way street that runs past my condominium building, and headed to the end of my single block street, expecting to turn right at a “T” junction onto another one way street. But when I got there, just to the right of the intersection the street was blocked by street construction, cordoned off, and with a small orange sign a foot or so off the ground saying “No Entry”.

The only truly legal choice for me to make was to stop. To go back from where I had come I needed to travel the wrong way on my street, facing either backwards or forwards, and either stopping at my garage, or continuing all the way to the street at the start of my street. Or I could turn left and go the wrong way on the street I had wanted to turn right onto, and after a block turn off onto a side street going in a legal direction.

A Careless might inform its human occupant of the quandry and ask for advice on what to do. That person might be able to do any of the social interactions needed should the Careless meet another car coming in the legal direction under either of these options.

But Carempty will need some extra smarts for this case.  Either hordes of empty cars eventually pile up at this intersection or each one will need to decide to break the law and go the wrong way down one of the two one way streets–that is what I had to do that morning.

The maps that a Carempty has won’t help it a whole lot in this case, beyond letting it know the minimum distance it is going to have to be in a transgressive state.

Hmmm.  It is OK for a Carempty to break the law when it decides it has to? Is it OK for a Careless to break the law when its human occupant tells it to? In the situation I found myself in above, I would certainly have expected my Careless to obey me and go the wrong way down a one way street. But perhaps the Careless shouldn’t do that if it knows that it is transporting a dementia patient.

The Police

How are the police supposed to interact with a Carempty?

While we have both driverful and driverless cars on our roads I think the police are going to assume that as with driverful cars they can interact with them by waving them through an intersection perhaps through a red light, stopping them with a hand signal at a green light, or just to allow someone to cross the road.

But besides being able to understand what an external human hand signaling them is trying to convey, autonomous cars probably should try to certify in some sense whether the person that is giving them those signals is supposed to be doing so with authority, with politeness, or with malice. Certainly police should be obeyed, and police should expect that they will be. So the car needs to recognize when someone is a police officer, no matter what additional weather gear they might be wearing. Likewise they should recognize and obey school crossing monitors. And road construction workers. And pedestrians giving them a break and letting them pass ahead of them. But should they obey all humans at all times? And what if in a Careless situation their human occupant tells them to ignore the taunting teenager?

Sometimes a police officer might direct a car to do something otherwise considered illegal, like drive up on to a sidewalk to get around some road obstacle. In that case a Carempty probably should do it. But if it is just the delivery driver whose truck is blocking the road wanting to get the Carempty to stop tooting at them, then probably the car should not obey, as then it could be in trouble with the actual police. That is a lot of situational awareness for a car to have to have.

Things get more complicated when it is the police and the car is doing something wrong, or there is an extraordinary circumstance which the car has no way of understanding.

In the previous section we just established that autonomous cars will sometimes need to break the law. So police might need to interact with law breaking autonomous cars.

One view of the possible conundrum is this cartoon from the New Yorker. There are two instantly recognizable Waymo style self driving cars, with no steering wheels or other controls, one a police car that has just pulled over the other car. They both had people in them, and the cop is asking the guy in the car that has just been pulled over, “Does your car have any idea why my car pulled it over?”.

If an autonomous car fails to see a temporary local speed sign and gets caught in a speed trap, how is it to be pulled over? Does it need to understand flashing blue lights and a siren, and does it do the pull to the side in a way that we have all done, only to be relieved when we realize that we were not the actual target?

And getting back to when I had to decide to go the wrong way down a one way street, what if a whole bunch of Carempties have accumulated at that intersection and a police officer is dispatched to clear them out? For driverful cars a police officee might give a series of instructions and point out in just a few seconds who goes first, who goes second, third, etc. That is a subtle elongated set of gestures that I am pretty sure no deep learning network has any hope at the moment of intpreting, of fully understanding the range of possibilities that a police officer might choose to use.

Or will it be the case that the police need to learn a whole new gesture language to deal with driverless cars? And will all makes all understand the same language?

Or will we first need to develop a communication system that all police officers will have access to and which all autonomous cars will understand so that police can interact with autonomous cars? Who will pay for the training? How long will that take, and what sort of legislation (in how many jurisdictions) will be required?

Getting Towed

A lot of cars get towed in Cambridge. Most streets get cleaned on a regular schedule (different sides of the same street on different days), and if your car is parked there at 7am you will get towed–see the sign in the left image. And during snow emergencies, or without the right sticker/permit you might get towed at any time. And then there are pop-up no parking signs, partially hand written, that are issued by the city on request for places for moving vans, etc. Will our autonomous cars be able to read these? Will they be fooled by fake signs that residents put up to keep pesky autonomous cars from taking up a parking spot right outside their house?

