Information Density: Rodney Brooks – Signal Evidence & AI Readability

Nothing is ever as good as it first seems and nothing is ever as bad as it first seems.
— A best memory paraphrase of advice given to me by Vice Admiral Joe Dyer, former chief test pilot of the US Navy and former Commander of NAVAIR.
[You can follow me on social media: @rodneyabrooks.bsky.social and see my publications etc., at https://people.csail.mit.edu/brooks]
[H5] Table of contents
Introduction
What I Nearly Got Wrong
What Has Surprised Me, And That I Missed 8 Years Ago
My Color Scheme and Past Analysis

My New Predictions
Quantum Computers
Self Driving Cars
Humanoid Robots
Neural Computation
LLMs

Self Driving Cars
A Brief Recap of what "Self Driving" Cars Means and Meant
My Own Experiences with Waymo in 2025
Self Driving Taxi Services
__Cruise
__Tesla
__Waymo
__Zoox
Electric Cars
Flying Cars

Robotics, AI, and Machine Learning
Capabilities and Competences
World Models
Situatedness vs Embodiment
Dexterous Hands

Human Space Flight
Orbital Crewed Flights
Suborbital Crewed Flights
Boeing's Starliner
SpaceX Falcon 9
NASA, Artemis, and Returning to the Moon
SpaceX Starship
Blue Origin Gets to Orbit
New Space Stations

