On January 1^st, 2018, I made predictions about self driving cars, Artificial Intelligence, machine learning, and robotics, and about progress in the space industry. Those predictions had dates attached to them for 32 years up through January 1^st, 2050.

I made my predictions because at the time I saw an immense amount of hype about these three topics, and the general press and public drawing conclusions about all sorts of things they feared (e.g., truck driving jobs about to disappear, all manual labor of humans about to disappear) or desired (e.g., safe roads about to come into existence, a safe haven for humans on Mars about to start developing) being imminent. My predictions, with dates attached to them, were meant to slow down those expectations, and inject some reality into what I saw as irrational exuberance.

I was accused of being a pessimist, but I viewed what I was saying as being a realist. Today, I am starting to think that I too, reacted to all the hype, and was overly optimistic in some of my predictions. My current belief is that things will go, overall, even slower than I thought four years ago today. That is not to say that there has not been great progress in all three fields, but it has not been as overwhelmingly inevitable as the tech zeitgeist thought on January 1^st, 2018.

As part of self certifying the seriousness of my predictions I promised to review them, as made on January 1^st, 2018, every following January 1^st for 32 years, the span of the predictions, to see how accurate they were. This is my fourth annual review and self appraisal, following those of 2019, 2020, and 2021. I am an eighth of the way there!  Sometimes I throw in a new side prediction in these review notes.

The biggest news across these three fields this year is what appears to be a breakthrough into space tourism. For a few minutes at one point this year there were 19 people weightless in space at the same time and 8 of them were not professional astronauts. That happened on December 11^th. There were six people aboard Blue Origin’s New Shepherd in a sub-orbital flight, three Chinese astronauts on the Chinese Tiangong space station, seven regular crew members on the International Space Station (three from the US, two from Russia, and one each from Japan and the European Space agency), and three visitors to that same station, one a professional Russian astronaut, accompanying a Japanese billionaire and his publicity person. As we will see below there were two other orbital flights and three other sub-orbital flights with tourists on board in 2021.

UPDATE of 2019’s Explanation of Annotations

As I said in 2018, I am not going to edit my original post, linked above, at all, even though I see there are a few typos still lurking in it. Instead I have copied the three tables of predictions below from 2021’s update post, and have simply added a total of twelve comments to the fourth columns of the three tables. As with last year I have highlighted dates in column two where the time they refer to has arrived.

I tag each comment in the fourth column with a Cyan (#00ffff) colored date tag in the form yyyymmdd such as 20190603 for June 3^rd, 2019. In 2022 I have started highlighting the new text put in for the current year in LemonChiffon (#fffacd) so that it is easy to pick out this year’s updates.

The entries that I put in the second column of each table, titled “Date” in each case, back on January 1^st of 2018, have the following forms:

NIML meaning “Not In My Lifetime, i.e., not until beyond December 31^st, 2049, the last day of the first half of the 21^st century.

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

BY some date, meaning “By” that date.

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

For now I am coloring those statements when it can be determined already whether I was correct or not.

I have started using LawnGreen (#7cfc00) for those predictions which were entirely accurate. For instance a BY 2018 can be colored green if the predicted thing did happen in 2018, as can a NET 2019 if it did not happen in 2018 or earlier. There are five predictions now colored green, the same ones as last year, with no new ones in January 2020.

I will color dates Tomato (#ff6347) if I was too pessimistic about them. Last year I colored one date Tomato. If something happens that I said NIML, for instance, then it would go Tomato, or if in 2020 something already had happened that I said NET 2021, then that too would have gone Tomato.

If I was too optimistic about something, e.g., if I had said BY 2018, and it hadn’t yet happened, then I would color it DeepSkyBlue (#00bfff). None of these yet either. And eventually if there are NETs that went green, but years later have still not come to pass I may start coloring them LightSkyBlue (#87cefa). I did that below for one prediction in self driving cars this year.

In summary then: Green splashes mean I got things exactly right. Red means provably wrong and that I was too pessimistic. And blueness will mean that I was overly optimistic.

So now, here are the updated tables.

Self Driving Cars

Spoiler alert: very little movement in deployment of actual, for real, self driving cars.

Way back four years ago when I made my predictions about “self driving cars” that term meant that the cars drove themselves, and that there was no one in the loop at a company office, or following in a chase car, or sitting in the drive or passenger seat ready to take over or punch a big red button. As I documented in last year’s update the AV company’s conveniently neglect to mention these uncomfortable truths when they give press releases about their great progress. Even tech savvy people do not realize how very far all the breathless press stories about self driving cars being deployed are from being accurate about what is really going on. Last year I gave nine questions that every reporter should ask at any announcement by an AV car company. I have started seeing some similar questions being asked by more savvy reporters this last year. I have no idea whether any of them read my blog.

As an example of how companies gloss ofter things, if you carefully watch Waymo’s breathless “ooh” and “ah” filled video about their first deployment in San Francisco you will occasionally see the hands and knees of the safety driver sitting in the driver’s seat, conveniently unmentioned in the video that is trying to give the impression that their service is deployed. If you carefully read about the Chandler, Arizona, deployment and watch videos from there you will see how there is constant contact with people watching from home base, there are rescue vehicles that come and deposit a human driver to take over in some cases, and most importantly that the scale of the number of rides they give per week is tiny, way too few to be anything but a giant cash drain. That is two years after it was announced that the service was real and deployed. Nope.  Still very early days.

And (h/t Mohamed Amer) here is Cruise(GM)’s very first driverless pickup less that two months ago, in San Francisco.  Note that the passenger has two Cruise employees talking to him  through the car. The passenger is also a Cruise employee, and notes that he was not allowed by Cruise to bring his three year old along for the ride. And there is a camera person at both the pick up and drop off location. No indication of how far the vehicle drove driverless before or after this ride, nor whether there is a chase vehicle (with a safety person on board ready to stop the driverless vehicle) or not, nor whether it could operate at any other pick up or drop off location (Steve Jobs’ demos of new Apple technology were famously scripted down to every key press as otherwise he would hit some bug or the other).  This is an important step towards real deployment, but it is a long long way from real deployment.

The Hubris Of Self Driving Cars

In fact driverless cars, with safer roads, have been predicted again and again, for over 80 years, and it is always just around the corner. I recommend the book Autonorama by Peter Norton for some balance to the recent hype that really, really, autonomous cars are just around the corner once again.  It will make many techno people uncomfortable and even angry.  But it is good to have your internal beliefs challenged once in a while.