If an unconventional Carempty is parked on the street, one assumes that it might at any time start up upon being summoned by its owner, or if it is a ride-share car when its services are needed. So now imagine that you are the tow truck operator and you are supposed to be towing such a car. Can you be sure it won’t try driving away as you are crawling under it connect the chains, etc., to tow it?  If a human runs out to move their car at the last minute you can see when things are going to start and adjust. How will it work with fully autonomous cars?

And what about a Carempty that has a serious breakdown, perhaps in its driving system, and it just sits there and can no longer safely move itself. That will need to be towed most likely. Can the tow truck operator have some way to guarantee that it is shut down and will not jump back to life, especially when the owner has not been contactable, to put it in safe mode remotely? What will be the protocols and regulations around this?

And then if the car is towed, and I know this from experience, it is going to be in a muddy lot full of enormous potholes in some nearby town, with no marked parking areas or driving lanes. The cars will have been dumped at all angles, higgledy-piggledy. And the lot is certainly not going have its instantaneous layout mapped by one of the mapping companies, providing the maps that autonomous cars rely on for navigation. To retrieve such a car a human is likely going to have to go do it (and pay before getting it out), but if it is an unconventional car it is certainly going to require some one in it to talk it through getting out of there without angering the lot owner (and again from experience, that is a really easy thing to do–anger the lot owner). Yes, in some distant future tow lots in Massachusetts will be clean, and flat with no potholes deeper than six inches, and with electronic payment systems, and all will be wonderful for our autonomous cars to find their way out.

Don’t hold your breath.

OTHER TRICKY SITUATIONS

What happens when a Carempty is involved in an accident? We know that many car companies are hoping that their cars will never be involved in an accident, but humans are dumb enough that as long as there are both human drivers and autonomous cars on the same streets, sometimes a human is going to drive right into an autonomous car.

Autonomous cars will need to recognize such a situation and go through some protocol. There is a ritual when a fender bender happens between two driverful cars. Both drivers stop and get out of their cars, perhaps blocking traffic (see above) and go through a process of exchanging insurance information. If one of the cars is an autonomous vehicle the the human driver can take a photo on their phone (technology to the rescue!) of the autonomous car’s license plate. But how is a Carempty supposed to find out who hit it? In the distant future when all the automobile stock on the road have transponders (like current airplanes) that will be relatively easy (though we will need to work through horrendous privacy issues to get there), but for the foreseeable future this is going to be something of a problem.

And what about refueling? If a ride-sharing car is gasoline powered and out giving rides all day, how does it get refueled? Does it need to go back to its home base to have a human from its company put in more gasoline? Or will we expect to have auto refueling stations around our cities? The same problem will be there even if we quickly pass beyond gasoline powered cars. Electric Carempties will still need to recharge–will we need to replace all the electric car recharging stations that are starting to pop up with ones that require no human intervention?

Autonomous cars are likely to require lots of infrastructure changes that we are just not quite ready for yet.

Impacts on the Future of Autonomous Cars

I have exposed a whole bunch of quandaries here for both Carempties and Carelesses. None rise to the moral level of the so called trolley problem (do I kill the one nun or seven robbers?) but unlike the trolley problem variants of these edge cases are very likely to arise, at least in my neighborhood. There will be many other edge case conundrums in the thousands, perhaps millions, of unique neighborhoods around the world.

One could try to have some general purpose principles that cars could reason from in any circumstances, perhaps like Asimov’s Three Laws^{\big 2}, and perhaps tune the principles to the prevailing local wisdom on what is appropriate or not. In any case there will need to be a lot of codifying of what is required of autonomous cars in the form of new traffic laws and regulations. It will take a lot of trial and error and time to get these laws right.

Even with an appropriate set of guiding principles there are going to be a lot of perceptual challenges for both Carempties and Carelesses that are way beyond those that current developers have solved with deep learning networks, and perhaps a lot more automated reasoning that any AI systems have so far been expected to demonstrate.

I suspect that to get this right we will end up wanting  our cars to be as intelligent as a human, in order to handle all the edge cases appropriately.

And then they might not like the wage levels that ride-sharing companies will be willing to pay them.



^{\big 1}But maybe not.  I may have one more essay on how driverless cars are going to cause major infrastructure changes in our cities, just as the original driverful cars did. These changes will be brought on by the need for geofencing–something that I think proponents are underestimating in importance.

^{\big 2}Recall that Isaac Asimov used these laws as a plot device for his science fiction stories, by laying out situations where these seemingly simple and straightforward laws led to logical fallacies that the story proponents, be they robot or human, had to find a way through.

