Big fan of your work Noah. Your comments on AI art made me think about some of the angst that was floating around the webs last year - the meme of robots becoming the poets, authors, and artists, while humans were relegated back to working in the fields - I wonder if art as a business will evolve in the same way that the music industry did, i.e. with less and less revenue over time from media sales and digital streaming, growth in importance of live touring, and the emergent dominance of mega-artists and mega-tours. You could see some parallels - if digital art was largely commoditized with close to zero residual commercial value, perhaps we'll see more growth of galleries and exhibition spaces for physical and performance artworks over time?
I think the highly unionized, non-profit model of classical music (symphonic, opera, and ballet) might be a reasonable future for a lot of other artforms. It's remarkable how much the classical music world operates on the aftermath of a strike in the '40s against broadcasting recordings on the radio without compensating the musicians. It's essentially a Luddite success story, keeping the emphasis on live performance over recordings through the power of the union.
I would not be surprised at all to see more of a division widen out between "art" films and "commercial" films as we've seen between classical and pop music, where acclaimed auteurs may increasingly be funded by /donors/, as opposed to producers expecting a return. As I understand it (totally hearsay, could be mistaken), this is already part of Wes Anderson's model with Indian Paintbrush - though much of the "donation" comes in the form of lodging for the cast and crew wherever they're filming.
This is literally already the case. I agree that machine learning tech is going to be big but y'all are so clueless about the art world and the broader creative industries it is cringe inducing at times.
Most artists already don't make a living and things have already been fully commoditized decades ago due to elite overproduction and dematerialization.
Live painting is a thing. It fills a tiny niche of festival art stuff. It isn't going to grow much more than that.
Visual art is a slow a painstaking process. . I don't think it would make for a great live show. Even artists who post their process on Youtube speed up the video 8X or so.
One other gamechanger that could come out of robotics: robot nannies and childcare workers. Given the cost of childcare, I'm genuinely convinced that their development would solve our birth rate crisis within a few years.
Also, thank you for pointing out the synergy between advancements in biology and in computing. CRIPSR gets the press (reasonably so) but it only solved the problem of how to make cheap and rapid targeted gene edits. The true revolution in biology will come about due to the advancements in -omics and bioinformatics we've been seeing the past decade, allowing us to accurately measure and model cell-wide changes in phenotype at the molecular level.
I don't want to live with adults who had robots for care-givers. The master/slave mentality is toxic. We have enough trouble with adults who were raised by poor, abused, uneducated and neglected servants. What we need instead of robots is a better education system. The USA Is currently rated 13th,(!) globally.
education rankings globally globally - Google Search
Robot nannies won't do anything. There are plenty of countries with free or heavily subsidised childcare and their birth rates aren't anything special.
My understanding (from an admittedly brief perusal of online sources) is that while childcare services in those countries are very low cost for individual parents who can get them, they also suffer from significant labor shortages and long wait times. So getting your kids into them in the first place is a serious hurdle for a lot of parents, which almost certainly pushes down the number of kids people like they can afford to have. Reducing the labor shortage with automation may (hopefully) ameliorate those issues.
EDIT: Commenters below have pointed out that this is not an issue in Scandinavian countries, where birth rates are still below replacement.
^What they said - the Nordic countries at least do not have any major availability issues for kindergarten, and in Finland you have a statutory right to receive a slot to a kindergarten starting at 1 years. Of course, for a family of two professional parents, having a daycare slot available from 9 to 5 is not nearly enough to solve a lot of practical time management issues, and comparatively low post-tax incomes can make it costly to hire extra help.
Fair points from both of you and I've edited my comment to reflect that. On the point of hiring extra help, I wonder if that could be the area where the development of robots capable of childcare (or at least a portion of it) would affect birth rates. A robot in the home 24/7 that can aid in taking care of the kids, as well as general household chores as Noah pointed out, could make the difference for some parents.