Norton points out that many companies have talked up the imminence of autonomous cars for a long time. Here is the list he documents for one such company, GM (which bought Cruise in 2016):

• 1939 World’s Fair at their Futurama exhibit, GM promised autonomous vehicles by 1960.
• 1964 World’s Fair in Futurama II, GM promised it again.
• 2010 GM and SAIC (Shanghai Automotive Industry Corporation) promised that it would arrive by 2030 as Xing! (autonomous Shanghai)
• 2017 GM 2017 Sustainability Report: Zero Crashes. Zero Emissions. Zero Congestion.

All worthy goals, but worthy goals doesn’t mean we yet no how to do it.  We certainly didn’t in 1939 and 1964. And we certainly didn’t know how to do it when Waymo started (inside Google) back in January 2009 (thirteen years ago and counting…).

We have seen the long tail of difficult situations delay deployments for years.  We know that humans can drive cars with “only” 35,000 annual fatalities just in the United States. We don’t know yet whether any of the sensor suites from any of the companies (all are different from the capabilities of human eyes) are sufficient to reach that goal or whether the presence of significant numbers of AVs on roads shared with humans will so impact the dynamics of driving that things get much worse.  We can have high faluting goals of zero this and zero that, but we actually have no idea whether the real world will allow that to happen without significant changes in approach.  WE JUST DON’T KNOW YET.

What we do know is that human drivers have many different capabilities, perceptual, reasoning, and an apparent ability to model what is happening and make predictions. Pure learning approaches are unlikely to capture that capability, so systems based largely on learning are making a big unproven bet. That hasn’t stopped massive funding going into two new startups in the last year, one for cars (Wayve) and one for trucks (Waabi) that try to solve the whole problem from pretty much a pure learning perspective. Didn’t we see this movie before? (Again, thanks to Mohamed Amer for pointing me at these two.)

Industry Misses

I believe that the hubris surrounding self driving cars has lead to delays in actual safety progress we could have already deployed if we had not been so distracted. I talk about this in the context of my own hubris after the table of the state of my predictions just below.

But first I will once again present a summary of predictions made by industry leaders that I scraped from a driverless future web site back in 2017. The site still exists but the last updates to it were made in late 2018. If you click on the predictions page you will see longer explanations of all these predictions that I gathered from it.

Each line is an individual prediction with the year it was made in parentheses. The orange arrows indicate that the companies later released updated later dates or cancelled their predictions completely, and that I personally noticed those updates. There may well be more that I have missed.

The years in blue indicate when the industry leaders thought these predictions would come to pass. I have highlighted all the dates up through 2021, now numbering 17 of the 23 predictions. Not one of them has happened or is even close to happening.

Hubris on a mass delusion scale. Audi fully autonomous by 2017?  That is Teslan in its delusion level.

Hubris

From my comments above the previous table it is clear that I think there has been too much hubris around driverless cars. But as I went through updating which predictions had stood the test of time I realized that I too had been guilty of hubris in my predictions.

There are two predictions about the earliest date that something might happen that I am now sure will not happen.  And it is not because of technical difficulty but because no one is working on them.

The first is about having dedicated lanes on freeways with special infrastructure installed that cars can interact with to make autonomous driving easier. Back in 1997 there was a project called the National Automated Highway System Consortium to do just that.  But when the hubris of fully self-driving cars made the need for external help for AVs seem redundant, such projects disappeared.

The second is the idea of leaving your car at the entrance to a garage and having it drive itself to park in a much tighter space than you could do.

Both these approaches to getting some sort of autonomy deployed in just a few years seem entirely doable to me, and I still believe that.  But the lure of full self driving (as in full self driving, not merely as a marketing ploy) drove everyone away from these approaches.

My hubris was to think that others would see things the same way and would work on them.  Given that no one was working on them do I really deserve credit for predicting they wouldn’t happen for a while?

Personal Gloating, An Irresistible Sin!!

Here I go.  For the last two years I have had a little rant that in April of 2019 the CEO of Tesla had said there would be one million autonomous Tesla taxis on the road by the end of 2020. Kai Fu Lee and I had agreed at the time to eat all such taxis on December 31^st, 2020. I.e., actually ingest them. All of them.  Of course there were zero such autonomous Tesla taxis on the road at the end of 2020. And a year later the number is still zero.

Note that in the last two weeks Tesla has agreed to eliminate the ability of drivers of their vehicles to play video games on the car’s display screen while the car is in motion. This was under regulatory pressure, indicating that “Full Self Driving” software is still not fully self driving. The latest spin from Tesla is that of course the words in the name of the software, “Full Self Driving”, could not possibly be interpreted to mean full self driving.

Defending Against Slings And Arrows, Another Irresistible Sin!!

Earlier this year someone used the fact that my prediction for when electric vehicles would account for 30% of automobile sales in the US as being no early than the year 2027 (see table above) as proof that I am a pessimist. Therefore, they said, my predictions about autonomous vehicles could not be trusted.

EV sales in the US were 1.7% of the total market in 2020 (up from 1.4% in 2019). We’ll need four doublings of that in seven years to get to 30%. It may happen. It may happen sooner than 2027.  But not by much. It would be a tremendous sustained growth rate that we have not yet seen.

Robotics, AI, and Machine Learning

With respect to my predictions for AI and ML there are only three that come close to being in play this year, either in terms of work that was done that impacts my predictions, or the date is close to when I said that something would or would not happen. I have annotated the table below in those three places; the next big thing, change in perspective of how to measure AI success, and robots that can really get around in our houses in a general purpose way.

The Next Big Thing

Back in 2018 I predicted that “the next big thing”, to replace Deep Learning, as the go to hot topic in AI would arrive somewhere between 2023 and 2027.  I was convinced of this as there has always been a next big thing in AI.  Neural networks have been the next big thing three times already. But others have had their shot at that title too, including (in no particular order) Bayesian inference, reinforcement learning, the primal sketch, shape from shading, frames, constraint programming, heuristic search, etc.

We are starting to get close to my window for the next big thing.  Are there any candidates?  I must admit that so far they all seem to be derivatives of deep learning in one way or another.  If that is all we get I will be terribly disappointed, and probably have to give myself a bad grade on this prediction.

