U.S. unemployment rate is steady at ~4% and we are now almost 4 years after the release of ChatGPT. The Jobs Apocalypse seems to be always on the horizon. UBI is being pushed by college educated workers burnt out of white collar work (understandably), but when you suggest to them working construction, teaching school, or working in a hospital, they deflect and go quiet.
1) initial signs of AI adoption harming job growth got the signal backwards: It's not that we didn't hire Jrs because we had AI now, its that we didn't hire Jrs because they used AI now, and it was hard to reconcile this with our standing hireability signals: code quality, side projects, portfolio, etc.
2) The TAM argument for AI replacing all work is not even required. Anthropic and company (and several future companies) can become fantastically wealthy beyond anyone's dreams just by following claude code subscription model to make life easier for workers, who in turn gladly pay for the subcription.
both are here with more details, apologies for links to personal (ad-free) blog
Correlation isn't causation; in the last few years we've experienced some significant economic headwinds like COVID and the current oil crisis. AI rollout is not happening in a vacuum.
No, that's not the usual response, because people in this situation have evaluated those alternatives and will be easily able to articulate what is wrong with them (pay). Also, the 4% figure treats doordash as employment. In a parallel universe where the pay was decent, that could be reasonable. In this universe, not so much.
These are boomer-tier talking points and if you bring them up to a kid stuck in this position you will get dunked on and you will deserve it. Unless you found a shy kid, in which case: congratulations on your new title as undefeated shadowboxing champion.
your argument suggests there's an opportunity for a trading strategy. It's also very common as a "technical analysis" investment strategy for any gullible person on the internet.
The problem is it looks "backwards" and tries to make the point something (in this case the argument made in the paper) is false because the backwards data doesn't support it. All while completely failing to address the crux of the point being made.
by some measures unemployment is at 25% and this aligns with my personal experience much better than frankly the bs 4% rate that is being pushed by this admin
i'm sure it's meant in some kind of technical sense that is precisely defined in the field of economics, but this use of the word "rational" is irritating:
> This demand externality traps rational firms in an automation arms race, displacing workers well beyond what is collectively optimal.
you can't be trapped in a sub-optimal situation and be rational at the same time. the rational thing for these rational competitors to do would be to agree to slow their deployment of ai, because the cost savings will be more than lost to demand destruction down the line. so if they won't or can't do that, they aren't rational. on the flip side, if it is actually impossible to slow ai deployment, then the firms are rational and the situation they find themselves in isn't actually sub-optimal.
it's irritating to me because i often see persuit of self-interest even to one's eventual detriment held up as "rationality", which may be true in some technical sense within the field of economics (i really couldn't say), but is certainly not true for, like, actual rationality. myopic selfishness is never rational.
> you can't be trapped in a sub-optimal situation and be rational at the same time.
In order for this statement to be true, you must also have perfect knowledge: something that is unachievable in most circumstances. Getting trapped in a local-optimal solution is perfectly possible if you are blindly following stochastic decent along a gradient: as it is too with life... getting trapped like this looks like being in the situation where there is no rational action (only irrational ones) that (eventually) lead to anything other that that status quo. a local optimum is where you have to do increasingly (in proportion to the level of optimization) non-positive actions to escape (likely to another local-optimal).
> the rational thing for these rational competitors to do would be to agree
This is the bit that is tripping you up. The market is composed of more than just the these potentially agreeing parties, so there is no case where all parties can even come together to agree. It's rational for any party to break the agreement or be stealthy and attempt to take the upper hand while the visible players are all patting themselves on the back for their self-inflicted restraint. It's rational for one party to encourage everyone to come together to agree, to distract from their own actions to undercut everyone else.
On top of that, such agreement would potentially qualify as collusion, allowing an upstart to cry foul.
Yeah it's a pet peev of mine as well. People define very tight space and then proclaim actions irrational if they don't make rational sense in that setting.
Which wouldn't be a problem in itself if there was discussion about the defined space and if it is an actually good model.
