Now Reading
Predictions Scorecard, 2023 January 01 – Rodney Brooks

Predictions Scorecard, 2023 January 01 – Rodney Brooks

2023-01-22 07:26:58


On January 1st, 2018, I made predictions about self driving automobiles, Synthetic Intelligence, machine studying, and robotics, and about progress within the house trade. These predictions had dates hooked up to them for 32 years up by means of January 1st, 2050.

As a part of self certifying the seriousness of my predictions I promised to assessment them, as made on January 1st, 2018, each following January 1st for 32 years, the span of the predictions, to see how correct they had been. That is my fifth annual assessment and self appraisal, following these of 201920202021, and 2022. I’m over a seventh of the way in which there!  Typically I throw in a brand new aspect prediction in these assessment notes.

I made my predictions as a result of on the time, similar to now, I noticed an immense quantity of hype about these three matters, and the final press and public drawing conclusions about all kinds of issues they feared (e.g., truck driving jobs about to vanish, all handbook labor of people about to vanish) or desired (e.g., protected roads about to return into existence, a protected haven for people on Mars about to start out creating) being imminent. My predictions, with dates hooked up to them, had been meant to decelerate these expectations, and inject some actuality into what I noticed as irrational exuberance.

I used to be accused of being a pessimist, however I considered what I used to be saying as being a realist. Within the final couple of years I’ve began to suppose that I too, reacted to all of the hype, and was overly optimistic in a few of my predictions. My present perception is that issues will go, total, even slower than I believed 5 years in the past. That’s not to say that there has not been nice progress in all three fields, but it surely has not been as overwhelmingly inevitable because the tech zeitgeist thought on January 1st, 2018.

UPDATE of 2019’s Clarification of Annotations

As I mentioned in 2018, I’m not going to edit my authentic publish, linked above, in any respect, though I see there are a couple of typos nonetheless lurking in it. As an alternative I’ve copied the three tables of predictions beneath from 2022’s replace publish, and have merely added feedback to the fourth columns of the three tables. I additionally spotlight dates in column two the place the time they check with has arrived.

I tag every remark within the fourth column with a Cyan (#00ffff) coloured date tag within the type yyyymmdd similar to 20190603 for June 3rd, 2019. As in 2022 I’ve highlighted the brand new textual content put in for the present 12 months in LemonChiffon (#fffacd) in order that it’s simple to pick this 12 months’s updates. There are 15 such updates this 12 months within the tables beneath.

The entries that I put within the second column of every desk, titled “Date” in every case, again on January 1st of 2018, have the next varieties:

NIML which means “Not In My Lifetime, i.e., not till past December 31st, 2049, the final day of the primary half of the 21st century.

NET some date, which means “No Earlier Than” that date.

BY some date, which means “By” that date.

Typically I gave each a NET and a BY for a single prediction, establishing a window during which I consider it’ll occur.

For now I’m coloring these statements when it may be decided already whether or not I used to be right or not.

I’ve began utilizing LawnGreen (#7cfc00) for these predictions which had been totally correct. As an illustration a BY 2018 might be coloured inexperienced if the expected factor did occur in 2018, as can a NET 2019 if it didn’t occur in 2018 or earlier. There are 14 predictions now coloured inexperienced, together with 4 new ones this 12 months.

I’ll colour dates Tomato (#ff6347) if I used to be too pessimistic about them. There’s one Tomato, with no new ones this 12 months. If one thing occurs that I mentioned NIML, as an illustration, then it could go Tomato, or if in 2020 one thing already had occurred that I mentioned NET 2021, then that too would have gone Tomato.

If I used to be too optimistic about one thing, e.g., if I had mentioned BY 2018, and it hadn’t but occurred, then I’d colour it DeepSkyBlue (#00bfff). The primary of those appeared this 12 months. And ultimately if there are NETs that went inexperienced, however years later have nonetheless not come to move I could begin coloring them LightSkyBlue (#87cefa). I did that beneath for one prediction in self driving automobiles final 12 months.

In abstract then: Inexperienced splashes imply I obtained issues precisely proper. Pink means provably improper and that I used to be too pessimistic. And blueness will imply that I used to be overly optimistic.

Self Driving Vehicles

ot the predictions had been made and the years in blue had been the initially predicted occasions when the aptitude or deployment would occur. (The orange arrows point out that I later discovered revised dates from the identical sources.) When a blue 12 months passes with out the prediction having been fulfilled I colour it pinkish. No new pink this 12 months, however as earlier than not a single one in all these predictions has come to fruition, up to now not ever, not simply that they’ve occurred, however later. These predictions had been a part of what made me make my predictions again in 2018 to mood them with actuality.

On the finish of 2024 and 2025, the subsequent years that present up in blue right here, I totally anticipate to have the ability to colour them pink.

However, I’m not alone.  Within the final 12 months there was a big shift, and perception in self driving automobiles being imminent and/or widespread has began to have its wheels fall off.

Bloomberg complained that regardless of 100 billion {dollars} having been spent on their growth self driving care are going nowhere.

Extra importantly varied efforts are being shut down, regardless of billions of {dollars} of sunk price.  After spending $3.6B Ford and VW shut down their three way partnership, Argo, because the story says, a self-driving road to nowhere. As this story says, Ford thought they may deliver the expertise to market in 2021 (as reported within the graphic above), however they suppose sustaining 2,000 workers looking for this chance will not be one of the best ways to serve their prospects.

Many different automotive corporations have pulled again from saying they’re engaged on driverless software program, or Stage 4.  Toyota and Mercedes are two such corporations, the place they anticipate a driver to be within the loop, and will not be attempting to take away them.