Addendum
[H5] Introduction
This is my eighth annual update on how my dated predictions from January 1st, 2018 concerning (1) self driving cars, (2) robotics, AI , and machine learning, and (3) human space travel, have held up. I promised then to review them at the start of the year every year until 2050 (right after my 95th birthday), thirty two years in total. The idea was to hold myself accountable for those predictions. How right or wrong was I?
The summary is that my predictions held up pretty well, though overall I was a little too optimistic. That is a little ironic, as I think that many people who read my predictions back on  January 1st, 2018 thought that I was very pessimistic compared to the then zeitgeist. I prefer to think of myself as being a realist.
And did I see LLMs coming? No and yes. Yes, I did say that something new and big that everyone accepted as the new and big thing in AI would come along no earlier than 2023, and that the key paper for its success had already been written by before I made my first predictions. And indeed LLMs were generally accepted as the next big thing in 2023 (I was lucky on that date), and the key paper, Attention Is All You Need, was indeed already written, and had first appeared in June of 2017. I wrote about this extensively in last year’s scorecard. But no, I had no idea it would be LLMs at the time of my correct prediction that something big would appear. And that lack of specificity on the details of exactly what will be invented and when is the case with all my predictions from the first day of 2018.
I did not claim to be clairvoyant about exactly what would happen, rather I was making predictions about the speed of new research ideas, the speed of hype generation, the speed of large scale deployments of new technologies, and the speed of fundamental changes propagating through the world’s economy. Those speeds are very different and driven by very different realities. I think that many people get confused by that and make the mistake of jumping between those domains of reality, thinking all the speeds will be the same.  In my case my estimates of those speeds are informed by watching AI and robotics professionally, for 42 years at the time of my predictions. I became a graduate student in Artificial Intelligence in January of 1976, just shy of 20 years after the initial public outing of the term Artificial Intelligence at the summer workshop in 1956 at Dartmouth. And now as of today I have been in that field for 50 years.
I promised to track my predictions made eight years ago today for 32 years. So I am one quarter of the way there. But the density of specific years of events or marking percentages of adoption that I predicted start to fall off right around now.
Sometime during 2026 I will bundle up all my comments over the eight years specifically mentioning years that have now passed, and put them in an archival mid-year post. Then I will get rid of the three big long tables that dominate the body of this annual post, and have short updates on the sparse dates for the next 24 years.
I will continue to summarize what has happened in self-driving cars generally, including electrification progress and the forever promised flying cars, along with AI and robotics, and human space flight. But early in 2025 I made five new predictions for the coming ten years, without specific dates, but which summarize what I think will happen.  I will track these predictions too.
What I Nearly Got Wrong
The day before my original prediction post in 2018 the price of Bitcoin had opened at $12,897.70 and topped out at $14,377.40 and 2017 had been the first year it had ever traded at over $1,000. The price seemed insane to me as Bitcoin wasn’t being used for the task for which it had been designed. The price seemed to me then, and now, to be purely about speculation. I almost predicted when it would be priced at $200, on the way down. But, fortunately, I checked myself as I realized that the then current state of the market made no sense to me and so any future state may not either. Besides, I had no experience or expertise in crypto pricing. So I left that prediction out. I had no basis to make a prediction. That was a wise decision, and I revisit that reasoning as I make new predictions now, and implore myself to only make predictions in fields where I know something.
What Has Surprised Me, And That I Missed 8 Years Ago
I made some predictions about the future of SpaceX although I didn’t always label them as being about SpaceX. A number of my predictions were in response to pronouncements by the CEO of SpaceX. My predictions were much more measured and some might say even pessimistic. Those predictions so far have turned out to be more optimistic than how reality has unfolded.
I had made no specific predictions about Falcon 9, though I did make predictions about the subsequent SpaceX launch family, now called Starship, but then known as BFR, which eight years later has not gotten into orbit.
In the meantime SpaceX has scaled the Falcon 9 launch rate at a phenomenal speed, and the magnitude of the growth is very surprising.
Eight years ago, Falcon 9 had been launched 46 times, all successful, over the previous eight years, and it had recently had a long run of successful landings of the booster whenever attempted. At that time five launches had been on a previously used booster, but there had been no attempts to launch Falcon Heavy with its three boosters strapped together.
Now we are eight years on from those first eight years of Falcon 9 launches. The scale and success rate of the launches has made each individual launch an unremarkable event, with humans being launched a handful of times per year. Now the Falcon 9 score card stands at 582 launches with only one failed booster, and there have been 11 launches of the three booster Falcon Heavy, all successful. That is a sustained growth rate of 38% year over year for eight years. And that it is a very high sustained deployment growth rate for any complex technology.
There is no other modern rocket with such a volume of launches that comes even close to the Falcon 9 record.  And I certainly did not foresee this volume of launches. About half the launches have had SpaceX itself as the customer, starting in February 2018, launching an enormous satellite constellation (about two thirds of all satellites ever orbited) to support Starlink bringing internet to everywhere on the surface of Earth.