Is War Now Post Kinetic?

rodneybrooks.com/is-war-now-post-kinetic/

When the world around us changes, often due to technology, we need to change how we interact with it, or we will not do well.

Kodak was well aware of the digital photography tsunami it faced but was not able to transform itself from a film photography company until too late, and is no more. On the other hand, Pitney Bowes started its transformation early from a provider of mail stamping machines to an eCommerce solutions company and remains in the S&P 500.

Governments and politicians are not immune from the challenges that technological change produces on the ground, and former policies and vote getting proclamations may lag current realities^{\big 1}.

I do wonder if war is transforming itself around us to being fought in a non-kinetic way, and which nations are aware of that, and how that will change the world going forward. And, importantly for the United States, what does that say about what its Federal budget priorities should be?

A Brief History of Kinetic War

The technology of war has always been about delivering more kinetic energy, faster, more accurately and with more remote standoff from the recipient of the energy, first to human bodies, and then to infrastructure and supply chains.

New technologies caused changes in tactics and strategies, and many of them eventually made old technologies obsolete, but often a new technology would co-exist with one that it would eventually supplant for long periods, even centuries.

One imagines that the earliest weapons used in conflicts between groups of people were clubs and axes of various sorts. These early wars were fought in close proximity, delivering kinetic blows directly to another’s body.

By about 4,400 years ago the first copper daggers appeared, and by 3,600 years ago, bronze swords appeared, allowing for an attack at a slightly longer distance, perhaps out of direct reach of the victim. Even today our infantries are equipped with bayonets on the ends of guns to deliver direct kinetic violence to another’s body through the use of human muscles. With daggers and swords the kinetic blows could be much more deadly as they needed less human energy to cause bleeding.

Simultaneously the first “stand off” weapons were developed; bows and arrows 12,000 years ago, most likely with a very limited range. The Egyptians had bows with a range of 100 meters a little less than 4,000 years ago. A bow stores the energy from human muscle in a single drawing motion, and then delivers it all in a fraction of a second. These weapons did not eliminate hand to hand combat, but they did allow engagement from a distance. With the introduction of horses and later chariots, there was added the element of speed of closing from too far away to engage to being in engagement range very quickly. These developments were all aimed at getting bleed-producing kinetic impacts on humans from a distance.

A little less than 3,000 years ago war saw a new way to use kinetic energy; thermally. No longer was it just the energy of human muscles that rained down on the enemy, but that from fire. First from burning crops, but soon by delivering  burning objects via catapults and other throwing devices. Those throwing devices started out just delivering heavy weights, though the muscle energy of many people stored over many minutes of effort. But once burning objects were being thrown they could deliver the thermal energy stored in the projectile, as well as unleash more thermal energy by setting things on fire in the landing area.

During the 8th to 16th century, hurled anti-personnel weapons, those aimed at individual people, were developed where projectiles full of hot pitch, oil, or resin, were thrown by mechanical devices, again with stored human energy, intended to maim and disable an individual human that they might hit.

The arrival of chemical explosives ultimately changed most things about warfare, but there was a surprisingly long coexistence with older weapons. The earliest form of gunpowder was developed in 9th century China, and it reached Europe courtesy of the Mongols in 1241. The cannon, which provided a way of harnessing that explosive power to deliver high amounts of kinetic energy in the form of metal or stone balls provided both more distant standoff and more destructive kinetics, and was well developed by the 14th century, with the first man portable versions coming of age in the 15th century.

But meanwhile the bow and arrow made a come back, with the English longbow, traditionally made from yew (and prompting a European wide trade network in that wood), having a range of 300 meters in the 14th and 15th centuries. It was contemporary with the cannon, but the agility of it being carried by a single bowman led to it being the major reason for victory in a large scale battle as late as the Battle of Agincourt in 1415.

The cannon changed the nature of naval warfare, and naval warfare itself was about logistics and supply lines, and later being a mobile platform to pound installations on the coast from the safety of the sea. Ships also changed over time due to new technologies for their propulsion, from oars, to sails, to steam, and ultimately to nuclear power, making them faster and more reliable. Meanwhile the mobile cannon was developed into more useful sorts of weapons, and with the invention of bullets (which combined the powder and projectile into a compact pre-manufactured expendable device), guns and then machine guns became the preferred weapon of the ground soldier.

Each of these technological developments improved upon the delivery of kinetic energy to the enemy, over time, in fits and starts making that delivery faster, more accurate, more energetic, and with more distant standoff.