A big chunk of economics is like that, if the data doesnt fit the model then too bad for the data. Same goes for metrics used, externalities are like a side note in economy books and not an example of fundamental issues with current economic toola and modeling.
And it's not limited to introductory level either. Nordhaus had most asinine assumptions in his model and he got the pseudo-Nobel.
It's rational at the individual level, irrational at the group level.
The issue is you can't easily enforce cooperation without some external overarching force (e.g. regulation), especially across competing powers.
Refer to nuclear arms race in general. A specific example on why not to be the only one to cooperate: Ukraine gave up nukes in order to push for group level rationality, gets walked over as an individual. Oversimplifying, but yeah that's the rough point.
> you can't be trapped in a sub-optimal situation and be rational at the same time
Yes you can that’s what tragedy of the commons is. Rational means you take the choice that provides the most value for yourself. Without a cartel (hard) or government regulation this situation happens all the time.
Yes and no. It's not a new problem for humans and it's a pretty solvable problem if social pressure and structures can be applied once it is identified. As long as the resource is to be used by a community that depends on it and not an outsider then we are able to keep ourselves mostly in line for the benefit of the group.
Tragedy of the commons is when a shared resource is rationally used by the individual, but harms the whole.
You can be trapped in sub-optimal situation because it's a local maxima; not because of tragedy of commons, which is something different.
Society can be trapped in local maxima and that would be because of tragedy of commons, and individuals can be trapped in local maxima but that's distinct from tragedy of commons.
Remember when the buttons test was circulating a few months ago? Where if you press the red button, you survive, but if you press the blue button, unless more than 50% of people chose the blue button you die? That was a case very similar to this. It is locally optimal for everyone to choose red; your own outcome is always better or equal to what it would have been if you chose blue, however everyone independently making this optimal choice is catastrophic on a global scale. Thats why its "rational".
The government will try at all costs to implement an AI tax, just because they can and because that's their job, to fleece us all. Then they will try in the name of 'peace/climate/asteroid/terrorism' to make the whole world comply or China will eat their lunch
Daron Acemoglu (2024 nobel economist) is on a book tour with some new ideas about fundamental changes to our economic theories from Ai (automation/technology are not uniform), this was a great interview
Not compelling for the basic reason laid out at the start of the article: this is just a re-phrasing of Marx
The guy was telling a story about capitalist economies in the 1800s. Little, if anything, about the next 100+ years has proven his stories or concerns valid.
Most of the alarmist writing I see about AI displacing workers or suddenly and drastically changing the economy seem to be written by people who are pretty removed from technology or the actual businesses in which AI is allegedly replacing people.
Their theses on AI are not as much a criticism of AI as much as a desire to frame capitalism itself within their own system of ethics and advocate for some form of reform; which is to say I feel that were AI as we know it to be inexistent, they would still find themselves arguing against technology in general, or more precisely against efficiency, with a presupposed socialist-adjacent premise that capitalism only incentivizes greed.
Most of the alarmist writing I see about AI displacing workers or suddenly and drastically changing the economy ... are from pro-ai boosters trying to create hype around AI.
They are not criticism of ai, they are openly stated goals and aspirations.
There's a lot of that type of slop on LinkedIn (with accompanying slop comments) but they're not taken anywhere as seriously as the more thoughtful blog posts like this
Also in technical circles, we're pretty aware of the marketing/self-promotional nuance of the "our AI accidentally hacked X company" type of press releases, which perhaps the general public may not take with as large a grain of salt
I dont read linkedin. I was not referencing linkedin. Among other sources, the primary one is literally AI companies themselves. And people close to them.
The "it will replace workers" is pro-ai talking and selling point.
I agree there, I think if we differentiate between first and second order AI companies (companies producing models vs applying those models to everyday tasks) I do think a lot of the hype comes from the 2nd order companies, since they necessarily must make a bold claim to be worthy of VC money. A lot of the 1st order companies are marketing their AI to consumers who are doing work, rather than looking to market their AI to business who are replacing those consumers. If anyone has the means to replace a worker, it's definitely OpenAI or Anthropic moreso than a 2nd order startup trying to replace the human doing the job.