I would imagine so, as a lot of parenting of toddlers involves the need to low-level engage with them (often in endless "yes-no-why" recursion loops) and to make sure that they do not injure themselves while mommy or daddy tries to answer emails
Trusting kids with robot caretakers seems like the last frontier - I think we'll see a lot more use cases before we get there. It's like self driving cars x 10 in terms of people's tolerance for errors.
But in theory, yes, this might be the way particularly considering the Baumol effect on cost of child care as everything else progresses.
I really dislike this narrative going that a declining human population is a "crises". It is necessary to avoid completely destroying the ecosystem. It is the BEST SOLUTION. Alternatives would be global disease, war, famine caused by overpopulation.
As someone who's been involved (as a hobbyist) in AI since the era of GPT-2... I do not trust those tech demos at all. Smoke and mirrors, I'd (literally) bet on it. The current techniques struggle with spatial sense so much that I simply cannot see a robot working in an arbitrary consumer kitchen in the next decade. If everything is consistently in its proper place, maybe, but in a house with children and/or cats? Forget about it.
For the same reason, I would bet against serious AI animation. It's no coincidence that all those AI videos you see are either psychedelic or consist of second-long clips. Show me a smooth animation lasting thirty seconds and I'll reconsider, but I don't see it happening this year. Barring a paradigm shift, not even this decade.
Biotech, on the other hand... talking with my sister, a pediatric nurse, about how many children will survive and even thrive because of technology and techniques that simply didn't exist when she started her job ten years ago was an eye-opener.
The most impressive professional use of AI I've seen recently was Corridor Digital's "Anime Rock Paper Scissors," which used an AI for rotoscoping (basically, turn a human performance on a greenscreen in to a fully-drawn anime character), to make an anime episode on a fraction of the budget that an actual studio would need. It's pretty cool, but there was still a lot of manual work - compositing the backgrounds, organizing all the shots to be animated, and of course writing a script that's worth animating.
I think better editing tools will be what really makes AI art take off. Currently you can generate an image that's 95% good looking but then the hands are weird or the pose is messed up or you look closer and realize that their hair melts into a necklace or something, and you can't fix it easily. Inpainting kind of works, but it's not good enough for fine details like that. We need a better way for the human to express their vision and inform the AI on what they're getting right or wrong.
I was thinking about posting some tempering of Noah's optimism, but I think it would be more interesting to here from him or other techno optimists what would be a disappointing result for them that would make them rethink their techno optimism. What would technology have to look like in 2030 for you to re-assess your optimism?
Great! I'm always for people who make predictions to set out endpoints for success or failure beforehand. It's why RCT's are better than retrospective studies.
On the techno-optimism front, I know that a decade ago they were predicting large quantities of self driving cars and much larger solar adoption by this point.
Sometimes, when I'm a churlish mood and dealing with the overly-optimistic, I'll send them https://xkcd.com/1623/ and ask them how those "surprisingly soon" self-driving cars are doing.
(I actually expected the same in 2016 -- the general failure of self-driving cars to remotely match the hype was a lesson in AI limitations I've since taken to heart.)
This is purely a legislative and "public perception sucks" thing - safety-wise, Waymo has been there for *years.*
The problem with self driving cars is, if you reduced the 40k annual road deaths to 10k, you don't get credit for saving 30k lives, you get pilloried for *killing* 10k people with your defective technology.
A self-driving car at more than 2x better than human driving capability still isn't viable due to these optics.
As of a few months ago, you can go out and buy a Mercedes-Benz with "Level 3" self-driving, which means it can drive itself under a lot of circumstances. You can also hire a "Level 4" self-driving taxi if you happen to be in the right place at the right time.
The technology continues to get better; the problem is that self-driving tech is one of those all-or-nothing technologies that only gets really interesting once it reaches a very high level of development. Though I guess you could also argue that many recent (significant) advances in vehicle safety are limited versions of self-driving too; a car that can slam on the brakes if it thinks it's about to hit something is doing a very limited version of self-driving.
Everybody believes self-driving cars are just five years away from mass adoption. Then they discover the base-rate fallacy and the prosecutor's fallacy.
This is always why "five years time" always becomes "give us another five years." Self-driving vehicles are optimized for highway driving, where vehicles are going in the same direction at a high speed and no stops or cross traffic to worry about.