In the meantime the warmth is on Tesla for naming their $15,000 software program Full Self Driving (FSD), and the state of California is preventing them in court docket.

See this very careful review of Telsa FSD driving in a Jacksonville, FL, neighborhood, with the automotive’s proprietor, airline pilot Chuck Cook dinner. He’s a real fan of Tesla and FSD, however assessments it rigorously, a lot in order that the CEO of Tesla has despatched engineers to work with him and tune up FSD for his neighborhood. As this assessment reveals it’s nonetheless not dependable in lots of circumstances. If it’s not dependable generally then the human needs to be paying consideration always and so it’s a great distance from truly being FSD. Individuals will make excuses for it, however that received’t work with regular individuals, like me. Paying $15,000 for software program which doesn’t do its major job, particularly giving the human full confidence, will not be a successful enterprise technique when you get out of the techy bubble.

Oh, and by the way in which, the  “one million robo-taxis” on the highway by 2020, promised by the CEO of Tesla in early 2019 have nonetheless not panned out. I consider the precise quantity continues to be solidly zero.

Different components of the good self driving experiment have additionally proved to be too costly. This 12 months Amazon shut down its final mile autonomous supply service undertaking, the place the concept was that driverless autos would get their packages on to homes.  Too exhausting for now.

However I Did Take Self-Driving Taxi Companies in 2022

Again in Might, on this weblog, I reported my experience taking a very driverless taxi service 3 times in a single night in San Francisco. It was with Cruise (owned my GM) who now cost cash for this service that runs on good climate nights from round 10:30pm to five:00am.  These are the hours of least site visitors in San Francisco.  It doesn’t function in all components of the town, particularly not the crowded congested components.

Right here is the transient abstract of my expertise.. The primary journey was to be about 20 blocks due south of my home. The Cruise car didn’t come to my home however informed me to go to a different road on the finish of my block. My block is sort of tough to drive on for a lot of causes and I by no means see any of Cruise, Waymo, or Zoox autos gathering knowledge on it, although they’ll all be seen only a block away.

Fairly than heading due south the Cruise car diverted 10 blocks west and again, likely to keep away from areas with site visitors. The few occasions it was in site visitors it jogged my memory of once I taught my children to drive with heavy pointless breaking when different automobiles had been close by. It additionally keep away from unprotected left turns and most popular to loop round by means of a collection of three proper turns.

On my third journey it elected to choose me up on the opposite aspect of the road than the place I had requested it, proper in entrance of building that compelled me to stroll out into lively lanes in an effort to get in. No human taxi driver would have ignored my waves and gestures to be picked up at a protected spot thirty ft additional alongside the highway.

My complete report contains many different particulars. Functionally it labored, but it surely did so by driving slowly and avoiding any indicators of congestion. The result’s that it was slower by an element of two over any human operated journey hailing service. Which may work for choose geographies, however it’s not going to compete with human operated programs for fairly some time.

That mentioned, I have to congratulate Cruise on getting it up to now alongside. That is far more spectacular and actual than another firm’s makes an attempt at deployment. However so far as I can inform it is just about 32 autos, and Cruise is shedding $5M per day.  That is many years away from profitability.

Right here is a part of a touch upon my weblog publish by a Glenn Mercer, who I don’t suppose that I do know (apologies Glenn, if I do or ought to!!) [Chorizo is the name of a car that took me on two of my three rides.]:

Your expertise appeared comparable: your rides appeared to “attempt” (sorry about imputing company right here) to be protected by avoiding difficult conditions. It will be underwhelming that if all AVs delivered to us was affirmation of the already-known indisputable fact that we will lower fatalities within the USA by obeying site visitors legal guidelines, driving cautiously, not consuming, and so forth. In fact ANY fatality averted is an efficient factor… however it’s a little underwhelming (from the attitude of a fan of all issues sci-fi) that we might keep away from them not as a result of some sentient AI detected by way of satellite tv for pc imagery a hurtling SUV 3 blocks away and calculated its trajectory in actual time to induce in Chorizo a spectacular life-saving 3-point flip… Fairly, that simply ensuring Chorizo stopped on the cease signal and “regarded each methods” could also be all that we wanted. Ah, effectively.

I feel this can be a nice remark. It suits effectively with my prediction that no self driving automotive goes to be confronted with an occasion of the trolley downside till past 2050. My recommendation on the so referred to as “trolley downside” is to simply stomp on the rattling brakes. That matches with Glenn’s advice above to only be persistently cautious, and also you’ll do a complete lot higher than most people.  This might be achievable.

[Self Driving Cars]
Date 2018 Feedback Updates
A flying automotive might be bought by any US resident in the event that they have the funds for. NET 2036 There’s a actual risk that this is not going to occur in any respect by 2050. 20230101

There’s at present frothy hype about coming flying electrical taxi providers. See the principle textual content for why I feel this hype is overblown. And notice that these eVTOL taxis are in no way what individuals used to imply (simply 5 years in the past) after they mentioned “flying automobiles”.

Flying automobiles attain 0.01% of US complete automobiles. NET 2042 That may be about 26,000 flying automobiles given immediately’s complete.
Flying automobiles attain 0.1% of US complete automobiles. NIML
First devoted lane the place solely automobiles in really driverless mode are allowed on a public freeway.