Rarely were the new technologies adopted quickly and universally, but over time they often made older technologies completely obsolete. One wonders how quickly people noticed the new technologies, how they were going to change war completely, and how they responded to those changes.

Latter Day WAR

In the last one hundred or so years, from the beginning of the Great War, also known as World War I, we have seen continued technological change in how kinetic energy is delivered during conflict. In the Great War we saw both the introduction of airplanes, originally as intelligence gathering machine conveyances, but later as deliverers of bullets and bombs, and the introduction of tanks. Even with mechanization, the United Army still had twelve horse regiments, each of 790 horses, at the beginning of World War II. They were no match for tanks, and hard to integrate with tank units, so eventually they were abolished.

By the end of World War II we had seen both the deployment of missiles (the V1 and V2 by Germany), and nuclear weapons (by the United States). Later married together, nuclear tipped missiles became the defining, but unused, technology that redefined the nature of war between superpowers. Largely that notion is obsolete, but North Korea, a small poor country, is actively flirting with it again these very days.

Another innovation in World War II, practiced by both sides, was massive direct kinetic hits on the civilian populations of the enemy, delivered through the air. For the first time kinetic energy could be delivered far inside territory still held by the enemy, and damage to infrastructure and morale could be wrought without the need to invade on the ground. Kinetically destroying large numbers of civilians was also part of the logic of MAD, or Mutually Assured Destruction, of the United States and the USSR pointing massive numbers of nuclear tipped missiles at each other during the cold war.

Essentially now war is either local engagements between smaller countries, or asymmetric battles between large powers and smaller countries or non-state actors. The dominant approach for the United States is to launch massive ship and air based volleys of Tomahawk Cruise Missiles, with conventional kinetic war heads, to degrade the war fighting infrastructure in the target territory, and then boots on the ground. The other side deploys harassing explosives both as booby traps, and to target both the enemy and local civilians through using human suicide bombers as a stand off mechanism for those directing the fight. As part of this asymmetry the non-state actors continually look for new ways to deliver kinetic explosions on board civilian aircraft which has had the effect of making air travel worldwide more and more unpleasant for the last 16 years.

In slow motion each class of combatant changes their behavior to respond to new, and past, technologies deployed or threatened by the other side.

But over the whole history of war, rulers and governments have had to face the issue of what war to prepare for and where to place their resources. When should a country stop concentrating on sources of yew and instead invest more heavily in portable cannons? When should a country give up on supporting regiments of horses? When should a country turn away from the ruinous expense of yet higher performance fighter planes whose performance is only needed to engage other fighter planes and instead invest more heavily in cruise missiles and drones with targeted kinetic capabilities?

How should a country balance its portfolio of spending on the old technologies of war, and putting enough muscle behind the new technologies so that it can ride up the curve of the new technology, defending against it adequately, and perhaps deploying it itself.

BUT HAS A NEW FORM OF WAR ARRIVED?

In the late nineteenth century fortunes were made in chemistry for materials and explosives. In the early part of the twentieth century extraordinary wealth for a few individuals came from coal, oil, automobiles, and airplanes. In the last thirty years that extraordinary wealth has come to the masters of information technology through companies such as Microsoft, Apple, Oracle, Google, and Facebook. Information technology is the cutting edge. And so, based on history, one should expect that technology to be where warfare will change.

Indeed, we saw in WW II the importance of cryptography and the breaking of cryptography, and the machines built at Bletchley Park in service of that gave rise to digital computers.

In the last few years we have seen how our information infrastructure has been attacked again and again for criminal reasons, with great amounts of real money being stolen, solely in cyberspace. Pacifists^{\big 2} might say that war is just crime on an international scale, so one should expect that technologies that start out as part of criminal enterprises will be adopted for purposes of war.

We have seen over the last half dozen years how non-state actors have used social media on the Internet to recruit young fighters from across the world to come and partake in their kinetic wars where those recruiters reside, or to wage kinetic violence inside countries far removed physically from where the recruiters reside. The Internet has been a wonderful new stand off tool, allowing distant ring-masters to burrow in to distant homelands and detonate kinetic weapons constructed locally by people the ring-masters have never met in person. This has been an unexpected and frightening evolution of kinetic warfare.

In the early parts of this decade a malicious computer worm named Stuxnet, most probably developed by the US and Israel, was deployed widely though the Internet. It infected Microsoft operating systems, and sniffed out whether they were talking to Siemens PLCs (Programmable Logic Controllers), and whether they were controlling nuclear centrifuges. Then it slowly degraded those centrifuges while simulating reports that said all was well with them. It is believed that this attack destroyed one fifth of Iran’s centrifuges. Here a completely cyber attack, with standoff all the way back to an office PC, was able to introduce a kinetic (slow though it may have been) attack in the core of an adversary’s secret facilities. And it was aimed at the production of the ultimate kinetic weapon, nuclear bombs. War is indeed evolving rapidly.