Most vehicle miles/kilometers traveled are actually in urban street driving conditions, with frequent stops, entry and exit from curb cuts, as well as sharing the street with vehicles that have to stop and start frequently -- buses, delivery vans, street sweepers, mail trucks, garbage trucks and the like. There are also motorcyclists, bicyclists, pedestrians in crosswalks, jaywalkers and the dreaded children's ball falling into the road. This is where self-driving vehicles fail. There is just too much information, while humans can process and react better than machines in this environment.
In order for self-driving technology to be economical to the point where a car can be driven better than a human, engineers would have to work from a statistic of Accidents % / Total VMT, not the "90% of accidents are caused by human error" metric, because it doesn't account for a driver's driving distance between accidents, and the total population of drivers' driving distances (e.g., is there a power law distribution of accidents caused by a certain group, like drunks or slow-reacting elderly). They have a very hard target to hit when they are trying to solve a very small problem from a very large population.
Ya this is why I got very skeptical with parts of this article. A lot of the optimism is based on projections, which I know have led us to be overconfident in the past. I do agree with him on somethings like Biotech and Space. But I'm still very skeptical on robotics (at least in the short term) and solar + battery capacity solving the intermittency issues.
"Some things have been disappointing in the past, therefore everything will be disappointing in the future" isn't a great assumption.
There have been overly optimistic predictions made in the past, just as there have been overly pessimistic predictions. Arguing about any given current prediction by vague analogy to times that predictions have been wrong in the past doesn't get us any closer to the truth.
Most of the technological predictions of my adult life have been very over optimistic. What underoptimistic predictions of the 21st were there? The only one I can think of is Krugman calling the internet the next fax machine. But if the internet is the best you can come up with, that's not much since it's unclear how beneficial it's been.
2. I like to think in terms of capabilities models... and think of capabilities in an investment context as where the technical potential of assets or innovations is mobilized by a management team and business model. So potential is a rock at the top of a hill but capability is that a man with a lever by a rock at the top of a hill.
3. Capabilities compete. Or one plugs capabilities of corporations into an iterative evolutionary framework where nothing makes sense without the iterative path dependence of evolution.
4. Second and third order effects of new capabilities always dominate first order effects, or the impacts of new capabilities increase over time/iteration and new un-pre-statable combinations of those capabilities cascade through ecosystems or economies (this is stuart kaufmann’s adjacent possible).
5. So your discussion of techno-optimistic ideas sets off its own cascade in my mind. And that’s fun for me. Thank you.
On techno-optimism, I think that techno-optimism is actually pragmatic realism. In an evolutionary framework, the future belongs to those who survive and one can’t survive without adapting and adding capabilities... and humans win by collaborating at scale and scale collaboration of humans doesn’t work if the humans don’t individually choose it - choice maximized agency which makes a group a more vigorous competitor... so the group that will inevitably win to survive is probably a techno-optimist group whose beliefs maximize human agency and the conversion of technical potential into economic capabilities...
In summary - appreciate what you are doing here and like how you think.
After reading "A City on Mars" I am much more skeptical that commercially viable space exploration is going to be occurring in the next 50 years, but the rest is really promising!
I'm skeptical about space colonization economically (there's about as much motivation to colonize Antarctica), but I think there's still a chance for orbital industry in the near future. The economic case is easier to make when you only need to ship stuff to and from LEO, the only question is if there's anything worth building up there besides more spaceships.
(I've read a few sci-fi stories that have "high-tech material that can only be made outside a gravity well" as a driving macguffin, in recognition of this problem)
In the beginning you mention the drop-off in productivity starting in 2005 and its reaccelearion in 2021-2 time frame. I think ZIRP had a lot to do with that. Scarcity as the interesting ability to focus minds. In this context capital was not scarce - it was cheap - and resulted in a lot of waste. That has now changed. I highly encourage you to check out the book by "Scarcity" by Mullainathan.
I stand corrected, you are right. I thought productivity had been suffering, but it has been going up for all workers. However, manufacturing productivity has been in a long term decline since 2010.