NET 2021

It is a bit like present day HOV lanes. My guess is the left most lane on 101 between SF and Silicon Valley (at present largely the area of rushing Teslas in any case). Individuals must have their palms on the wheel till the automotive is within the devoted lane. 20210101 It did not occur any sooner than 2021, so I used to be technically right. However I actually thought this was the trail to getting autonomous automobiles on our freeways safely. Nobody appears to be engaged on this…

Maybe I used to be projecting my answer to how one can get self driving automobiles to occur earlier than the one for one substitute method that the Autonomous Automobile corporations have been taking. The left lanes of 101 are being rebuilt at this second, however solely as a toll lane–no particular help for AVs. I’ve turned the colour on this one to “too optimistic” on my half.

Such a devoted lane the place the automobiles talk and drive with decreased spacing at larger velocity than individuals are allowed to drive NET 2024
First driverless “taxi” service in a serious US metropolis, with devoted choose up and drop off factors, and restrictions on climate and time of day.

NET 2021

The choose up and drop off factors is not going to be parking spots, however like bus stops they are going to be marked and restricted for that function solely. 20190101 Though a couple of such providers have been introduced each one in all them operates with human security drivers on board. And a few function on a hard and fast route and so don’t rely as a “taxi” service–they are shuttle buses. And people which are “taxi” providers solely let a really small variety of rigorously pre-approved individuals use them. We’ll have extra to argue about when any of those providers do really go driverless. Which means no human driver within the car, and even working it remotely.
Throughout 2019 Waymo began working a ‘taxi service’ in Chandler, Arizona, with no human driver within the autos. Whereas this can be a large step ahead see feedback beneath for why this isn’t but a driverless taxi service.
20210101 It wasn’t true final 12 months, regardless of the headlines, and it’s nonetheless not true. No, not, no.

It nonetheless did not occur in any significant approach, even in Chandler. So I can name this prediction as right, although I now suppose it’ll end up to have been wildly optimistic on my half.


There was motion right here, in that Cruise began a service in San Francisco, for a couple of hours per night time, with about 32 autos. I rode it earlier than it was charging cash, but it surely now’s doing so. Cruise continues to be shedding $5M per day, nonetheless. See the principle textual content for particulars.

Such “taxi” providers the place the automobiles are additionally used with drivers at different occasions and with prolonged geography, in 10 main US cities NET 2025 A key predictor right here is when the sensors get low cost sufficient that utilizing the automotive with a driver and never utilizing these sensors nonetheless makes financial sense.
Such “taxi” service as above in 50 of the 100 largest US cities. NET 2028 Will probably be a really gradual begin and roll out. The designated choose up and drop off factors could also be utilized by a number of distributors, with communication between them in an effort to schedule automobiles out and in.
Devoted driverless package deal supply autos in very restricted geographies of a serious US metropolis.

NET 2023

The geographies must be the place the roads are huge sufficient for different drivers to get round stopped autos. 20220101

There are not any autos delivering packages anyplace. There are some meals robots on campuses, however nothing near delivering packages on metropolis streets. I am not seeing any indicators that this can occur in 2022.


It did not occur in 2022, so I can name it. It appears unlikely to occur any time quickly as some main gamers working on this space, e.g., Amazon, deserted their makes an attempt to develop this functionality (see the principle textual content). It’s a lot more durable than individuals thought, and enormous scale good religion makes an attempt to do it have proven that.

A (worthwhile) parking storage the place sure manufacturers of automobiles might be left and picked up on the entrance and they’re going to go park themselves in a human free setting.

NET 2023

The financial incentive is way larger parking density, and it’ll require communication between the automobiles and the storage infrastructure. 20220101

There has not been any seen progress in the direction of this that I can see, so I feel my prediction is fairly protected. Once more I used to be maybe projecting my very own ideas on how one can get to something worthwhile within the AV house in an affordable period of time.


It did not occur in 2022, so I can name this prediction as right. I feel Tesla’s driving system might be ok for his or her automobiles to do that, however nobody appears to be going after this small change utility.

A driverless “taxi” service in a serious US metropolis with arbitrary choose and drop off areas, even in a restricted geographical space. NET 2032 That is what Uber, Lyft, and standard taxi providers can do immediately.
Driverless taxi providers working on all streets in Cambridgeport, MA, and Greenwich Village, NY. NET 2035 Until parking and human drivers are banned from these areas earlier than then.
A significant metropolis bans parking and automobiles with drivers from a non-trivial portion of a metropolis in order that driverless automobiles have free reign in that space. NET 2027
BY 2031
This would be the place to begin for a turning of the tide in the direction of driverless automobiles.
The vast majority of US cities have the vast majority of their downtown underneath such guidelines. NET 2045
Electrical automobiles hit 30% of US automotive gross sales. NET 2027 20230101

I made this prediction 5 years in the past immediately. Whether it is correct we’ll know 5 years from immediately. Once I made the prediction many individuals on social media mentioned I used to be approach too pessimistic. Trying on the present numbers (see the principle textual content) I feel this stage of gross sales may very well be plausibly reached in one in all 2026, 2027, or 2028. Any small perturbance will knock 2026 out of rivalry. Within the worst (for me) case maybe I used to be pessimistic in predicting ten years somewhat than 9 for this to occur.

Electrical automotive gross sales within the US make up primarily 100% of the gross sales. NET 2038
Individually owned automobiles can go underground onto a pallet and be whisked underground to a different location in a metropolis at greater than 100mph. NIML There could be some small demonstration tasks, however they are going to be simply that, not actual, viable mass market providers.
First time {that a} automotive geared up with some model of an answer for the trolley downside is concerned in an accident the place it’s virtually invoked. NIML Recall {that a} variation of this was a key plot facet within the film “I, Robotic”, the place a robotic had rescued the Will Smith character after a automotive accident on the expense of letting a younger woman die.