But… there is one historical rocket, a suborbital one which has a much higher record of use than Falcon 9 over a much briefer period. The German V-2 was the first rocket to fly above the atmosphere and the first ballistic missile to be used to deliver bombs. It was fueled with ethanol and liquid oxygen, and was steered by an analog computer that also received inputs from radio guide signals–it was the first operational liquid fueled rocket. It was developed in Germany in the early 1940’s and after more than a thousand test launches was first put into operation on September 7th, 1944, landing a bomb on Paris less than two weeks after the Allied liberation of that city. In the remaining 8 months of the war 3,172 armed V-2 rockets were launched at targets in five countries — 1,358 were targeted at London alone.
My Color Scheme and Past Analysis
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.
Last year I added hemming and hawing. This is for when something looks just like what I said would take a lot longer has happened, but the underlying achievement is not what everyone expected, and is not what was delivered. This is mostly for things that were talked about as being likely to happen with no human intervention and it now appears to happen that way, but in reality there are humans in the loop that the companies never disclose. So the technology that was promised to be delivered hasn’t actually been delivered but everyone thinks it has been.
When I quote myself I do so in orange, and when I quote others I do so in blue.
I have not changed any of the text of the first three columns of the prediction tables since their publication on the first day of 2018. I only change the text in the fourth column to say what actually happened.  This meant that by four years ago that fourth column was getting very long and skinny, so I removed them and started with fresh comments two years ago. I have kept the last two year’s comments and added new ones, with yellow backgrounds, for this year, removing the yellow backgrounds from 2025 comments that were there last year. If you want to see the previous five years of comments you can go back to  the 2023 scorecard.
[H5] My NEW PREDICTIONS
On March 26th I skeeted out five technology predictions, talking about developments over the next ten years through January 1st, 2036. Three weeks later I included them in a blog post. Here they are again.
1. Quantum Computers. The successful ones will emulate physical systems directly for specialized classes of problems rather than translating conventional general computation into quantum hardware. Think of them as 21st century analog computers. Impact will be on materials and physics computations.
2. Self Driving Cars. In the US the players that will determine whether self driving cars are successful or abandoned are #1 Waymo (Google) and #2 Zoox (Amazon). No one else matters. The key metric will be human intervention rate as that will determine profitability.
3. Humanoid Robots. Deployable dexterity will remain pathetic compared to human hands beyond 2036. Without new types of mechanical systems walking humanoids will remain too unsafe to be in close proximity to real humans.
4. Neural Computation. There will be small and impactful academic forays into neuralish systems that are well beyond the linear threshold systems, developed by 1960, that are the foundation of recent successes. Clear winners will not yet emerge by 2036 but there will be multiple candidates.
5. LLMs. LLMs that can explain which data led to what outputs will be key to non annoying/dangerous/stupid deployments. They will be surrounded by lots of mechanism to keep them boxed in, and those mechanisms, not yet invented for most applications, will be where the arms races occur.
These five predictions are specifically about what will happen in these five fields during the ten years from 2026 through 2035, inclusive. They are not saying when particular things will happen, rather they are saying whether or not  certain things will happen in that decade. I will do my initial analysis of these five new predictions immediately below. For the next ten years I will expand on each of these reviews in this annual scorecard, along with reviews of my earlier predictions. The ten years for these predictions are up on January 1st, 2036. I will have just turned 81 years old then, so let’s see if I am still coherent enough to do this.
Quantum Computers
The successful ones will emulate physical systems directly for specialized classes of problems rather than translating conventional general computation into quantum hardware. Think of them as 21st century analog computers. Impact will be on materials and physics computations.
The original excitement about quantum computers was stimulated by a paper by Peter Shor in 1994 which gave a digital quantum algorithm to factor large integers much faster than a conventional digital computer. Factoring integers is often referred to as “the IFP” for the integer factorization problem.
So what? The excitement around this was based on how modern cryptography, which provides our basic security for on-line commerce, works under the hood.
Much of the internet’s security is based on it being hard to factor a large number. For instance in the RSA algorithm Alice tells everyone a large number (in different practical versions it has 1024, 2048, or 4096 bits) for which she knows its prime factors. But she tells people only the number not its factors. In fact she chose that number by multiplying together some very large prime numbers — very large prime numbers are fairly easy to generate (using the Miller-Rabin test). Anyone, usually known as Bob, can then use that number to encrypt a message intended for Alice. No one, neither Tom, Dick, nor Harry, can decrypt that message unless they can find the prime factors of Alice’s public number. But Alice knows them and can read the message intended only for her eyes.
So… if you could find prime factors of large numbers easily then the backbone of digital security would be broken. Much excitement!
Shor produced his algorithm in 1994. By the year 2001 a group at IBM had managed to find the prime factors of the number 15 using a digital quantum computer as published in Nature. All the prime factors. Both 3 and 5. Notice that 15 has only four bits, which is a lot smaller than the number of bits used in commercial RSA implementations, namely 1024, 2048, or 4096.
Surely things got better fast.  By late 2024 the biggest numbers that had been factored by an actual digital quantum computer had 35 bits which allows for numbers no bigger than 34,359,738,367. That is way smaller than the size of the smallest numbers used in RSA applications. Nevertheless it does represent 31 doublings in magnitude of numbers factored in 23 years, so progress has been quite exponential. But it could take another 500 years of that particular version of exponential growth rate to get to conquering today’s smallest version