Toyota pomoses solid state batteries every 3-4 years. Since 2010. Don't hold your breath. SSB will come but not from Toyota and unlikely to be commercialized this decade. Batteries are incredibly hard to make, not to mention improve (all 5 issues that count and not just one of the 5).
BNEFs forecasts on EVs have been notoriously wrong on the LOW side. In China EV sales is projected to reach 80% in December 2025. Pure BEV sales will be around 65%. Looking back by 2027 I think people will find it laughable that in 2023 estimates for EV adoption would be so low. BEV adoption has grown 100 fold from 2013-2022. A 10x every 5 years. By the end of the next 5 year period in 2027 BEV adoption globally will likely be around 60% to 70%.
I don't electric vehicles will make up more than 30 percent of light vehicle sales in the US in 2026. That is a number that I don't think would be crossed, and it serves as a marker for my techno-pessimism.
I don't necessarily disagree. But whether it's 25% or 35% is irrelevant. What matters are the global numbers. The US has been a laggard on the EV revolution (with some very valid reasons for it but also a lot of lies and FUD being spread by the fossil fuel industry and their mouthpiece, the mainstream media).
And what people are missing now is that EV adoption numbers don't just go up as EV sales increase, it also goes up as ICE sales drop. And (global) ICE sales dropping will only increase every year.
Great list! Didn’t see a lot of neural related technologies though. Do you not see a lot of brain augmented technologies coming out this decade? Like neuralink type technologies or more advanced telepathy?
The one think I remember from the Jetsons is George complaining about his gruelling (I think), 3 hour days. And then there were commentators who spoke of how people would use all their extra free time. But in the working world hours spent at work have seemed to go in the other direction.
>But in the working world hours spent at work have seemed to go in the other direction.
Maybe in some industries? Though honestly I can't think of any.
The big story on this in the last few years is the de facto reduction of hours for office workers, as WFH/Remote/Hybrid means your hours 'at work' are now aligned with how many hours of work you actually have to do, rather than hours spent in the office doing nothing.
My hot take is that life is actually a little too long to experience fully, and we need to fill some of it with mindless monotony to make it through all 80 years. I think Zen is a more positive framing of this idea. It’s a delicate balance. For me a 30 hour work week comes out as about optimal, and less would actually make me less happy.
100% in agreement that AI tech that doesn't depend so much on language will evolve and improve faster than LLMs, which are dying on the vine, except for search and walled gardens. Thank you as always for your optimism! And rabbits!
Big fan of your work Noah. Your comments on AI art made me think about some of the angst that was floating around the webs last year - the meme of robots becoming the poets, authors, and artists, while humans were relegated back to working in the fields - I wonder if art as a business will evolve in the same way that the music industry did, i.e. with less and less revenue over time from media sales and digital streaming, growth in importance of live touring, and the emergent dominance of mega-artists and mega-tours. You could see some parallels - if digital art was largely commoditized with close to zero residual commercial value, perhaps we'll see more growth of galleries and exhibition spaces for physical and performance artworks over time?
That's a really interesting thought! Now I kind of want to write a post about that! I'll give you credit of course!!
I would love to hear your take on what the T-Swift of abstract expressionism would be like.
Hahahaha that's a little TOO far outside my lane... :D
Memes.
I think the highly unionized, non-profit model of classical music (symphonic, opera, and ballet) might be a reasonable future for a lot of other artforms. It's remarkable how much the classical music world operates on the aftermath of a strike in the '40s against broadcasting recordings on the radio without compensating the musicians. It's essentially a Luddite success story, keeping the emphasis on live performance over recordings through the power of the union.
I would not be surprised at all to see more of a division widen out between "art" films and "commercial" films as we've seen between classical and pop music, where acclaimed auteurs may increasingly be funded by /donors/, as opposed to producers expecting a return. As I understand it (totally hearsay, could be mistaken), this is already part of Wes Anderson's model with Indian Paintbrush - though much of the "donation" comes in the form of lodging for the cast and crew wherever they're filming.
https://en.wikipedia.org/wiki/1942%E2%80%931944_musicians%27_strike
This is literally already the case. I agree that machine learning tech is going to be big but y'all are so clueless about the art world and the broader creative industries it is cringe inducing at times.