So far the things that I see bubbling around and getting people excited are transformers, foundation models, and unsupervised learning.

Transformers provide a front end to deep learning that lets it handle sequential data rather than all input at once. This lets it be applied to natural language and to image sequences. We have seen tremendous hype about large language models using this mechanism. There is something going on there, but nothing as astounding or important as people imagine.

These language models are over interpreted by people as understanding what they are spitting out, especially when the press writes stories where they have cherry picked responses. But they come with incredible problems, including copyright violations, intellectual theft of code, and even outright life threatening danger when they find their way into consumer products. Tech companies have a real problem in rushing some of these systems to market.

Overall the will to believe in the innate superiority of a computer model is astounding to me, just as it was to others back in 1966 when Joseph Weizenbaum showed off his Eliza program which occupied a just a few kilobytes of computer memory. Joe, too, was astounded at the reaction and spent the rest of his career sounding the alarm bells. We need more push back this time around.

Foundation models are large trained models that start out as a basis for tuning particular applications.  There has been some self important announcements with a sort of me too feel (“Hey, I produced a foundation model too!!”), which don’t amount to much of an intellectual contribution.  If this turns out to be the next big thing I am going to have to rip off my mask of equanimity and revert to my natural state of being a grumpy old man.

Unsupervised learning is an idea that has been around for a long time.  Not a big intellectual jump to want to get it into deep learning–may be a hard technical problem, but not an intellectual breakthrough this time around.  (RIP Teuvo Kohonen who passed away last month.)

So, am I wrong, or was I just too optimistic about how soon the next big thing will show up?  I’ve given myself another 28 years of reviewing my predictions so perhaps there is time for something real and big to show up, and then I’ll be able to claim that I was simply too optimistic…

Beyond The Turing Test And Asimov’s Laws 

I start to see the serious technical press really trying to come to terms with judging whether systems are intelligent or not, and the number of articles questioning the Turing Test seem to be on the rise. Perhaps this is driven by the perceptions of what transformer based natural language systems can do. They are not intelligent but they can can fool an awful lot of the people an awful lot of the time.

As is often the case, Melanie Mitchell has a really good analysis, this time on the Turing Test and GPT-3. She is a renowned academic but she now also writes for Quanta Magazine.

Getting Around In A House

The first mass market robots that could get around reliably in ordinary people’s houses were introduced by my company iRobot at a two hundred dollar price point twenty years ago this year, in September of 2002. iRobot has now sold around 35 million Roombas and I am guessing that the follow on brands have together sold about the same number.

But we all know that Roombas could never be Rosie the Robot from the cartoon series The Jestsons. They are limited to a single flat floor and can’t get up high enough to really see objects that matter to humans in the houses that they bumble around in.  One of my predictions is that A robot that can navigate around just about any US home, with its steps, its clutter, its narrow pathways between furniture, etc. won’t even be a lab demo until 2026, and not be deployed for real until 2030 and at low cost in 2035.

Amazon just made a good and necessary step towards this capability with release of the Astro home robot. I got to spend time with some Astros before they were announced and I was impressed with the technical capabilities and at their price for all that technology. Whether they will be a successful product or not I can not hazard a guess.

The impressive step that Amazon has made is in getting a really reliable SLAM (Simultaneous Localization And Mapping) system that can quickly build a very good map of a house without any help from humans. It starts up in an unknown place in an unknown environment, and within minutes has a solid useable map. It is still somewhat low to the ground but its camera on a mast lets it get views up to table top level. No way to handle steps yet, and it mostly avoids clutter, but this is definitely progress.

The Problem As I See It

I have often stated that I think the field of AI, despite the great practical successes recently of Deep Learning, is probably a few hundred years away from where most people think it is. We’re still back in phlogiston land, not having yet figured out the elements, including oxygen.

AI, Robotics, and Machine Learning are areas that I have a real personal investment in.  I wrote a terrible Masters thesis on ML back in 1977.  I joined the Stanford AI Lab later that year, then the MIT AI Lab four years later, and became director of that lab in 1997, merging it with LCS (Lab for Computer Science) to form MIT CSAIL in 2003, the largest lab at MIT, still today.  I have founded six AI and robotics companies.  After 45 years in the academic and industry trenches can I be unbiased?  Probably not.

I know that many who disagree with me will dismiss me for all that experience that I have. Perhaps those who agree with me should also dismiss me for the same reason!!

My current belief is that it all gets back to the symbol grounding problem, and even more deeply to adopting a computational approach to AI, Robotics, and ML (and I expect almost no one will agree with that latter claim).

Traditional AI put off the very hard problem of perception, and assumed that later that would be solved and a perception system would deliver symbols describing the world. What are symbols? Well, a convenient abstraction was Lisp symbols, and Lisp symbols really come down to whether two things have the same address in computer memory (with some technical fixes for copying garbage collectors moving stuff around everywhere…), with a little bit of dictionary like descriptions of each symbol in terms of other symbols. (Real dictionaries used by real people always bottom out in the physical experience of those people; not so a Lisp program.) Perhaps that certainty of symbols and separation from perception, and even the metaphor of computation, has lead people astray in the way they try to solve the problems of common sense, qualitative reasoning, inference, etc.  Many dear friends may take this as an attack on their work, but that is not how I mean it.  I mean it as more general questioning.

Along came Deep Learning and many thought that the perception problem had been solved, and that DL gave us those symbols, recognizing objects in the world. But I don’t think that is right either. I think what DL has done is solve the labeling problem, not the object recognition problem.  It gives good label, and it gives poor object. (And yes I meant label and object rather than labels and objects.)

If you want to expand your mind, try reading The Promise Of Artificial Intelligence by Brian Cantwell Smith. This is by far his shortest and most readable book from his forty years of producing them. It is exceptionally erudite. But it is still a hard slog, and even native English speakers will be consulting a dictionary quite often.

In this book Smith introduces the idea of registration, as a maintained relationship between an object outside of us and what goes on inside our head (and he would have it also in a classical computer) despite changes in perception and even context.  It feels a bit like autopoiesis for thinking rather than Maturana and Varela’s conception of it for being. It is a hard concept to grasp and hold onto.  It will be even harder to make it real.  But perhaps that is what AI will need.