But now in the 2016 US presidential election, and again in the 2017 French presidential election we have seen, and all the details are not yet out, a glimpse of a future warfare where kinetic warfare is not used at all. Nevertheless it has been acts of war. US intelligence services announced in 2016 that there had been Russian interference in the US election.  The whole story is still to come out, but in both the US and French elections there were massive dumps of cyber-stolen internal emails from one candidate’s organization, timed exquisitely in both cases down to just a few minutes’ window of maximum impact. This was immediately, minutes later, followed by seemingly unrelated thousands of people looking through those emails claiming clues to often ridiculous malevolence. In both elections the mail dumps included faked emails which had sinister interpretations, uncovered by the armies of people looking through the emails for a smoking gun. These attacks most probably changed the outcome of the US election, but failed in France. This is post kinetic war waged in a murky world where the citizens of the attacked country can never know what to believe.

Let us be clear about the cleverness and monumental nature of these attacks. An adversary stands off, thousands of miles away, with no physical intrusion, and changes the government of its target to be more sympathetic to it than the people of the target country wanted. There are no kinetic weapons. There are layers of deception and layers of deniability. The political system of the attacked country has no way to counteract the outcome desired and produced by the enemy. The target country is dominated by the attacking adversary. That is a successful post kinetic war.

Technology changes how others act and how we need to act. Perhaps the second amendment to the US Constitution, allowing for an armed civilian militia to fight those who would destroy our Republic, is truly obsolete. Perhaps the real need is to equip the general population of the United States with tools of privacy and cyber security, both at a personal level, and in the organizations where they work. Just as WW II showed the obsolescence of physical borders to protect against kinetic devices raining from the sky, so too now we have seen that physical borders no longer protect our fundamental institutions of civil society and of democracy.

We need to learn how to protect ourselves in a new era of post kinetic war.

We see a proposed 2018 US Federal budget building up the weapons of kinetic war way beyond their current levels. Kinetic war will continue to be something we must protect against–it will remain an avenue of attack for a long time. We saw above how the English long bow was still a credible weapon, coexisting with cannon and other uses of gun powder for centuries, though now its utility is well gone.

However, we must not give up worrying about kinetic war, but we must start investing in strength and protection against a new sort of post kinetic war that has really only started in the last twelve months. With $639B slated for defense in the proposed 2018 budget, and even $2.6B for a border fence, surely we can spend a few little billions, maybe even just one or two, on figuring out how to protect the general population from this newly experienced form of post kinetic war. I have recommendations^{\big 3}.

We don’t want the United States to have its own Kodak moment.



^{\big 1}For instance, in just six months from this last October to April, more jobs were lost in retail in the US than the total number of US coal jobs. Not only did natural gas, wind, and solar technology decimate coal mining, jobs never to return, but information technology has enabled fulfillment centers, online ordering, and delivery to the home, completely decimating the US retail sector, a sector that is many times bigger than coal.

^{\big 2}I do not count myself as a pacifist.

^{\big 3}Where in the Federal Government should such money be spent? The NSA (National Security Agency) has perhaps the most sophisticated group of computer scientists and mathematicians working on algorithms to wage and protect against cyber war. But it is not an agency that shares that protection with the general population and businesses, just as the US Army does not protect individual citizens or even recommend how they should protect themselves. No, the agency that does this is NIST, the National Institute of Standards and Technology, part of the Department of Commerce.  It provides metrology standards which enable businesses to have a standard connection to the SI units of measurement.  But it also has (with four Nobel prizes under its belt) advanced fundamental physics so that we can measure time accurately (and hence have working GPS), it has been a key contributor, through its measurements of radio wave propagation. to the 3G, 4G, and coming 5G standards for our smart phones, and it is contributing more and more to biological measurements necessary for modern drug making.  But for the purpose of this note its role in cybersecurity is omni important. NIST has provided a Cybersecurity Framework for businesses, now followed by half of US companies, giving them a set of tools and assessments to know whether they are making their IT operations secure. And, NIST is now the standards generator and certifier for cryptography methods.  The current Federal budget proposal makes big cuts to NIST’s budget (in the past its total budget has been around $1B per year).  Full disclosure: I am a member of NIST’s Visiting Committee on Advanced Technology (VCAT). That means I see it up close. It is vitally important to the US and to our future. Now is not the time to cut its budget but to support it as we find our way in our future of war that is post kinetic.