Most artists already don't make a living and things have already been fully commoditized decades ago due to elite overproduction and dematerialization.
Live painting is a thing. It fills a tiny niche of festival art stuff. It isn't going to grow much more than that.
Visual art is a slow a painstaking process. . I don't think it would make for a great live show. Even artists who post their process on Youtube speed up the video 8X or so.
Man, remember those MJ holograms? Things are going to get weird.
One other gamechanger that could come out of robotics: robot nannies and childcare workers. Given the cost of childcare, I'm genuinely convinced that their development would solve our birth rate crisis within a few years.
Also, thank you for pointing out the synergy between advancements in biology and in computing. CRIPSR gets the press (reasonably so) but it only solved the problem of how to make cheap and rapid targeted gene edits. The true revolution in biology will come about due to the advancements in -omics and bioinformatics we've been seeing the past decade, allowing us to accurately measure and model cell-wide changes in phenotype at the molecular level.
I don't want to live with adults who had robots for care-givers. The master/slave mentality is toxic. We have enough trouble with adults who were raised by poor, abused, uneducated and neglected servants. What we need instead of robots is a better education system. The USA Is currently rated 13th,(!) globally.
education rankings globally globally - Google Search
https://www.google.com/search?q=education+rankings+globally+globally&client=ms-android-xiaomi-rvo3&sca_esv=597153205&biw=375&bih=695&sxsrf=ACQVn09NGaI8Dl483DKEAB4tp2M7BNGOEg%3A1704878201624&ei=eWCeZbbdJeqkkdUPmPO-yAc&oq=education+rankings+globally+globally&gs_lp=EhNtb2JpbGUtZ3dzLXdpei1zZXJwIiRlZHVjYXRpb24gcmFua2luZ3MgZ2xvYmFsbHkgZ2xvYmFsbHkyCBAAGIAEGKIEMggQABiABBiiBEjVbVDGQ1iSW3ABeAGQAQCYAc0BoAGaCKoBBTAuNi4xuAEDyAEA-AEBwgIKEAAYRxjWBBiwA8ICBxAAGIAEGA3CAgsQABiABBiKBRiGA-IDBBgAIEGIBgGQBgg&sclient=mobile-gws-wiz-serp
Is #13 good or bad?
I don't know. Numbers (to me) are spaces
By the way, do you listen to GasslitNation?
https://podcasts.google.com/feed/aHR0cHM6Ly9nYXNsaXRuYXRpb24ubGlic3luLmNvbS9yc3M/episode/NmRhZWZlZWQtODdhZS00ODc1LWIzMDktY2E0MTI5OTgyZTc5?ep=14
Yeah!!
Robot nannies won't do anything. There are plenty of countries with free or heavily subsidised childcare and their birth rates aren't anything special.
Well they would certainly make life easier.
My understanding (from an admittedly brief perusal of online sources) is that while childcare services in those countries are very low cost for individual parents who can get them, they also suffer from significant labor shortages and long wait times. So getting your kids into them in the first place is a serious hurdle for a lot of parents, which almost certainly pushes down the number of kids people like they can afford to have. Reducing the labor shortage with automation may (hopefully) ameliorate those issues.
EDIT: Commenters below have pointed out that this is not an issue in Scandinavian countries, where birth rates are still below replacement.
^What they said - the Nordic countries at least do not have any major availability issues for kindergarten, and in Finland you have a statutory right to receive a slot to a kindergarten starting at 1 years. Of course, for a family of two professional parents, having a daycare slot available from 9 to 5 is not nearly enough to solve a lot of practical time management issues, and comparatively low post-tax incomes can make it costly to hire extra help.
Fair points from both of you and I've edited my comment to reflect that. On the point of hiring extra help, I wonder if that could be the area where the development of robots capable of childcare (or at least a portion of it) would affect birth rates. A robot in the home 24/7 that can aid in taking care of the kids, as well as general household chores as Noah pointed out, could make the difference for some parents.