And on a restrospective point, I think one of the easiest jobs for AI to do right now is software engineering. But on the whole, while there certainly have been layoffs, software engineering hiring has still been mostly a net positive[0].
And that's just for formal software engineering jobs; if we define a software engineer as a human creating software, then I think AI as a tool has created a LOT more (perhaps not employed as one, but say, a realtor creating their own sofware for personal use) software engineers than the amount that it's actually replaced. In a sense, because of AI, there are more software engineers now than ever before.
One of the people in the company I work for/with is a non-programmer that has a “vibe coding consulting” gig on the side.
Basically half of their Linkedin is about how they’re shipping features at half the speed as developers, with quality, etc.
They’re not.
They took about three months to prototype something that was rejected for not working/unreadable code. Another team built the same feature from scratch in a week.
Seeing the other side of the grift is kind of pathetic.
I think there is a time and place where the outputs of LLMs belong.
The issue is this requires NUANCE.
And the more certain people expose themselves to LLMs and forgo that ability to sit there and think painstakingly about nuanced things - the more useless they become IMO.
U.S. unemployment rate is steady at ~4% and we are now almost 4 years after the release of ChatGPT. The Jobs Apocalypse seems to be always on the horizon. UBI is being pushed by college educated workers burnt out of white collar work (understandably), but when you suggest to them working construction, teaching school, or working in a hospital, they deflect and go quiet.
I have two ideas you might like:
1) initial signs of AI adoption harming job growth got the signal backwards: It's not that we didn't hire Jrs because we had AI now, its that we didn't hire Jrs because they used AI now, and it was hard to reconcile this with our standing hireability signals: code quality, side projects, portfolio, etc.
2) The TAM argument for AI replacing all work is not even required. Anthropic and company (and several future companies) can become fantastically wealthy beyond anyone's dreams just by following claude code subscription model to make life easier for workers, who in turn gladly pay for the subcription.
both are here with more details, apologies for links to personal (ad-free) blog
https://jodavaho.io/posts/ai-signalling.html
https://jodavaho.io/posts/ai-jobpocolypse.html
But total unemployment is going up: https://fred.stlouisfed.org/series/U6RATE
Correlation isn't causation; in the last few years we've experienced some significant economic headwinds like COVID and the current oil crisis. AI rollout is not happening in a vacuum.
No, that's not the usual response, because people in this situation have evaluated those alternatives and will be easily able to articulate what is wrong with them (pay). Also, the 4% figure treats doordash as employment. In a parallel universe where the pay was decent, that could be reasonable. In this universe, not so much.
These are boomer-tier talking points and if you bring them up to a kid stuck in this position you will get dunked on and you will deserve it. Unless you found a shy kid, in which case: congratulations on your new title as undefeated shadowboxing champion.
your argument suggests there's an opportunity for a trading strategy. It's also very common as a "technical analysis" investment strategy for any gullible person on the internet.
The problem is it looks "backwards" and tries to make the point something (in this case the argument made in the paper) is false because the backwards data doesn't support it. All while completely failing to address the crux of the point being made.
You realize that this number has been 'kept' at 4% by continuiously qualifying the unemployed as "not looking".