Different Automobile Predictions — Electrical Vehicles

In my 2018 predictions I mentioned that electrical automobiles wouldn’t make up 30% of US vehicle gross sales till 2027 (and notice that the 2027 gross sales stage can solely be recognized  someday in 2028, ten years after my prediction).

This prediction was at odds with the froth about Tesla, and certainly its inventory value over the previous couple of years, till This autumn 2022, when issues modified dramatically.  I puzzled whether or not I had been too optimistic.

Nevertheless, there was a giant uptick in electrical automotive gross sales within the US in Q3 2022, as might be seen by evaluating the Blue E book numbers on gross sales of all vehicles with these of electric vehicles. In Q3 electrical car gross sales had been 205,682 (a 12 months on 12 months improve of 68%) in comparison with a complete of three,418,718 for all car gross sales (a mere 0.1% 12 months on 12 months improve), making electrical gross sales 6% of all US gross sales. Clearly one thing is occurring.

If electrical gross sales proceed to develop at 68% per 12 months then 2025 would hit 28% of recent automobiles within the US being electrical. Even at 50% 12 months over 12 months development we’d get to 30% in 2026, a 12 months sooner than I had predicted.  However these are each large ifs, and one quarter of huge development doesn’t a make a development. We have to see this sustained for some time.

In the meantime there are forces able to decelerate electrical automotive adoption, together with dropping gasoline costs. Or maybe if among the vehicle battery crops are delayed for any of 100 causes. If all the pieces goes proper I feel there’s a likelihood of assembly 30% electrical autos within the US someday within the 2026 to 2028 interval.

Not everybody agrees nonetheless, together with the chief of the largest automotive firm on the earth. Simply this final week Akio Toyoda, CEO of Toyota, has been quoted attempting to mood expectations: “Identical to the totally autonomous automobiles that we had been all presupposed to be driving by now, I feel BEVs are simply going to take longer to develop into mainstream than the media would really like us to consider.” (Observe that he additionally casually mentions that self driving automobiles are a lot more durable than everybody thought. Additionally he calls them Battery Electrical Automobiles as Toyota additionally produces a gas cell primarily based electrical car that consumes hydrogen.)

So…electrical automotive adoption is rising, however the jury continues to be out on whether or not we’ll get to 30% US market penetration by 2027.

Different Automobile Predictions — Flying Vehicles

Manner again once I made my predictions in 2018 “flying automobiles” meant automobiles that might drive on common roads after which take to the air and fly someplace. Over the previous couple of years the which means of “flying automobiles” has drifted to issues that aren’t automobiles in any respect, however somewhat very mild electrical vertical take off and touchdown (eVTOL) flying machines that can be flying taxis whisking individuals over the crowded freeways to their locations. A utopian transport construction.

Evidently there was froth galore on simply how large this trade goes to be and simply how quickly it will occur, and simply how large a revolution will probably be.


Lately I did an analysis, revealed on this weblog, of a McKinsey report saying all these items.  I allowed that maybe they had been 100 occasions too optimistic (what’s an element of 100 too large as a market evaluation between buddies, in any case), and if we allowed for 400% development per 12 months to satisfy that 100th sized market then immediately there have to be seven flights per day of eVTOLs, following a industrial profile and with individuals onboard. So… there must be movies of such flights, someplace, proper?

No, not anyplace.  There are some uncrewed flights, and there are some only a few tens of meters excessive over water,  however none over populated areas and none at  the tons of of meters that may have to be maintained. And the movies on the web sites of corporations which have taken tons of of tens of millions of {dollars} in VC funding present solely two meter excessive crewed hops.  And maybe extra tellingly, later in 2022, Larry Web page pulled the plug on Kitty Hawk, an organization he had been funding for 12 years to develop such capabilities.

A technique or one other, at scale eVTOL taxis will not be occurring quickly.

Robotics, AI, and Machine Studying

British science fiction and science author Arthur C. Clarke formulated three adages which have come to be often called Clarke’s three laws. The third of those is:

Any sufficiently superior expertise is indistinguishable from magic.

As I mentioned in my publish seven deadly sins of predicting the future of AI.

It is a downside all of us have with imagined future expertise. Whether it is far sufficient away from the expertise now we have and perceive immediately, then we have no idea its limitations. It turns into indistinguishable from magic.

When a expertise passes that magic line something one says about it’s not falsifiable, as a result of it’s magic.

When expertise is sufficiently totally different from what now we have skilled up to now, we people can’t make good guesses about its limitations. And so we begin imbuing it with magical powers, believing that it’s extra highly effective than something now we have beforehand imagined, so it turns into without delay each extremely highly effective and extremely harmful.

A extra detailed downside is what I name the efficiency/competence confusion, additionally from that seven lethal sins weblog publish.

We people have a great mapping for individuals in understanding what a specific efficiency by an individual in some area implies in regards to the basic competence of that individual in that area. As an illustration, if an individual is ready to keep on a dialog a couple of radiological photos that they’re wanting it’s a protected guess that that individual would be capable to look out of the window and inform you what the climate is on the market. And much more, inform you the types of climate that one sometimes sees over the interval of a 12 months in that very same location (e.g., “no snow right here, however plenty of rain in February and March”).

The identical can’t be mentioned for our AI programs. Efficiency on some specific activity in no indicator of any human like basic competence in associated areas.