I would imagine so, as a lot of parenting of toddlers involves the need to low-level engage with them (often in endless "yes-no-why" recursion loops) and to make sure that they do not injure themselves while mommy or daddy tries to answer emails
It can often cost all of one spouces after tax wages esentially defeating the short term benifit of working.
Sounds like we used to understand that trade off and had a division of labor system for parents.
Trusting kids with robot caretakers seems like the last frontier - I think we'll see a lot more use cases before we get there. It's like self driving cars x 10 in terms of people's tolerance for errors.
But in theory, yes, this might be the way particularly considering the Baumol effect on cost of child care as everything else progresses.
I really dislike this narrative going that a declining human population is a "crises". It is necessary to avoid completely destroying the ecosystem. It is the BEST SOLUTION. Alternatives would be global disease, war, famine caused by overpopulation.
As someone who's been involved (as a hobbyist) in AI since the era of GPT-2... I do not trust those tech demos at all. Smoke and mirrors, I'd (literally) bet on it. The current techniques struggle with spatial sense so much that I simply cannot see a robot working in an arbitrary consumer kitchen in the next decade. If everything is consistently in its proper place, maybe, but in a house with children and/or cats? Forget about it.
For the same reason, I would bet against serious AI animation. It's no coincidence that all those AI videos you see are either psychedelic or consist of second-long clips. Show me a smooth animation lasting thirty seconds and I'll reconsider, but I don't see it happening this year. Barring a paradigm shift, not even this decade.
Biotech, on the other hand... talking with my sister, a pediatric nurse, about how many children will survive and even thrive because of technology and techniques that simply didn't exist when she started her job ten years ago was an eye-opener.
The most impressive professional use of AI I've seen recently was Corridor Digital's "Anime Rock Paper Scissors," which used an AI for rotoscoping (basically, turn a human performance on a greenscreen in to a fully-drawn anime character), to make an anime episode on a fraction of the budget that an actual studio would need. It's pretty cool, but there was still a lot of manual work - compositing the backgrounds, organizing all the shots to be animated, and of course writing a script that's worth animating.
I think better editing tools will be what really makes AI art take off. Currently you can generate an image that's 95% good looking but then the hands are weird or the pose is messed up or you look closer and realize that their hair melts into a necklace or something, and you can't fix it easily. Inpainting kind of works, but it's not good enough for fine details like that. We need a better way for the human to express their vision and inform the AI on what they're getting right or wrong.
The AI art proves it: the man loves his bunnies.
Here for the bun art.
I was thinking about posting some tempering of Noah's optimism, but I think it would be more interesting to here from him or other techno optimists what would be a disappointing result for them that would make them rethink their techno optimism. What would technology have to look like in 2030 for you to re-assess your optimism?
I can write a post about that!
Great! I'm always for people who make predictions to set out endpoints for success or failure beforehand. It's why RCT's are better than retrospective studies.
On the techno-optimism front, I know that a decade ago they were predicting large quantities of self driving cars and much larger solar adoption by this point.
Sometimes, when I'm a churlish mood and dealing with the overly-optimistic, I'll send them https://xkcd.com/1623/ and ask them how those "surprisingly soon" self-driving cars are doing.
(I actually expected the same in 2016 -- the general failure of self-driving cars to remotely match the hype was a lesson in AI limitations I've since taken to heart.)
This is purely a legislative and "public perception sucks" thing - safety-wise, Waymo has been there for *years.*
The problem with self driving cars is, if you reduced the 40k annual road deaths to 10k, you don't get credit for saving 30k lives, you get pilloried for *killing* 10k people with your defective technology.
A self-driving car at more than 2x better than human driving capability still isn't viable due to these optics.
As of a few months ago, you can go out and buy a Mercedes-Benz with "Level 3" self-driving, which means it can drive itself under a lot of circumstances. You can also hire a "Level 4" self-driving taxi if you happen to be in the right place at the right time.