Real unemployment has been steadily growing along U6 numbers
by some measures unemployment is at 25% and this aligns with my personal experience much better than frankly the bs 4% rate that is being pushed by this admin
https://www.lisep.org/tru
in your link, the "true rate" has also not risen since the rise of AI, was higher in 1995
i'm sure it's meant in some kind of technical sense that is precisely defined in the field of economics, but this use of the word "rational" is irritating:
> This demand externality traps rational firms in an automation arms race, displacing workers well beyond what is collectively optimal.
you can't be trapped in a sub-optimal situation and be rational at the same time. the rational thing for these rational competitors to do would be to agree to slow their deployment of ai, because the cost savings will be more than lost to demand destruction down the line. so if they won't or can't do that, they aren't rational. on the flip side, if it is actually impossible to slow ai deployment, then the firms are rational and the situation they find themselves in isn't actually sub-optimal.
it's irritating to me because i often see persuit of self-interest even to one's eventual detriment held up as "rationality", which may be true in some technical sense within the field of economics (i really couldn't say), but is certainly not true for, like, actual rationality. myopic selfishness is never rational.
> you can't be trapped in a sub-optimal situation and be rational at the same time.
In order for this statement to be true, you must also have perfect knowledge: something that is unachievable in most circumstances. Getting trapped in a local-optimal solution is perfectly possible if you are blindly following stochastic decent along a gradient: as it is too with life... getting trapped like this looks like being in the situation where there is no rational action (only irrational ones) that (eventually) lead to anything other that that status quo. a local optimum is where you have to do increasingly (in proportion to the level of optimization) non-positive actions to escape (likely to another local-optimal).
> the rational thing for these rational competitors to do would be to agree
This is the bit that is tripping you up. The market is composed of more than just the these potentially agreeing parties, so there is no case where all parties can even come together to agree. It's rational for any party to break the agreement or be stealthy and attempt to take the upper hand while the visible players are all patting themselves on the back for their self-inflicted restraint. It's rational for one party to encourage everyone to come together to agree, to distract from their own actions to undercut everyone else.
On top of that, such agreement would potentially qualify as collusion, allowing an upstart to cry foul.
It's all game theory.
Yeah it's a pet peev of mine as well. People define very tight space and then proclaim actions irrational if they don't make rational sense in that setting.
Which wouldn't be a problem in itself if there was discussion about the defined space and if it is an actually good model.
A big chunk of economics is like that, if the data doesnt fit the model then too bad for the data. Same goes for metrics used, externalities are like a side note in economy books and not an example of fundamental issues with current economic toola and modeling.
And it's not limited to introductory level either. Nordhaus had most asinine assumptions in his model and he got the pseudo-Nobel.
It's not pretty.
It's rational at the individual level, irrational at the group level.
The issue is you can't easily enforce cooperation without some external overarching force (e.g. regulation), especially across competing powers.
Refer to nuclear arms race in general. A specific example on why not to be the only one to cooperate: Ukraine gave up nukes in order to push for group level rationality, gets walked over as an individual. Oversimplifying, but yeah that's the rough point.
Meditations on Moloch is a favourite of mine on this (albeit very depressing): https://www.slatestarcodexabridged.com/Meditations-On-Moloch
Something in the ballpark of a Nash equilibrium. Kinda amazing how often you run into those things, in situations like groupthink.
> you can't be trapped in a sub-optimal situation and be rational at the same time
Yes you can that’s what tragedy of the commons is. Rational means you take the choice that provides the most value for yourself. Without a cartel (hard) or government regulation this situation happens all the time.
Yes and no. It's not a new problem for humans and it's a pretty solvable problem if social pressure and structures can be applied once it is identified. As long as the resource is to be used by a community that depends on it and not an outsider then we are able to keep ourselves mostly in line for the benefit of the group.
Tragedy of the commons is when a shared resource is rationally used by the individual, but harms the whole.
You can be trapped in sub-optimal situation because it's a local maxima; not because of tragedy of commons, which is something different.
Society can be trapped in local maxima and that would be because of tragedy of commons, and individuals can be trapped in local maxima but that's distinct from tragedy of commons.
The term you are looking for is what Douglas Hofstadter called "superrationality" [https://en.wikipedia.org/wiki/Superrationality].
Remember when the buttons test was circulating a few months ago? Where if you press the red button, you survive, but if you press the blue button, unless more than 50% of people chose the blue button you die? That was a case very similar to this. It is locally optimal for everyone to choose red; your own outcome is always better or equal to what it would have been if you chose blue, however everyone independently making this optimal choice is catastrophic on a global scale. Thats why its "rational".