And yes this little section is obscure, but the best I can do for a rather short blog post.

Space

There are three big stories in the space industry from 2021, and they all impact one or more predictions that I made.  The first is that space tourism notched up significantly from where it had been, both for sub-orbital and orbital trips.  The second is that Boeing had another serious setback on launching people into space; I had said by 2022, now I think that is less likely to happen.  And third, a lot of visibility into progress on SpaceX’s Starship (what was once called the BF Rocket) second stage progress in the first half of the year, and much less than people had expected for the first stage, with no launch yet.

Space Tourism Sub-Orbital

There were two players in each of sub-orbital and orbital space tourism.  Virgin Galactic and Blue Origin for sub-orbital, and Roscosmos and SpaceX for orbital trips.

Virgin Galactic had two manned launches of their sub-orbital Unity each of which got to a bit less than 90km in altitude (note that way back in 2004 Virgin Galactic had three manned flights that got over 100km). The first of this year’s flights was with two professional pilots only, but then on July 11^th two pilots were joined by four civilians, including billionaire founder Richard Branson. This flight was only announced after Jeff Bezos had stated his intention to get into space as the first sub-orbital tourist, and Branson rushed in and beat him by nine days.  All six people on board were part of the Virgin Galactic organization, so none paid for their flights explicitly.

Blue Origin was a little like an actual tourism company in that all three of its manned sub-orbital space flights in 2021 had between one and four paying customers on board, though each launch also had non-paying celebratory guests. Jeff Bezos was on board the first of these flights which launched on July 20^th, the 52^nd anniversary of Armstrong and Aldrin landing on the Moon.  A total of 14 people flew on these three flights, none of whom were professional space travelers. As best as I can tell six of those seats were paid for. Each of the Blue Origin flights topped 100km in altitude.

This is the first year that there have been paying customers on sub-orbital flights and it looks like six of them had paid seats, so not yet the few handfuls of paying customers that I predicted would not happen until at least 2022. So my prediction of no earlier than 2022 was correct. Blue Origin looks like it has a chance to get a few handfuls of paying customers in 2022, and we’ll see whether Virgin Galactic launches its first paying customer this coming year.

There is a question in my mind whether there will be enough appetite for these very short experiences of only about five minutes of weightlessness.  The price per minute is much more than the rate for an orbital flight. Perhaps that will push the market more towards orbital tourism. In any case both the companies involved have much bigger space aspirations than just sub-orbital flights and may lose interest if either the demand or profit margins are not high enough. Sub-orbital tourism may never reach a breakneck pace as airplane pleasure flights did long ago–the market will decide.

Space Tourism Orbital

In 2021 we saw eight orbital space tourists, two of whom paid for their own rides, however all eight rides were paid for, unlike the sub-orbital cases.  The only previous space tourists were eight orbital seats, spread over seven individuals, between 2001 and 2009. In this one year, 2021, we had a doubling of the number of all time orbital space tourists.

The first orbital tourist flight of 2021 was carried out in September by SpaceX, when four people launched on the third flight of a Falcon 9 booster, aboard the second flight of a Dragon capsule. They were aloft just under 72 hours, or three days. One of the people onboard paid for the flight, and the other three were his guests. This is the only fully commercial orbital flight ever to date.

All previous orbital tourist flights had been on Soyuz vehicles going to the International Space Station (ISS), operated by Roscosmos, the same organization that sends all other Russian manned flights aloft.  The other four 2021’s tourists were on those same vehicles to the same destination, though three of the people were working for their flights. In October a Russian actor along with a movie director/camera operator to film her journeyed to shoot scenes for a movie. In December a Japanese billionaire, Yusaku Maezawa, along with his publicist, also went to the ISS. Both visits lasted 12 days. Maezawa is also signed up to loop around the Moon with SpaceX in 2023, delayed from the original announced goal of the fourth quarter of 2018.

Clearly 2021 was the best year ever for space tourism.

Boeing’s Woes

NASA funded SpaceX and Boeing to develop two reusable capsules for launching NASA astronauts. Originally they were neck and neck on schedule. Each had some setbacks and things have taken longer than expected.

The SpaceX vehicle, Crew Dragon, first took people to space in 2020, and has now done so five times, including hosting the first non-Russian orbital tourist flight.

Boeing has not been so lucky. In their first unmanned orbital test flight in December 2019 there were serious software problems and the vehicle did not reach the ISS as planned.  During an August 2021 launch window for a re-fly of the unmanned test, problems with valves deep in the craft were discovered and it was removed from the launch pad. Current plans call for a launch in May 2022, but that is by no means certain.  Meanwhile NASA has reassigned crew members from each of the first two manned launches as those astronauts were getting stale waiting for Boeing to fix the problems. This is not a good sign.

Starship

SpaceX Starship (as distinct from Boeing Starliner) was an a very rapid launch, blow-up, try again, pace as 2020 turned into 2021. Starship will be the largest rocket ever flown and both its first and second stage are intended to be fully re-usable.

After a series of fiery flights which resulted in loss of vehicle, usually in spectacular explosions, SpaceX launched the second stage of Starship in early May 2021 and had it land softly back at the launch pad, remain standing, and survive without blowing up.

Attention shifted to the first stage and people were expecting it to launch soon, but it was not to be in 2021.  Three times, a second stage has been mated to a first stage, and sometimes a first stage has fired some engines in a static test. So far, however, there has been no launch attempt.  It thus remains to be seen whether the lessons from the second stage can lead to a faster test campaign for the first stage.

As usual with the CEO of SpaceX it is sometimes hard to discern fact from fantasy/trolling.  On November 30th (in the middle of the Thanksgiving break) he warned of potential bankruptcy for SpaceX unless all hands were on deck to fix problems with producing enough Raptor engines for Starship’s first stage. In the linked story “according to Musk’s email, SpaceX needs to launch Starship at least once every two weeks next year to keep the company afloat”. Next year in that context is 2022.

As a comparison, SpaceX launched its first Falcon 9 in June 2010.  They did not reach an annual rate of one launch every two weeks until 2020 with 26 launches. That was a ten year ramp up. So to go from zero launches to a run rate of one every two weeks by the end of the year seems rather ambitious even by SpaceX record breaking standards. [[The Falcon 9 launch rate in 2021 improved to once every 12 days.]]