The technology continues to get better; the problem is that self-driving tech is one of those all-or-nothing technologies that only gets really interesting once it reaches a very high level of development. Though I guess you could also argue that many recent (significant) advances in vehicle safety are limited versions of self-driving too; a car that can slam on the brakes if it thinks it's about to hit something is doing a very limited version of self-driving.
Everybody believes self-driving cars are just five years away from mass adoption. Then they discover the base-rate fallacy and the prosecutor's fallacy.
https://en.wikipedia.org/wiki/Base_rate_fallacy
This is always why "five years time" always becomes "give us another five years." Self-driving vehicles are optimized for highway driving, where vehicles are going in the same direction at a high speed and no stops or cross traffic to worry about.
Most vehicle miles/kilometers traveled are actually in urban street driving conditions, with frequent stops, entry and exit from curb cuts, as well as sharing the street with vehicles that have to stop and start frequently -- buses, delivery vans, street sweepers, mail trucks, garbage trucks and the like. There are also motorcyclists, bicyclists, pedestrians in crosswalks, jaywalkers and the dreaded children's ball falling into the road. This is where self-driving vehicles fail. There is just too much information, while humans can process and react better than machines in this environment.
In order for self-driving technology to be economical to the point where a car can be driven better than a human, engineers would have to work from a statistic of Accidents % / Total VMT, not the "90% of accidents are caused by human error" metric, because it doesn't account for a driver's driving distance between accidents, and the total population of drivers' driving distances (e.g., is there a power law distribution of accidents caused by a certain group, like drunks or slow-reacting elderly). They have a very hard target to hit when they are trying to solve a very small problem from a very large population.
Ya this is why I got very skeptical with parts of this article. A lot of the optimism is based on projections, which I know have led us to be overconfident in the past. I do agree with him on somethings like Biotech and Space. But I'm still very skeptical on robotics (at least in the short term) and solar + battery capacity solving the intermittency issues.
"Some things have been disappointing in the past, therefore everything will be disappointing in the future" isn't a great assumption.
There have been overly optimistic predictions made in the past, just as there have been overly pessimistic predictions. Arguing about any given current prediction by vague analogy to times that predictions have been wrong in the past doesn't get us any closer to the truth.
Most of the technological predictions of my adult life have been very over optimistic. What underoptimistic predictions of the 21st were there? The only one I can think of is Krugman calling the internet the next fax machine. But if the internet is the best you can come up with, that's not much since it's unclear how beneficial it's been.
1. Thank you.
2. I like to think in terms of capabilities models... and think of capabilities in an investment context as where the technical potential of assets or innovations is mobilized by a management team and business model. So potential is a rock at the top of a hill but capability is that a man with a lever by a rock at the top of a hill.
3. Capabilities compete. Or one plugs capabilities of corporations into an iterative evolutionary framework where nothing makes sense without the iterative path dependence of evolution.
4. Second and third order effects of new capabilities always dominate first order effects, or the impacts of new capabilities increase over time/iteration and new un-pre-statable combinations of those capabilities cascade through ecosystems or economies (this is stuart kaufmann’s adjacent possible).
5. So your discussion of techno-optimistic ideas sets off its own cascade in my mind. And that’s fun for me. Thank you.
On techno-optimism, I think that techno-optimism is actually pragmatic realism. In an evolutionary framework, the future belongs to those who survive and one can’t survive without adapting and adding capabilities... and humans win by collaborating at scale and scale collaboration of humans doesn’t work if the humans don’t individually choose it - choice maximized agency which makes a group a more vigorous competitor... so the group that will inevitably win to survive is probably a techno-optimist group whose beliefs maximize human agency and the conversion of technical potential into economic capabilities...
In summary - appreciate what you are doing here and like how you think.
Thanks!!
After reading "A City on Mars" I am much more skeptical that commercially viable space exploration is going to be occurring in the next 50 years, but the rest is really promising!
I'm skeptical about space colonization economically (there's about as much motivation to colonize Antarctica), but I think there's still a chance for orbital industry in the near future. The economic case is easier to make when you only need to ship stuff to and from LEO, the only question is if there's anything worth building up there besides more spaceships.
