Quick Summary: I am going to continue to do my yearly update on three sets of technologies every January 1st. I am going to decouple them from my prediction tables from the first post on January 1st, 2018, as the dates for predicted events are getting sparser, and at the same time many things have gone in different directions than where everyone believed the would go eight and a half years ago. I will also bring into play the five new directional predictions I made a year ago and first included in my annual update at the start of this year. The long, long tables from previous posts are heading out the door with this post.
A BRIEF HISTORY OF My PREDICTIONS
The acronyms I used for predictions in my original post were as follows.
NET year means it will not happen before that year (No Earlier Than)
BY year means I predict that it will happen by that year.
NIML, Not In My Lifetime, i.e., not before 2050.
As time passes mentioned years I color then as accurate, too pessimistic, or too optimistic
18 months ago I added hemming and hawing. This is for when something looks just like what I said would take a lot longer than has happened, but the underlying achievement is not what everyone expected, and is not what was delivered. People may be happy with what was delivered but it was not was promised. So when I put further out predictions for things that haven’t happened the way everyone was saying they would happen, I do want to claim accuracy in my predictions.
Here are two examples.
1. When everyone was promising self driving taxi services they were saying that the cars would drive themselves. Waymo appears to have delivered this, but in fact it has three separate crews of humans who intervene with the cars while they are driving. There are customer agents you can call through a system in the car to report failures of all sorts. There is a separate set of people who can give the cars commands, such as “ignore the lack of a green signal, as I can see from the images that the power is out”, or “stop trying to turn right, and instead turn left”. (Bystanders have no way to communicate with the cars as they do with human driven cars, so there is no car/bystander cooperation.) And then there are gig workers using the Honk app, who sign up to go shut a door left open in a car, or take care of other details the car cannot do itself. [[I contrast that with my own company Robust AI where we have mobile robots deployed all over North America, and we have no remote operations or communications with those robots (for those that the customers do not insist on air gapping we do receive reports from the robots which helps spot operational problems). They just work. But anyone in the warehouse or factory where they are deployed can grab them (either their bodies or their force multiplying handlebar) and redirect them should they wish.]]
2. When I talked about flying cars it was the old fashioned idea that they would be things that could drive on a road and fly in the air. Over the past eight and a half years that definition has changed and they have become multi-rotor electric helicopters that take off and land vertically. And they have no wheels and can not move at all once on the ground. BUT, there are not even any of them in commercial service yet. There is one company that has a single seater that you need many days of training to use, and the CEO of that company told me in March that they had had six sales at four million dollars a piece to wealthy individuals. And watching one take off made me feel seasick.
Overview of My 2018 Predictions
Below I have edited down the generations of long, long tables to be shorter and sweeter and easier to read on a small device. Each of the three tables has the original predictions with their original dates. The colored dates with asterisks have been incrementally colored in my eight previous updates. In this post I am coloring in seven further predictions with dates 2026, 2027, and 2028, and put one or two asterisks against each of them. In the three sections below I explain why it is reasonable to evaluate those future outcomes already.
That leaves no predictions before 2030 uncolored, and a total of 19 left, five of which are labeled NIML (not in my lifetime). [[I had already colored the Hyperloop prediction of NIML when the last of the hyperloop companies went out of business–it never made technical sense and that particular fantasy is dead. But don’t worry, there are plenty more fantasies coming from the same place. Most of them will die too.]]
So my blog going forward will summarize progress and failures in my three areas:
- Self driving cars (and electric, and flying ones too).
- Robotics, AI, and machine learning.
- Human spaceflight.
In the future I will mention each of my specific predictions when its date comes up. And each year until January 1st, 2037, I will continue to discuss progress and failures in the areas of my five new ten year long predictions, made about a year ago:
- Quantum Computers
- Self Driving Cars
- Humanoid Robots
- Neural Computation
- LLMs
Note that self driving cars appear in both lists.
Self Driving Cars
I have pre-called three predictions here, one which would have been decided in 18 months, and two that were only six months away. There are nine dated predictions left now, with six ranging from 2031 to 2045, and three that are “not in my lifetime” (and recall that I am using my 95th birthday, just days before my 32nd promised update on January 1st , 2050 as when I can give up on tracking these things).