Lately we noticed this with Deep Studying (and Reinforcement Studying for that matter, although the small print of the Alpha sport gamers had been far more advanced than most individuals understood). DL was presupposed to make each radiologists and truck drivers redundant in only a few years. As an alternative now we have a scarcity of individuals in every of those occupations. Then again DL has given each these professions helpful instruments.

This previous 12 months has seen Dall-E 2, the most recent picture from pure language system, and ChatGPT, the most recent massive language mannequin. When individuals see cherry picked examples from these they suppose that basic machine intelligence, with each promise and hazard, is simply across the nook.

There’s a veritable cottage trade on social media with two sides; one gushes over virtuoso performances of those programs, maybe cherry picked, and the opposite reveals how incompetent they’re at quite simple issues, once more cherry picked.

The issue is that as a person you don’t know upfront what you’ll get. So people nonetheless want their palms on the data house steering wheel always so as to not find yourself with a complete wreck of an final result in any purposes constructed on these programs.

Vice Admiral Joe Dyer, former chief check pilot of the US Navy, as soon as jogged my memory that nothing is ever pretty much as good because it first appears, nor as unhealthy. That’s an extremely useful adage.

Relax individuals. We neither have tremendous highly effective AI across the nook, nor the top of the world brought on by AI about to return down upon us.

And that just about sums up the place AI and Machine Studying have gone this 12 months. Plenty of froth and never a lot precise deployed new exhausting core dependable expertise.

[AI and ML]
Date 2018 Feedback Updates
Educational rumblings in regards to the limits of Deep Studying

BY 2017

Oh, that is already occurring… the tempo will choose up. 20190101 There have been loads of papers revealed on limits of Deep Studying. I’ve offered hyperlinks to some proper beneath this desk. 20200101
Return to last year’s update to see them.
The technical press begins reporting about limits of Deep Studying, and limits of reinforcement studying of sport play.

BY 2018

20190101 Likewise some technical press tales are linked beneath. 20200101
Return to last year’s update to see them.
The favored press begins having tales that the period of Deep Studying is over.

BY 2020

20200101 We’re seeing increasingly more opinion items by non-reporters saying this, however nonetheless not fairly on the tipping level the place reporters come at and say it. Axios and WIRED are getting shut.
20210101 Whereas hype stays the main matter of AI tales within the widespread press, some retailers, similar to The Economist (see after the desk) have come to phrases with DL having been oversold. So we’re there.
VCs work out that for an funding to repay there must be one thing greater than “X + Deep Studying”.

NET 2021

I’m being just a little cynical right here, and naturally there can be no method to know when issues change precisely. 20210101 That is the primary place the place I’m admitting that I used to be too pessimistic. I wrote this prediction once I was annoyed with VCs and let that frustration get the higher of me. That was silly of me. Many VCs found out the hype and are specializing in fundamentals. That’s good for the sector, and the world!
Emergence of the widely agreed upon “subsequent large factor” in AI past deep studying.

NET 2023

BY 2027

No matter this seems to be, will probably be one thing that somebody is already engaged on, and there are already revealed papers about it. There can be many claims on this title sooner than 2023, however none of them will pan out. 20210101 To date I do not see any actual candidates for this, however that’s OK. It could take some time. What we’re seeing is new understanding of capabilities lacking from the present hottest components of AI. They embrace “widespread sense” and “consideration”. Progress on these will most likely come from new methods, and maybe a kind of methods will grow to be the brand new “large factor” in AI.

There are two or three candidates effervescent up, however all popping out of the now custom of deep studying. Nonetheless no fully new “subsequent large factor”.


Plenty of individuals appear to be converging on “neuro-symbolic”, because it addresses issues lacking in massive language fashions.

The press, and researchers, usually mature past the so-called “Turing Take a look at” and Asimov’s three legal guidelines as legitimate measures of progress in AI and ML.

NET 2022

I want, I actually want. 20220101

I feel we’re proper on the cusp of this occurring. The intense tech press has run tales in 2021 about the necessity to replace, however each the Turing Take a look at and Asimov’s Legal guidelines nonetheless present up within the widespread press. 2022 would be the switchover 12 months. [Am I guilty of confirmation bias in my analysis of whether it is just about to happen?]


The Turing Take a look at was lacking from all of the breathless press protection of ChatGPT and buddies in 2022. Their efficiency, although not constant, pushes well past the outdated comparisons.

Dexterous robotic palms usually accessible. NET 2030
BY 2040 (I hope!)
Regardless of some spectacular lab demonstrations now we have not truly seen any enchancment in extensively deployed robotic palms or finish effectors within the final 40 years.
A robotic that may navigate round nearly any US house, with its steps, its litter, its slender pathways between furnishings, and so forth. Lab demo: NET 2026
Costly product: NET 2030
Inexpensive product: NET 2035
What is straightforward for people continues to be very, very exhausting for robots. 20220101

There was some spectacular progress on this path this 12 months with the Amazon’s launch of Astro. A crucial step in the direction of these a lot more durable objectives. See the principle textual content.


Astro doesn’t appear to have taken off, and there are not any new demos of it for public consumption. We might have to attend for another person to choose up the mantle.

A robotic that may present bodily help to the aged over a number of duties (e.g., entering into and away from bed, washing, utilizing the bathroom, and so forth.) somewhat than simply some extent answer. NET 2028 There could also be level answer robots earlier than that. However quickly the homes of the aged can be cluttered with too many robots.
A robotic that may perform the final 10 yards of supply, getting from a car right into a home and placing the package deal contained in the entrance door. Lab demo: NET 2025
Deployed programs: NET 2028
A conversational agent that each carries long run context, and doesn’t simply fall into recognizable and repeated patterns.