An Astounding Historical Rocket Ramp Up

Incidentally, though by modern standards SpaceX develops and deploys at incredible speed, there is one historical instance of even faster rocket development, production, and deployment.

Werner Von Braun designed the V-2 rocket (originally designated A-4), and carried out the first test flight on October 3^rd, 1942. That sub-orbital flight reached an altitude of 192km, roughly twice what Virgin Galactic and Blue Origin’s suborbital rockets have achieved.

The V-2 is the progenitor of all human spaceflight rockets, and the team that designed and produced them ended up as principals in both the Soviet and US space programs. Werner von Braun was the architect of the Apollo lunar landing program. However, back in 1942 he promoted the V-2 as a weapon, and by December of 1942 Hitler ordered them into mass production.

Over the next two years there were more than 300 further test flights, and then in the last few months of the war, starting on September 8^th 1944 in an attack on newly liberated Paris, Germany launched over 3,000 of them on operational warhead delivery missions.  Three thousand operational flights in eight months.

On January 1^st, 2018, I made predictions about self driving cars, Artificial Intelligence and machine learning, and about progress in the space industry. Those predictions had dates attached to them for 32 years up through January 1^st, 2050.

I made my predictions because at the time I saw an immense amount of hype about these three topics, and the general press and public drawing conclusions about all sorts of things they feared (e.g., truck driving jobs about to disappear, all manual labor of humans about to disappear) or desired (e.g., safe roads about to come into existence, a safe haven for humans on Mars about to start developing) being imminent. My predictions, with dates attached to them, were meant to slow down those expectations, and inject some reality into what I saw as irrational exuberance.

I was accused of being a pessimist, but I viewed what I was saying as being a realist. Out of emotion I wanted to give some other predictions that went against the status quo expectations at the time. Fortunately I was able to be rational and so I only made one prediction that I view in retrospect as being hot headed (and I ‘fess up to it in the AI&ML table in this post). Any one can, and they do, make predictions. But usually they are not held to those predictions. So I am holding myself accountable to my own predictions.

As part of self certifying the seriousness of my predictions I promised to review them, as made on January 1^st, 2018, every following January 1^st for 32 years, the span of the predictions, to see how accurate they were. This is my third annual review and self appraisal, following those of 2019 and 2020. Only 29 to go!

I think in the three years since my predictions, there has been a general acceptance that certain things are not as imminent or as inevitable as the majority believed just then. So some of my predictions now look more like “of course”, rather than “really, that long in the future?” as they did in 2018.

This is a boring update. Despite lots of hoopla in the press about self driving cars, Artificial Intelligence and machine learning, and the space industry, this last year, 2020, was not actually a year of any surprises.

The biggest news of all is that SpaceX launched humans into orbit twice in 2020. It was later than they had predicted, but within my predicted timeframe. Neither they nor I are surprised or shocked at their accomplishment, but it is a big deal. It changes the future space game for the US, Canada, Japan, and Europe (including the UK…). Australia and New Zealand too. It gives countries in Asia, South America and Africa more options on how they plot their space ambitions. My heartfelt congratulations to Elon and Gwynne and the whole gang at SpaceX.

This year’s summary indicates that so far none of my predictions (except for the one hot headed one I mentioned above) have turned out to be too pessimistic. As I said last year, overall I am getting worried that I was perhaps too optimistic, and had bought into the hype too much.

Repeat of 2019’s Explanation of Annotations

As I said in 2018, I am not going to edit my original post, linked above, at all, even though I see there are a few typos still lurking in it. Instead I have copied the three tables of predictions below from 2020’s update post, and have simply added a total of eight comments to the fourth columns of the three tables. As with last year I have highlighted dates in column two where the time they refer to has arrived.

I tag each comment in the fourth column with a cyan colored date tag in the form yyyymmdd such as 20190603 for June 3^rd, 2019.

The entries that I put in the second column of each table, titled “Date” in each case, back on January 1^st of 2018, have the following forms:

NIML meaning “Not In My Lifetime, i.e., not until beyond December 31^st, 2049, the last day of the first half of the 21^st century.

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

BY some date, meaning “By” that date.

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

For now I am coloring those statements when it can be determined already whether I was correct or not.

I have started using LawnGreen (#7cfc00) for those predictions which were entirely accurate. For instance a BY 2018 can be colored green if the predicted thing did happen in 2018, as can a NET 2019 if it did not happen in 2018 or earlier. There are five predictions now colored green, the same ones as last year, with no new ones in January 2020.

I will color dates Tomato (#ff6347) if I was too pessimistic about them. No Tomato dates yet. But if something happens that I said NIML, for instance, then it would go Tomato, or if in 2020 something already had happened that I said NET 2021, then that too would have gone Tomato.

If I was too optimistic about something, e.g., if I had said BY 2018, and it hadn’t yet happened, then I would color it DeepSkyBlue (#00bfff). None of these yet either. And eventually if there are NETs that went green, but years later have still not come to pass I may start coloring them LightSkyBlue (#87cefa).

In summary then: Green splashes mean I got things exactly right. Red means provably wrong and that I was too pessimistic. And blueness will mean that I was overly optimistic.

So now, here are the updated tables.

Self Driving Cars

Oh gosh.  The definition of a self driving car keeps changing.  When I was a boy, three of four years ago, it meant that the car drove itself. Now it means that a company gives a press release that they have deployed self driving cars, reporters give breathless headlines about deployment of said self driving cars, and then bury deep within the story that actually there is a person in the loop, sometimes in the car itself, and sometimes remotely.  Oh, and the deployments are in the most benign public suburbs imaginable, and certainly nowhere that a complete and sudden emergency stop is an unacceptable action.  And no mention  of restrictions on time of day, or weather, or particular roads within the stated geographical area.

Dear technology reporters, after this table, where almost nothing has changed, save for two comments dated Jan 1^st 2021, I’ll give you some questions to ask before you anoint corporate press releases as breaking technology stories.  Just saying.