Self-driving Taxi Services
Here I have a two asterisk (2028) early calling of one of my predictions.
I had said that that it would be at least until 2028 before there were 50 major US cities with driverless taxi services. At the time I meant really driverless. But I’m going to relax that (since the world has changed out from under me) and say that even if they are Waymo style services where a remote human operator can intervene I will allow that watered down system to qualify for when I predicted truly driverless taxis. I won’t allow services where there is a human safety driver sitting in either of the front seats to qualify (that one is for you Tesla and your gold painted driverless taxis that all have human drivers sitting in them or driving right behind in a chase car).
There are only two players in the US at the moment. In 2025 Waymo had service in five major cities, and Zoox had smaller services areas in two (San Francisco appeared on both lists). Waymo said it would double to deployments in ten major cities. in 2026 and I believe they will. Though, they have recently backed off from their cars driving on freeways on May 19the 2026, due to a number of safety incidents. Oh, and Waymo is still getting into trouble with city administrations for causing horrible multi-hour traffic jams, as happened right across the north side of San Francisco on July 4th, 2026, as people flocked to see fireworks centered around the Golden Gate bridge.
Meanwhile Zoox has increased its footprint in San Francisco in 2026, but it still a much smaller geographical area than for Waymo, and still not really available to the public.
But I just don’t see how Waymo can go from a doubling in 2026, including some service level retrenchments, to a quintupling in 2027. So that is why I have a two asterisk green label on the NET 2028 (No Earlier Than) prediction for fifty US cities. It is not going to happen in 2027, and I have not heard Waymo say they will try to do that. I made it easier to pass, but it is still not going to happen.
Parking Bans
In my original predictions I had said that at some point some (major) cities might ban human driven cars from some large area to allow driverless cars free reign (or rein) there. My reasoning was that with no human driven cars around the driverless cars could communicate with each other or talk to a cloud-based orchestrator to cooperate and be sure that some bad human driver wouldn’t do stupid things.
I was overly optimistic and so my prediction of NET 2027 is safe, as no cities anywhere (except perhaps Singapore) have even given this any thought. My expectation that somewhere will do it by 2032 is in grave danger. Come on Singapore!!
Electric Car Sales
I had predicted that it would take until at least 2027 before electric cars comprised 30% of all car sales in the US. I remember some people in 2018 taking great offense that I was being so pessimistic and was obviously wrong.
In 2025 the fraction was less than 8.5%, although Q3 of 2025 hit 10.5%, a record, before falling again. That strong quarter was people taking advantage of US government subsidies before they were withdrawn as part of the so-called “One Big Beautiful Bill” act. And, in 2026 there has only been bad news for electric car sales in the US as many manufacturers have cut their production numbers. That means that (apart from Cyber Trucks) there probably aren’t many vehicles to be sold should sales pick up. They might well pick up due to high gas prices as fallout from the US-Iran war. But with supply chains in shambles due to tariff turbulence there is just no way that we could get more than a point or two lift.
So… no, 30% of car sales being electric models in the US in 2026 by a factor of close to three if not more. NET 2027 is a solid prediction. And it is itself probably wildly optimistic.
| Prediction [Self Driving Cars] | Date |
|---|---|
| A flying car can be purchased by any US resident if they have enough money. | NET 2036 |
| Flying cars reach 0.01% of US total cars. | NET 2042 |
| 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 2021 |
| Such a dedicated lane where the cars communicate and drive with reduced spacing at higher speed than people are allowed to drive | NET 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 2021 |
| Such "taxi" services where the cars are also used with drivers at other times and with extended geography, in 10 major US cities | NET 2025 |
| Such "taxi" service as above in 50 of the 100 biggest US cities. | NET 2028 ** |
| Dedicated driverless package delivery vehicles in very restricted geographies of a major US city. | NET 2023 |
| 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 2023 |
| A driverless "taxi" service in a major US city with arbitrary pick and drop off locations, even in a restricted geographical area. | NET 2032 |
| Driverless taxi services operating on all streets in Cambridgeport, MA, and Greenwich Village, NY. | NET 2035 |
| 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 |
| 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. | NIML |
| 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. | NIML |
ROBOTICS, AI, AND MACHINE LEARNING
In this section I am calling two NET 2028 predictions 18 months early. Both of these are because in 8 of the 10 years at least that I predicted they would take to reach there has been no real visible progress towards them. Both are in robotics (recently re-branded by people with no understanding of robotics as “physical AI”; that should fix their ignorance problems!).