Lab demo: NET 2023

Deployed programs: 2025

Deployment platforms exist already (e.g., Google Dwelling and Amazon Echo) so will probably be a quick observe from lab demo to huge unfold deployment. 20230101

Regardless of the efficiency of ChatGPT and different pure language transformer fashions, nobody appears to have linked them to spoken language programs, nor have their textual content variations demonstrated any understanding of interactional context over durations of time with people. Maybe individuals have tried however the shortcomings of the programs dominate on this mode.

An AI system with an ongoing existence (no day is the repeat of one other day because it at present is for all AI programs) on the stage of a mouse. NET 2030 I’ll want a complete new weblog publish to elucidate this…
A robotic that appears as clever, as attentive, and as trustworthy, as a canine. NET 2048 That is a lot more durable than most individuals think about it to be–many suppose we’re already there; I say we aren’t in any respect there.
A robotic that has any actual thought about its personal existence, or the existence of people in the way in which {that a} six 12 months outdated understands people. NIML

The Subsequent Huge Factor in AI

Each Dall-E 2 and ChatGPT have much more mechanism in them than simply plan again propagation deep neural networks. To make Robotics or AI programs do stuff we all the time want lively mechanism in addition to any form of studying algorithm.

Within the outdated days of AI we constructed lively mechanisms with symbols, utilizing methods often called Symbolic Artificial Intelligence. Symbols are atomic computational objects which stand instead of one thing out within the bodily world.

Over the past 12 months there was a brand new time period making the rounds, “neuro-symbolic” or “neurosymbolic”, as a approach of merging the deep studying revolution with symbolic synthetic intelligence. Though everybody may be very protecting of their distinctive angle on this merger plainly plenty of individuals from everywhere in the area of Synthetic Intelligence consider there could also be one thing there. One may even classify the 2 media darlings above as situations of neurosymbolic AI.

In Q3 of 2022 a brand new journal Neurosymbolic Artificial Intelligence confirmed up. It has an editorial board with illustration from first fee establishments throughout the globe.

Observe that the wikipedia web page on symbolic AI, referenced above, was edited in August 2022 to incorporate the time period “neuro-symbolic” for the primary time. Whether or not neuro-symbolic turns into an actual dominant method continues to be up within the air. However it has a whiff of risk about it.

[[I don’t happen to think that neuro-symbolic will past the test of time, but it will lead to more short term progress. I’ll write a longer post on why I think it still misses out on being a solid foundation for artificially intelligent machines over the next few centuries.]]

The largest story of the final 12 months is the dearth of any additional flights of SpaceX’s Starship. Late in 2021 the CEO was speaking about attending to a flight each two weeks by the top of 2022.  However none, nada, zippo. Usually that could be thought-about an unbelievable distinction in what was promised and what was delivered. However nowadays CEOs of automotive corporations, social media corporations, and sure house transportation corporations, have been recognized to repeatedly make outrageous guarantees that don’t have anything to do with actuality.

Within the case of SpaceX this can be problematic, because the US return to the Moon is premised upon SpaceX delivering. If the federal government can’t consider the CEO’s guarantees that’s going to make coverage selections and letting contracts very tough.

NASA has had some success (un-crewed Artemis) in the direction of lunar objectives, and Blue Origin has lastly delivered engines for a big booster constructed by others.

SpaceX itself has had a banner 12 months with its Falcon 9 booster with 60 profitable launches (plus one Falcon Heavy launch); most likely the final time we had a launch fee that enormous was the V2 within the closing 12 months of the Second World Conflict (although the V2 fee was about 50 occasions larger and its success fee was a lot decrease than that of SpaceX).

House tourism had an ideal 2021 however not such a giant 12 months in 2022.  It’s nonetheless not a big scale enterprise and sure received’t be for some time, if ever.

Abstract: typical stuff goes nice, new stuff is tough and solely limping alongside.

House Tourism Sub-Orbital

In July 2020 Virgin Galactic had their first suborbital flight which took non-flight crew alongside, together with Richard Branson the founder. It was anticipated that this could mark the beginning of their industrial operations however there have been zero additional flights of any type since then with many reasons for the delays.

In 2022 Blue Origin had three flights with individuals aboard for a complete of 18 individuals. At most 14 of those paid for his or her tickets whereas the others had been an worker of Blue Origin, or had been chosen by different organizations, in order that they primarily received their seats.

All six crewed Blue Origin flights up to now have shared the identical {hardware} together with New Shepard 4 because the booster. In September 2022, a month after the sixth crewed flight, an older booster, New Shepard 3, failed throughout ascent and the flight was aborted and the un-crewed capsule landed safely. There have been no additional crewed flights for the remainder of the 12 months.

See Also

In abstract crewed sub-orbital flights had been down from 2021, and there was no measurable uptick in paying passengers. These flights will not be actually taking off at scale.

House Tourism Orbital

There had been eight non-governmental paid orbital flights up by means of 2009, all on Russia’s Soyuz. Then none till the final 4 months of 2021 when there have been three; two on Soyuz to the Worldwide House Station and one in a Crew Dragon on a SpaceX Falcon 9 that was independent with no docking in house. A complete of eight non-professionals flew to house on these three missions.

This sudden burst of exercise made it appear to be maybe issues may actually be taking off, so to talk.  However in 2022 there was solely one paid orbital flight, a SpaceX Dragon to the Worldwide House Station, carrying three non-professionals and one former NASA Astronaut working for the corporate that organized the flight and bought it from SpaceX.