I’m still not counting the Waymo taxi service in Chandler, a suburb of Phoenix, as a driverless taxi service. It has finally been opened to members of the public, and has no human driver actually in the car, but it is not a driverless service yet. What about the October 8^th headline: Waymo finally launches an actual public, driverless taxi service with lede Fully driverless technology is real, and now you can try it in the Phoenix area. No. It is not fully driverless. When you read down in the story you eventually find that, just as for the last year with a more restricted customer base, there is still a human in the loop. The story proudly proclaims that “[m]embers of the public who live in the Chandler area can hail a fully driverless taxi today”, but a few sentences later say the economics may not work out “because Waymo says the cars still have remote overseers”. That would be a person in the loop. Although Waymo doesn’t give a straight answer of how many cars a single overseer oversees, they refuse to say directly that is more than one. I.e., they are not willing to say that there is less than one full person devoted to overseeing the control of each “driverless” taxi. This is not a driverless taxi service. Tech press headlines be damned. Oh, and by the way, Waymo’s aim was to get to 100 rides per week by the end of 2020. Not exactly a large scale deployment, and not open to the public at any sort of scale, and certainly not driverless.

But what about this December 9^th headline: Cruise is now testing fully driverless cars in San Francisco? Again it pays to read the story. Carefully. There is a Cruise employee in the passenger seat of every one of the “fully driverless” cars and they can stop the car whenever they choose. And just to be sure, every vehicle is also monitored by a remote employee. Two safety humans for every one of these so called fully driverless cars. A story from the same day in the Washington Post reports that the testing will only be in the Sunset district of the city, and only on certain streets. SF residents know that that is a very residential area with very little traffic, and four way stop signs at intersections every block. The traffic moves very slowly even when the streets are deserted, and it is always safe to come to a complete stop.

Reporters should ask all companies saying they are testing fully driverless cars, whether as a taxi service or not, the following questions, at least.

• Is there an employee in the “driverless car”? Are they able to stop the car or take over in any way?
• Is there a remote person able to monitor the car? Are they able to take over in any way, even just commanding that the car stop, or being able to set a way point which will let the car get out of a stuck situation? How many cars does such a remote person monitor today (as distinct from aspirations for how many in the future)?
• Is there a chase vehicle following the driverless car to prevent rear end collisions from other random drivers? (I understand that Cruise does this in San Francisco; and I think I have seen it myself with a Waymo vehicle in the last week, but I may have misinterpreted.) If there is a chase vehicle does someone in that vehicle have the ability to command the driverless vehicle to stop? Note that a chase vehicle is a whole second car driven by a person just to enable the driverless car to be safe…
• Are there any restrictions on the time of day that the system operates?
• Will testing change if there is really bad weather or will it go on regardless?
• What is the area that the test extends over, and are there any streets that the cars will not go on?
• If it is a taxi service are there designated pickup and drop off sites, or can those be anywhere? And that includes 3 meters further (regular taxis and rideshare services do this) at the passenger’s request.
• Once a passenger gets in can they change the desired destination? Can they abort the ride and get out whenever they want to?
• Can the passenger communicate with the car (not a remote person) by voice?

Most of these questions will have answers that the PR people, at this point, will not want to be forthcoming about. Sometimes they will be “unable to answer” even when it is a simple yes/no question (e.g., see the Cruise story linked in the questions above)–that is likely not a sign of them not actually knowing the answer…

The only company “that says it has” tested fully driverless cars, with no chase vehicles on California roads is Nuro, but that is a long way from a taxi service since they use special purpose tiny vehicles that can not accommodate even one person.

Towards the end of the year AutoX, a company based in China, has said they are operating a driverless taxi service in a number of cities. First note, however, that the service is not open to the public, and is only available for demonstrations by company employees. They say they have removed both drivers in the cars and remote drivers, but none of the articles I have seen have answered the comprehensive set of questions I have above, including whether there is a person doing remote monitoring rather than remote driving. The one video I have seen is from inside an empty vehicle as it drives on empty(!!) roads. I have spent a lot of time in China in a lot of different cities, and I have never seen an empty road.

I am not saying there is anything wrong with these small steps towards driverless cars. They are sensible, and necessary steps along the way. But if you look at the headlines, and even the first few paragraphs of most stories about driverless cars, they breathlessly proclaim that the future is already here. No it is not. Read the fine print.  We are very far from an actual driverless taxi service that can compete with a traditional taxi company (which has to make money), or even one of the ride share companies which themselves are still losing money. The deployed demonstration systems have many more people per car working full time than any of these other systems, are only operating a tiny number of rides per day, and would not be profitable even if each ride was charged at a thousand times that charged by the incumbents. We are still a long, long way from a viable profitable autonomously driven taxi service.

The last two years have seen a real shakeout in expectations for deployment of self driving cars.  The most recent example of this is was in December 2020 when Uber sold their self driving car efforts for minus $ 400 million. That’s right.  They gave away the division working on self driving cars, once thought to be the way to finally make Uber profitable, and sent along with it $ 400 million in cash.  [[… If only they had asked me. I might have bought it for minus $ 350 million! A relative bargain for Uber!]]

To illustrate how predictions have been slipping, here is a slide that I made for talks, based on a snapshot of predictions about driverless cars from March 27, 2017. The web address still seems to give the same predictions with a couple more at the end that I couldn’t fit on my slide. In parentheses are the years the predictions were made, and in blue are the dates for when the innovation was predicted to happen. The orange arrows point at predictions that I have since seen companies retract. I may have missed some retractions. The dates highlighted in red have passed, without coming to fruition. None of the other predictions have so far borne fruit.

In summary, twelve of the twenty one predictions have passed and vanished into the ether. Of the remaining nine I am confident that the three for 2021 will also not be attained. And then there are none that come due until 2024. I think it is fair to say that predictions for autonomous vehicles in 2017 were wildly overoptimistic. My blog posts of January 2017 and July 2018 tried to tamp them down by talking about deployment issues that were not being addressed, and that, in my opinion, needed large amounts of research to solve. It turns out that before you get to my concerns just the long tail of out of the ordinary situations, and not even the really long tail, just some pretty common aspects of normal driving in most cities and on freeways, have delayed the revolution.

Two more car topics

A year ago, in my update at the start of 2020, I had a little rant about a prediction that Elon Musk had made about how there would be a million Tesla robo-taxis on the road by the end of 2020. His prediction certainly helped Tesla’s stock price. Here is my rant.