You can... check tragedy of the commons, and prisoners' dilemma
> A Pigouvian automation tax can
The government will try at all costs to implement an AI tax, just because they can and because that's their job, to fleece us all. Then they will try in the name of 'peace/climate/asteroid/terrorism' to make the whole world comply or China will eat their lunch
Daron Acemoglu (2024 nobel economist) is on a book tour with some new ideas about fundamental changes to our economic theories from Ai (automation/technology are not uniform), this was a great interview
https://www.youtube.com/watch?v=e5O1Y-5phEU
Not compelling for the basic reason laid out at the start of the article: this is just a re-phrasing of Marx
The guy was telling a story about capitalist economies in the 1800s. Little, if anything, about the next 100+ years has proven his stories or concerns valid.
Most of the alarmist writing I see about AI displacing workers or suddenly and drastically changing the economy seem to be written by people who are pretty removed from technology or the actual businesses in which AI is allegedly replacing people.
Their theses on AI are not as much a criticism of AI as much as a desire to frame capitalism itself within their own system of ethics and advocate for some form of reform; which is to say I feel that were AI as we know it to be inexistent, they would still find themselves arguing against technology in general, or more precisely against efficiency, with a presupposed socialist-adjacent premise that capitalism only incentivizes greed.
Most of the alarmist writing I see about AI displacing workers or suddenly and drastically changing the economy ... are from pro-ai boosters trying to create hype around AI.
They are not criticism of ai, they are openly stated goals and aspirations.
There's a lot of that type of slop on LinkedIn (with accompanying slop comments) but they're not taken anywhere as seriously as the more thoughtful blog posts like this
Also in technical circles, we're pretty aware of the marketing/self-promotional nuance of the "our AI accidentally hacked X company" type of press releases, which perhaps the general public may not take with as large a grain of salt
I dont read linkedin. I was not referencing linkedin. Among other sources, the primary one is literally AI companies themselves. And people close to them.
The "it will replace workers" is pro-ai talking and selling point.
I agree there, I think if we differentiate between first and second order AI companies (companies producing models vs applying those models to everyday tasks) I do think a lot of the hype comes from the 2nd order companies, since they necessarily must make a bold claim to be worthy of VC money. A lot of the 1st order companies are marketing their AI to consumers who are doing work, rather than looking to market their AI to business who are replacing those consumers. If anyone has the means to replace a worker, it's definitely OpenAI or Anthropic moreso than a 2nd order startup trying to replace the human doing the job.
And on a restrospective point, I think one of the easiest jobs for AI to do right now is software engineering. But on the whole, while there certainly have been layoffs, software engineering hiring has still been mostly a net positive[0].
And that's just for formal software engineering jobs; if we define a software engineer as a human creating software, then I think AI as a tool has created a LOT more (perhaps not employed as one, but say, a realtor creating their own sofware for personal use) software engineers than the amount that it's actually replaced. In a sense, because of AI, there are more software engineers now than ever before.
[0]: https://newsletter.pragmaticengineer.com/p/state-of-the-job-...
About pro-AI boosters:
One of the people in the company I work for/with is a non-programmer that has a “vibe coding consulting” gig on the side.
Basically half of their Linkedin is about how they’re shipping features at half the speed as developers, with quality, etc.
They’re not.
They took about three months to prototype something that was rejected for not working/unreadable code. Another team built the same feature from scratch in a week.
Seeing the other side of the grift is kind of pathetic.
I think there is a time and place where the outputs of LLMs belong.
The issue is this requires NUANCE.
And the more certain people expose themselves to LLMs and forgo that ability to sit there and think painstakingly about nuanced things - the more useless they become IMO.
I agree. It requires good judgement, domain knowledge, good work ethic.
Grifters who think the LLM can do it all for them will definitely not survive in a competitive market.