The house tourism enterprise continues to be on the sputtering preliminary steps section.

SpaceX Falcon 9

SpaceX had a spectacular 12 months with its Falcon 9 having 60 profitable launches, with no launch failures and all boosters that had been meant to return to Earth efficiently doing so. It additionally had its first Falcon Heavy launch since 2019. As with all 4 Falcon Heavy launches this was profitable and the 2 aspect boosters landed again on the launch web site as anticipated.

Date 2018 Feedback Updates
Subsequent launch of individuals (check pilots/engineers) on a sub-orbital flight by a non-public firm.

BY 2018

20190101 Virgin Galactic did this on December 13, 2018.
20200101 On February 22, 2019, Virgin Galactic had their second flight, this time with three people on board, to house of their present car. So far as I can inform that’s the solely sub-orbital flight of people in 2019. Blue Origin’s new Shepard flew 3 times in 2019, however with no individuals aboard as in all its flights up to now.
20210101 There have been no manned suborbital flights in 2020.
Just a few handfuls of shoppers, paying for these flights.

NET 2020

20210101 Issues must velocity up if that is going to occur even in 2021. I could have been too optimistic.

It appears like six individuals paid in 2021 so nonetheless not a couple of handfuls. Believable that it occurs in 2022.


There have been three such flights in 2022, with maybe 14 of the 18 passengers paying. Not fairly there but.

A daily sub weekly cadence of such flights.

NET 2022

BY 2026


On condition that 2021 solely noticed 4 such flights, it’s unlikely that this can be achieved in 2022.


Solely three flights complete in 2022. An extended method to go to get to sub weekly flights.

Common paying buyer orbital flights. NET 2027 Russia provided paid flights to the ISS, however there have been solely 8 such flights (7 totally different vacationers). They’re now suspended indefinitely. 20220101

We went from zero paid orbital flights since 2009 to a few within the final 4 months of 2021, so undoubtedly an uptick in exercise.


Not so fast! Just one paid flight in 2022.

Subsequent launch of individuals into orbit on a US booster.

NET 2019

BY 2021

BY 2022 (2 totally different corporations)

Present schedule says 2018. 20190101 It did not occur in 2018. Now each SpaceX and Boeing say they are going to do it in 2019.
20200101 Each Boeing and SpaceX had main failures with their programs throughout 2019, although no people had been aboard in both case. So this objective was not achieved in 2019. Each corporations are optimistic of getting it accomplished in 2020, as they had been for 2019. I am positive it’ll occur ultimately for each corporations. 20200530 SpaceX did it in 2020, so the primary firm obtained there inside my window, however two years later than they predicted. There’s a actual threat that Boeing is not going to make it in 2021, however I feel there may be nonetheless a robust likelihood that they are going to by 2022.

Boeing had one other large failure in 2021 and now 2022 is wanting unlikely.


Boeing made up some floor in 2022, however my 2022 prediction was too optimistic.

Two paying prospects go on a loop across the Moon, launch on Falcon Heavy.

NET 2020

The latest prediction has been 4th quarter 2018. That’s not going to occur. 20190101 I am calling this one now as SpaceX has revised their plans from a Falcon Heavy to their nonetheless creating BFR (or no matter it will get referred to as), and predict 2023. I.e., it has slipped 5 years within the final 12 months.

With Starship not but having launched a primary stage 2023 is beginning to look unlikely, as one would anticipate the paying buyer (Yusaku Maezawa, who simply went to the ISS on a Soyuz final month) would wish to see a profitable re-entry from a Moon return earlier than going himself. That’s a whole lot of check program to get to there from right here in underneath two years.


With zero 2022 launch exercise of Starship this implies it could not presumably occur till 2024.

Land cargo on Mars for people to make use of at a later date NET 2026 SpaceX has mentioned by 2022. I feel 2026 is optimistic but it surely could be pushed to occur as an announcement that it may be accomplished, somewhat than for an urgent sensible cause. 20230101

The CEO of SpaceX has a sample of over-optimistic timeframe predictions . This didn’t occur in 2022 as he predicted. I am now pondering that my 2026 prediction is approach over optimistic, as the one present plan is to make use of Starship to attain this.

People on Mars make use of cargo beforehand landed there. NET 2032 Sorry, it’s simply going to take longer than each one expects.
First “everlasting” human colony on Mars. NET 2036 Will probably be magical for the human race if this occurs by then. It’ll really encourage us all.
Level to level transport on Earth in an hour or so (utilizing a BF rocket). NIML This is not going to occur with out some main new breakthrough of which we at present haven’t any inkling.
Common service of Hyperloop between two cities. NIML I can not assist however be reminded of when Chuck Yeager described the Mercury program as “Spam in a can”.

Boeing’s Woes

The Boeing Starliner is a capsule for use by NASA on an identical industrial foundation that NASA makes use of Crew Dragon from SpaceX. Initially the event of the 2 programs was neck and neck, however each had been delayed past authentic expectations and Boeing’s Starliner fell behind Crew Dragon.

The primary un-crewed flight was in December 2019 and there have been many severe issues revealed. Boeing determined that it wanted to re-fly that mission earlier than placing NASA astronauts aboard and there have been repeated delays with the second flight lastly occurring in Might of 2022, the place it autonomously docked with the Worldwide House Station after which safely returned to a splashdown on Earth.

The primary crewed flight is at present scheduled for April 2023, a few years later than the unique expectation.


NASA efficiently flew its Artemis 1 mission, on NASA’s new House Launch System, and the Orion Capsule, in November and December of 2022.  It flew to the Moon and and went into a really elliptical orbit, lastly returned to a protected splash down on Earth after 25 days.