(I've read a few sci-fi stories that have "high-tech material that can only be made outside a gravity well" as a driving macguffin, in recognition of this problem)
In the beginning you mention the drop-off in productivity starting in 2005 and its reaccelearion in 2021-2 time frame. I think ZIRP had a lot to do with that. Scarcity as the interesting ability to focus minds. In this context capital was not scarce - it was cheap - and resulted in a lot of waste. That has now changed. I highly encourage you to check out the book by "Scarcity" by Mullainathan.
That's interesting, because productivity accelerated in the late 2010s when we had ZIRP.
I stand corrected, you are right. I thought productivity had been suffering, but it has been going up for all workers. However, manufacturing productivity has been in a long term decline since 2010.
LFG!!!
Though I'm sure my cats will not get along with the housework robots. Maybe we need robot cats instead?
Cats are only useful as mousers, so a robot one would be a better mousetrap. My robot dog on the other hand, can fetch my slippers.
I agree about robots. I think that they will be the first killer application for LLMs, unless something unforeseen happens first.
Toyota pomoses solid state batteries every 3-4 years. Since 2010. Don't hold your breath. SSB will come but not from Toyota and unlikely to be commercialized this decade. Batteries are incredibly hard to make, not to mention improve (all 5 issues that count and not just one of the 5).
BNEFs forecasts on EVs have been notoriously wrong on the LOW side. In China EV sales is projected to reach 80% in December 2025. Pure BEV sales will be around 65%. Looking back by 2027 I think people will find it laughable that in 2023 estimates for EV adoption would be so low. BEV adoption has grown 100 fold from 2013-2022. A 10x every 5 years. By the end of the next 5 year period in 2027 BEV adoption globally will likely be around 60% to 70%.
Look at the old 2050 predictions on BEV sales from EIA if you want a good chuckle. Very flat line.
I don't electric vehicles will make up more than 30 percent of light vehicle sales in the US in 2026. That is a number that I don't think would be crossed, and it serves as a marker for my techno-pessimism.
I don't necessarily disagree. But whether it's 25% or 35% is irrelevant. What matters are the global numbers. The US has been a laggard on the EV revolution (with some very valid reasons for it but also a lot of lies and FUD being spread by the fossil fuel industry and their mouthpiece, the mainstream media).
And what people are missing now is that EV adoption numbers don't just go up as EV sales increase, it also goes up as ICE sales drop. And (global) ICE sales dropping will only increase every year.
Great list! Didn’t see a lot of neural related technologies though. Do you not see a lot of brain augmented technologies coming out this decade? Like neuralink type technologies or more advanced telepathy?
I think they're not quite ready for me to be optimistic about yet, but progress is being made!
Thanks for the perspective. I’m a fan of your columns.
Did you see the new Rabbit personal AI assistant device. $200 and interfaces to phone and PC based apps with a natural voice interface.
https://x.com/nickfloats/status/1744843239125819468?s=61&t=VH-7ngz75UVBiu_X4I1aWw
RABBIT!!!
The one think I remember from the Jetsons is George complaining about his gruelling (I think), 3 hour days. And then there were commentators who spoke of how people would use all their extra free time. But in the working world hours spent at work have seemed to go in the other direction.
>But in the working world hours spent at work have seemed to go in the other direction.
Maybe in some industries? Though honestly I can't think of any.
The big story on this in the last few years is the de facto reduction of hours for office workers, as WFH/Remote/Hybrid means your hours 'at work' are now aligned with how many hours of work you actually have to do, rather than hours spent in the office doing nothing.
My hot take is that life is actually a little too long to experience fully, and we need to fill some of it with mindless monotony to make it through all 80 years. I think Zen is a more positive framing of this idea. It’s a delicate balance. For me a 30 hour work week comes out as about optimal, and less would actually make me less happy.
100% in agreement that AI tech that doesn't depend so much on language will evolve and improve faster than LLMs, which are dying on the vine, except for search and walled gardens. Thank you as always for your optimism! And rabbits!