Well, surprise!, everyone.  2020 has come and gone and there are not a million Tesla robo-taxis on the road. There are zero. I, and Kai-Fu Lee, both had complete confidence in our predictions of zero. And we were right. Don’t worry. Undaunted by complete and abject failure to predict accurately, twice again, Elon has new predictions. In July he predicted level 5 autonomy in 2020, and in December he made predictions of full autonomy in 2021, with his normal straight face. Those predictions have again have pumped up Tesla’s stock price. Never give up on a winning strategy!

Finally I want to say something about flying cars.  As with “self driving cars” the definition has been changing as PR departments for various companies try to claim progress in this domain. My predictions were about cars that could drive on roads, and could take to the air and fly when their driver wanted them to. Over the last couple of years “flying car” has been perverted to mean any small one or two person flying machine, usually an oversize quadcopter or octodecacopter, or some such, that takes off and lands and then stays exactly where it is until it takes off again. These are not cars. They are helicopters. With lots of rotors. They are not flying cars.

Artificial Intelligence and Machine Learning

I had not predicted any big milestones for AI and machine learning for the current period, and indeed there were none achieved. In the world of AI hype that is a bold statement to make. But I really do believe it. There have been overhyped stupidities, and I classify GPT-3 as one, see the analysis at the end of this section.

No research results from this last year are easily understood by someone not in the thick of the actual research. Compare this to the original introduction of the perceptron, or back propagation, or even deep learning. Geoff Hinton was able to convey the convergence of three of four ideas that came to be known as deep learning, in a single one hour lecture, that a non-specialist could understand and appreciate. There have been no such innovations in the last year. They are all down in the weeds.

I am not criticizing research being down in the weeds. That is how science and engineering gets done. Lots and lots of grunt work making incremental improvements. But thirty years from now we probably won’t be saying, in a positive way, “remember AI/ML advances in 2020!”  This last year will blend and become indistinguishable from many adjacent years.  Compare that to 2012, the first time Deep Learning was applied to ImageNet. The results shocked everyone and completely changed two fields, machine learning and computer vision. Shock and awe that has stood the test of time.

In last year’s review I had some thoughts and links to things such as the growing carbon foot print of deep learning, on comparison to how quickly people learn, on non generalizability of games that are learned with reinforcement learning. I also argued why game playing has been such a success for DeepMind, and why it will probably not generalize as well as people expect. These comments occurred both just before and just after last year’s version of the table below. I stand by all those comments and invite you to go look at them.

In the table below I have added only three items, two of which relate to how people look at AI and ML, and one to a missing capability, rather than technical advances.

There have been many criticisms in the popular press over the last year about how AI is not living up to its hype. Perhaps the most influential such one is from the The Economist. The headline was: An understanding of AI’s limitations is starting to sink in, with a lede of After years of hype many people feel that AI has failed to deliver. Such rationality has not stopped breathless other stories in outlets that should know better, such as the AAAS journal Science, and sometimes even in Nature. The ongoing amount of hype is depressing. And it is mostly inaccurate.

Attention

In the table above I mentioned that both common sense and attention have surfaced as something that current ML methods are lacking and at which they do not do well.

Although learning everything from scratch has become a fetish in ML (see Rich Sutton and my response, both from 2019) there are advantages to using extra built-in machinery. And of course, when ever any ML-er shows off their system learning something new they have been the external built-in machinery. They, the human, chose the task, the reward, and the data set or simulator, with what was included and what was not. Then they say “see, it did it all by itself!”. Ahem.

We have seen reinforcement learning (RL) being used to learn manipulation tasks, but they are both contrived, by the humans writing the papers, and can be extraordinarily expensive in terms of cloud computation time. How do children learn to manipulate?  They do it in evolutionarily pre-guided ways. They learn simple grasps, then they learn the pinch between forefinger and thumb, and along the way they learn to direct force to those grasps from their arm muscles. And in some seemingly magical way, they are able to use these earlier skills as building blocks.

By the time they are 18 months (and I just observed this in my own grandson), presented with a door locked with an unknown mechanism (four different instances in this particular anecdote), children use affordances to know where to grasp and how to grasp. They ignore applying arbitrary forces to the wood of the door, except to try to pull it open, testing whether they have gotten the lock in the unlocked state, and instead apply seemingly random forces and motions (just like a machine RL learner!) to just the lock mechanism, experimenting with different grasps and forces. They direct their manipulation attention to the visible lock mechanism, rather than treating the whole thing as one big point cloud as a machine RL learner might. They know where to pay attention to.

Of course this strategy might fail, and that is precisely the allure of secret lock wooden boxes. But usually it works. And the search space for the learner is a couple of orders of magnitude smaller than the “learn everything” approach. When the bill for cloud computer time is tens of millions of dollars, as it is for some “advances” promoted by large companies, an order of magnitude or two can make a real cash difference.

We need to figure out the right mechanisms for attention and common sense and build those into our learning systems if we are to build general purpose systems.

GPT-3 (or the more things change the more they are the same…)

Will Bridewell has likened GPT-3 to a ouija board, and I think that is very appropriate. People see in it what they wish, but there is really nothing there.

<rant>

GPT-3 (Generative Pre-Trained Transformer 3) is a BFNN (that’s a technical term) that has been fed about 100 billion words of text from many sources. It clusters them, and then when you give it a few words it rambles on completing what you have said. Reporters have seen brilliance in these rambles (one should look at the reports about Eliza from the 1960’s), and some have gone as far as taking various output sentences, putting them together in an order chosen by the journalist, and claiming that the resultant text is what GPT-3 said.  Disappointingly, The Guardian did precisely this getting GPT-3 to generate 8 blocks of text of about 500 words each, then the journalists edited them together to about 1,100 words and said that GPT-3 had written the essay. Which was true in only a very sort-of way. And some of these journalists (led on by the PR people at Open AI, a misnamed company that produced GPT-3) go all tingly about how it is so intelligent that it is momentous and dangerous all at once; oh, please, the New York Times, not you too!! This is BS (another technical term) plain and simple. Yes, I know, the Earth is flat, vaccines will kill you, and the world is run by pedophiles. And if I wasn’t such a bitter pessimist I would hop on the GPT-3 train of wonder, along with all those glorious beliefs promoted by a PR department.