The Artemis 2 mission will take astronauts to lunar orbit in Might 2024, to be adopted by Artemis 3 touchdown two astronauts on the Moon a while in 2025. There are a complete of ten crewed Artemis missions at present deliberate, out to 2034, with missions 3 by means of 11 all touchdown astronauts on the Moon, generally for a lot of months at a time. All astronauts will depart Earth on NASA’s House Launch System. Lunar landings for missions 3 and 4 are at present scheduled to make use of SpaceX’s Starship. See beneath.  SpaceX Falcon Heavy (triple Falcon 9 boosters) will take {hardware} to Lunar orbit for Gateway, a small house station.


Starship is SpaceX’s meant workhorse to interchange each Falcon 9 and Falcon Heavy.

It’s two levels, each of that are meant to return to Earth. The second stage had many check flights up by means of early 2021, and after a lot of prototypes being destroyed in explosions one efficiently landed again at its launch web site in Texas. None of those flights left the environment so the warmth shields haven’t but been examined.

The primary stage is huge, with 37 Raptor engines which are a era later than the Merlins used on Falcon 9. To date it has not flown, nor has it efficiently lit as many as 30 of its engines in a floor check.

Over US Thanksgiving in 2021 the CEO of SpaceX urged his workers to desert their households and are available in to work to spice up the manufacturing fee of those engines. In his e mail he mentioned:

What it comes all the way down to is that we face real threat of chapter if we can’t obtain a Starship flight fee of not less than as soon as each two weeks subsequent 12 months.

“Subsequent 12 months” could be 2022.  There have been zero check flights of both the second stage or the primary stage and it’s the finish of 2022. The primary stage has by no means flown in any respect. Not even powered up all its engines. The primary stage is far more advanced than any earlier booster ever constructed. It’s not flying as soon as each two weeks. Does that imply SpaceX is at risk of chapter? Or was the SpaceX CEO utilizing hyperbole to extract work from his workers?

Testing of the Raptor engines continues on the McGregor launch web site and generally they blow up (most just lately on Dec 21st, 2022)–they might be being examined to destruction when that occurs.

I’m involved that Starship is not going to be prepared for the proposed 2025 Lunar Touchdown of astronauts and that may actually delay humankind’s return the to Moon.

Blue Origin’s Orbital Class Rockets

One other competitor for touchdown astronauts on the Moon has been Blue Origin. Whereas they’ve demonstrated suborbital flights, with vertical landings, they haven’t but demonstrated heavy elevate functionality. Their New Glenn booster has had many delays and now appears to be scheduled for a primary launch late in 2023.

Blue Origin has developed fully new engines, the BE-4 for the primary stage which can be powered by seven of them.  In October of 2022 Blue Origin delivered two of those engines to a buyer, ULA, for the brand new era Vulcan Centaur rocket. It’s anticipated to launch with them early in 2023. This rocket has strong boosters in addition to the BE-4 engine so makes use of much less BE-4’s than New Glenn.

In any case, the supply of those engines to a paying buyer looks as if a constructive growth for the eventual flight of New Glenn.

Generic Predictions (Jan 01, 2023)

Listed here are some new predictions, probably not as detailed as those from 2018, however they’re solutions to a lot of the hype that’s on the market. I’ve not given time traces as these are largely “not going to occur the way in which the zeitgeist suggests”.

  1. The metaverse ain’t going anyplace, regardless of the tens of billions of {dollars} poured in. If something just like the metaverse succeeds it’ll from a brand new small participant, a small workforce, that’s not yoked down by an present behemoth.
  2. Crypto, as in all of the currencies on the market now, are going to fade away and lose their remaining worth. Crypto might rise once more but it surely wants a brand new set of algorithms and functionality for scaling. The more than likely path is that present nationwide currencies will morph into crypto forex as contactless fee develop into widespread in increasingly more international locations. It could result in one of many present nationwide currencies changing into far more accessible world huge.
  3. No automotive firm goes to provide a humanoid robotic that may change manufacturing in any respect. Dexterity is a great distance off, and improvements in manufacturing will take very totally different purposeful and course of varieties, maybe hardly seeming in any respect like a robotic from widespread creativeness.
  4. Massive language fashions might discover a area of interest, however they don’t seem to be the inspiration for usually clever programs.  Their novelty will put on off as individuals attempt to construct actual scalable programs with them and discover it very tough to ship on the hype.
  5. There can be human drivers on our roads for many years to return.
And One Final BITE

I do know, individuals are dissatisfied that I didn’t say something on this publish about ChatGPT extra substantial than “settle down”.  However actually I don’t suppose there’s a lot value saying.  Persons are making the identical mistake that they’ve made repeatedly and once more, fully misjudging some new AI demo because the signal that all the pieces on the earth has modified.

It hasn’t.

In essence ChatGPT is choosing the more than likely phrase to observe all people who precede it, whether or not it’s the phrases in a human enter immediate, or what it has already generated itself. The very best analogy that I see for it’s that ChatGPT is a dream generator, the desires that you’ve at night time if you end up asleep.

Typically out of your desires you may pick an perception about your self or the world round you. Or generally your desires make no sense in any respect, and generally they are often scary nightmares. Basing your life, unquestioned,  on what you dreamed final night time, and each night time, will not be a successful technique.


Source Link

What's Your Reaction?
In Love
Not Sure
View Comments (0)

Leave a Reply

Your email address will not be published.

2022 Blinking Robots.
WordPress by Doejo

Scroll To Top