That leaves seven of my predictions open, with one of them being “not in my lifetime” and the other six ranging from 2030 to 2048.
Of those remaining six, the two that are about products in 2030 and 2035 for my prediction about
A robot that can navigate around just about any US home, with its steps, its clutter, its narrow pathways between furniture, etc.
are in real danger, as there has been no lab demonstration of this capability. There is plenty of humanoid theater out there but the “homes” are movie sets, have wide clear walkways, there are no humans around, and they have none of the messy reality of human homes. Real homes are hard. One of the companies I founded has put 55 million cleaning robots in real homes, so I can say this with conviction.
Physical Assistance to the Elderly
I predicted that we would have:
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 2028. So far we don’t have a solid demonstration, i.e., one that works again and again, for any of single one of these as point solutions. So we are not going to have a robot that can do even a few, or even one, of these that will be deployed by 2028.
Back in 2018 I was way too optimistic that the world of engineering research funding would wake up to an unfolding crisis and fund the basic research that is needed for this. But it has not happened. So I am saying today that I was right and it certainly won’t happen this year and next. It really is NET 2028. And in reality much, much further away.
And that is a real shame. The world-wide demographic inversion is going to leave most people currently aged between 40 and 65 with no care workers to look after them as they get to end of life, health wise.
Last 10 Yards of Delivery
As fulfilment services were ramping up it became clear that there would be a lot more human delivery workers needed. I predicted in 2018 that:
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.
would be deployed NET 2028. This idea was in the air when I made my predictions at the start of 2018. I thought then that it would take at least 10 years, but it will, in fact, take a lot longer than that.
To show that the idea and goal was around then, here is a movie made jointly by a major US car manufacturer and a humanoid robot company in 2019 showing what it might look like for a humanoid to get out of a van and deliver a package to the front stoop of fairly benign environment (big open side walk, just one electric scooter on the ground in the way, and only three steps up). This was one of the first instances of humanoid robot theater. It has gotten way slicker in the intervening seven years, but we are really not that much closer to even this demo being a reality. Let alone the messiness of most urban neighborhoods.
There have not been any plausible demos of this capability. The food delivery robots that you might see are far from the autonomous delivery of packages ideal. They have at least three humans involved for every delivery. One to put the food into the robot, one (the customer) to take it out, and always a remote human operator who can take over when there are any problems.
| Prediction [Robotics, AI, and ML] | Date |
|---|---|
| Academic rumblings about the limits of Deep Learning | BY 2017 |
| 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 2021 |
| Emergence of the generally agreed upon "next big thing" in AI beyond deep learning. | NET 2023 BY 2027 |
| 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 2022 |
| Dexterous robot hands generally available. | NET 2030 BY 2040 (I hope!) |
| 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 2030Affordable product: NET 2035 |
| 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 2028 ** |
| 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 |
| 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 2030 |
| A robot that seems as intelligent, as attentive, and as faithful, as a dog. | NET 2048 |
| 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 |
Human Spaceflight
In this section I have claimed validity on two predictions which I would have been able to call one way or the other by the end of this year. Both can be called now because in each case there was an unexpected event which guarantees that my predictions were correct.
Suborbital Tourism
The first is about when there would be a regular sub weekly cadence of sub-orbital tourist flights. I had previously said that it would happen NET 2022, which is accurate. But it still hasn’t happened and hanging there was my prediction that it would happen by 2026. At the end of last year it looked like there was a small chance still that it might happen.