Not surprisingly, given the wide variety of sources of its 100 billion training words, GPT-3’s language is sometimes sexist, racist, and otherwise biased. But people who believe we are on the verge of fully general AI see the words it generates, which are rehashes of what it has seen, and believe that it is intelligence. A ouija board. Some of its text fragments have great poetry to them, but they are often unrelated to reality, and if anyone is even a little skeptical of it, they can input a series of words where its output clearly shows it has zero of the common human level understanding of those words (see Gary Marcus and Ernie Davis in Technology Review for some examples). It just has words and their relationships with no model of the world. The willfully gullible (yes, my language is harsh here, but the true believers or journalists who rave about GPT-3 are acting like cult followers–it is a mirage, people!) see true intelligence, but its random rants that it produces (worse than my random rants!) will not let it be part of any serious product.

And finally, GPT-n was actually invented, using neural networks, back in 1956. It just needed more and bigger computers to fool more of the people more of the time. In that year, Claude Shannon and John McCarthy published a set of papers (by thirteen mostly very famous authors including themselves, John von Neumann, and Marvin Minsky) titled Automata Studies, under the imprint of Princeton University Press (Annals of Mathematical Studies, Number 34).

I love that the first paragraph of their preface begins and ends:

Among the most challenging scientific questions of our time are the corresponding analytic and synthetic problems. How does the brain function? Can we design a machine which will simulate a brain? … Currently it is fashionable to compare the brain with large scale electronic computing machines.

Apart from the particular sparse style of their language, an equivalent passage could easily be thought to have been written by any number of modern researchers, sixty four years later.

On page vi of their preface they criticize a GPT-n like model of intelligence, one that is based on neural networks (the words used in their volume) and able to satisfy the Turing Test (not yet quite called that) without really being a thinking machine:

A disadvantage of the Turing definition of thinking is that it is possible, in principle, to design a machine with a complete set of arbitrarily chosen responses to all possible input stimuli (see, in this volume, the Culbertson and the Kleene papers). Such a machine, in a sense, for any given input situation (including past history) merely looks up in a “dictionary” the appropriate response. With a suitable dictionary such a machine would surely satisfy Turing’s definition but does not reflect our usual intuitive concept of thinking.

The machine they refer to here, a neural network, is just like the trained GPT-3 network, though Culbertson and Kleene, in 1956, had not figured out how to train it.

Steven Kleene was a very famous logician and academic Ph.D. sibling of Alan Turing, under Alonzo Church. James Culbertson is a forgotten engineer who speculated on how to build intelligent machines using McCulloch-Pitts neurons, here and elsewhere–I rehabilitate him a little in my upcoming book.

S. C. Kleene, Representation of Events in Nerve Nets and Finite Automata, pp 3-41.

James T. Culbertson, Some Uneconomical Robots, pp 99-116.

</rant>

AlphaFold

Finally, I want to mention AlphaFold, which right at the end of the year got incredible press, breathless incredible press I would say, about “solving” an outstanding question in science.

Given a sequence of amino acids making up a protein, how will that protein fold due to physical forces between parts of nearby acids? This is a really important question. The three dimensional structure, what is inside, and what is on the outside, and what adjacencies are there on the surface of the folded molecule, determines how that protein will interact with other molecules. Thus it gives a clue as to how drugs will interact with a protein.  People in Artificial Intelligence have worked on this since at least the early 1990’s (e.g., see the work of Tomás Lozano-Pérez and Tom Dietterich).

AlphaFold (from DeepMind, the subsidiary of Google that is losing the better part of a billion dollars per year) is a real push forward. But it does not solve the problem of predicting protein folds.  It is very good at some cases, and very poor at other cases, and you don’t know which it is unless you know the real answer already. You can see a careful (and evolving) analysis here. It is long so you can skip to the four paragraphs of conclusion to hear about both the goods and the bads. It likely will not revolutionize drug discovery without a lot more work on it, and perhaps never.

While AlphaFold is another interesting “success” for machine learning, it does not advance the fields of either AI or ML at all. And its long term impact is not yet clear.

Space

In 2020 SpaceX came back from the explosion of a Crewed Dragon capsule that happened during a ground test on April 20^th, 2019.  They were able to launch their first two people to space aboard Demo-2 on May 30^th, 2020.  Bob Behnken and Doug Hurley, both NASA astronauts, docked with the ISS on May 31st, and stayed there until their return and splashdown on August 2^nd.  On November 16^th, Crew-1 launched with three NASA and one JAXA astronauts on board, and is currently docked to ISS where the astronauts are working.

Boeing’s entry into manned space flight, the CST-100, is behind SpaceX. A software problem in December 2019 meant that an unmanned flight did not make it to the ISS, though it did return safely to Earth. In April of 2020 Boeing announced it would refly the unmanned test in the fourth quarter of 2020. That has now slipped to the first quarter of 2021 due to further software issues. I think Boeing is one glitch away from not to getting a manned flight in 2021. So my optimism for them is going to be tested.

There was only one attempt at a manned suborbital flight in 2020 but it did not get to space. On December 12^th the rocket motor of Virgin Galactic’s SpaceShipTwo Unity did not ignite after the craft was dropped as planned from its mothership at over 40,000 feet in altitude.

On October 13^th, 2020, Blue Origin’s New Shepard 3 had its 7^th successful unmanned test flight. A new craft, New Shepard 4, was slated to undergo an unmanned test flight in the last two months of the year, as the 14^th test flight of a New Shepard system. That did not happen. New Shepard 4 is currently expected to fly the first manned mission for Blue Origin.

As noted in the table SpaceX originally predicted that they would send two paying customers around the Moon in 2018, using Falcon 9 Heavy as their launch vehicle. During 2018 they revised that to a new vehicle, now known as Starship, with a date of 2023. Progress is being made on Starship, and the 55m long second stage flew within the atmosphere for more than a simple hop in 2020. It did controlled aerodynamic flight as it came into land, and impressively and successfully flipped to vertical for a landing. There were some fuel control issues and it did not decelerate enough for a soft landing, and blew up. This is real and impressive progress.

So far we haven’t seen the 64m first stage, though there are rumors of a first flight in 2021. This will be a massive rocket, and given the move fast and blow up strategy that SpaceX has very successfully used for the second stage, we should expect its development to have some fiery moments.

A 55m long vehicle to loop around the Moon seems like a bit of overkill, but that could be the point. However, I will not be surprised at all if the mission slips beyond 2023.  If fantastic progress happens in 2021 I will get more confident about 2023, but 2021 will have to be really spectacular.