There are only two companies operating in this space, Virgin Galactic and Blue Origin. Both companies have flown more than 50 people but less than 100. Virgin Galactic last flew their first generation vehicle in June 2024, and announced there would be a pause as they built a new class of vehicles, and said they would start test flights in late 2026. So it won’t be them that gets to regular more often than weekly flights this year.
Late last year Blue Origin announced ambitious plans to move from one to three launch sites and build a bunch more of the flight and customer proven New Shepard vehicle. That was the glimmer of a chance that I might have been spot on in predicting 2026 as the year that it might happen. But at the end of January they announced a two year pause of that program so that the company could concentrate on getting people to the Moon.
Why the change? Originally SpaceX was for sure going to be the company to land the next NASA astronauts to walk on the Moon. Last year the schedule for SpaceX to do that was not looking at all good. So an interim NASA administrator had reopened the contract award to give Blue Origin a chance to compete. But then on December 18th of 2025 a new permanent NASA administrator was sworn in. It was Jared Isaacman who had previously bought and flown on two SpaceX orbital flights on their ultra-reliable Falcon 9 rocket. Isaacman announced that he would continue the open contract negotiations and said he welcomed a competition between the two companies. And he wanted both companies to be ready with their landers for 2027 for an Earth orbital trial of docking with the Artemis III rocket, a crewed NASA mission.
So, that external change made Blue Origin change their course–they already had a NASA contract for landing cargo on the Moon, but now they had a shot at landing the first astronauts on the Moon in over 50 years.
They had a successful orbital flight of a previously flown first stage on April 19th of this year. But then on May 28th a fully fueled New Glenn exploded on its launch pad during a test. There were no casualties but the launch complex was almost completely destroyed.
NASA is sticking with their support of both SpaceX and Blue Origin. This is what NASA said about Blue Origin on July 24th.
It turns out that I was too optimistic.
External events happen all the time. And usually they delay the deployment of new technologies. If the external events are bad enough, like a way, they sometimes speed it up.
Orbital Tourism
In the past, Russia had an orbital tourism business taking paying customers on Soyuz vehicles to the International Space Station. The only other avenue for paid orbital tourism was to use the services of SpaceX with their Falcon 9 launch vehicle and their Dragon crew vehicle.
The company Axiom has become an aggregator for this, providing flights to the International Space Station, paying both SpaceX and NASA. Each flight has been commanded by a former NASA astronaut (two each by Michael López-Alegria and Peggy Whitson), with three passengers. There was one flight in each year from 2022 to 2025, none in 2026, and the next one is scheduled for 2027. The passengers have mostly been astronauts from the official space agencies of countries who have no access to launch services themselves. So it is not exactly space tourism in the sense of some rich person paying for a ride for themselves as in the Soyuz days.
There have been three orbital SpaceX flights paid for by rich individuals, and none of them went to the International Space Station. There was one in 2025 paid for my a new individual, and two previous flights in 2021 and 2024 paid for by one person–Jared Isaacson. So he is out of that business (see his new job above).
I am therefore comfortable that regular paid orbital tourist flights will not happen before 2027.
What’s Left?
There are only two dated predictions left for human spaceflight, from my 2018 predictions. They were wildly pessimistic compared to what the CEO of SpaceX was predicting at the time. But now I think I got sucked into the hype and that my predictions for humans on Mars were wildly optimistic. I framed them both as “no earlier than”, so I am guessing now that I will be able to check them off in 2032 and 2036 as having been not wrong.
| Prediction [Space] | Date |
|---|---|
| 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 2027 * |
| Next launch of people into orbit on a US booster. | NET 2019 BY 2021 BY 2022 (2 different companies) |
| Two paying customers go on a loop around the Moon, launch on Falcon Heavy. | NET 2020 |
| Land cargo on Mars for humans to use at a later date | NET 2026 |
| Humans on Mars make use of cargo previously landed there. | NET 2032 |
| First "permanent" human colony on Mars. | NET 2036 |
| Point to point transport on Earth in an hour or so (using a BF rocket). | NIML |
| Regular service of Hyperloop between two cities. | NIML |