I mean, it's nice that the article starts listing its assumptions, but then it forgets it's talkung about a small probability slice and acts lime this is the one problem to solve.
We do have a recent example of industrial revolution. I don't know how US is doing, but Europe is still split in productive urban centers and mostly unproductive rural. Their solution was to throw money at the problem - if you drive in a small european village, chances are it has better roads and utilities than a large city, even if they were built at 10-50x the cost per capita. And in eastern europe this is mixed with corruption and political capture.
Bigger problem is that it's been over half a century since the latest agricultural mecanisation, and the problem is still mitigated, not solved. There is a bit of agriculture, a bit of tourism, with better transport some villages are now becoming suburbs (especially in denser areas like Switzerland), but all in all people there are poor purposeless, and vote accordingly.
A likely path forward is the same, only with white colar workers.
I'm not sure I understand your point here. Are you saying that the essay should've set out a bunch of assumptions, and then gone on to answer things beyond those assumptions?
No, it's a fair article in itself, and makes its point well. I'm only pointing out a slighly different PoV - pouring money at the problem will only mask it.
Or, more generaly: building a society without work as a foundation is something we know we don't know how to do. Large scale socialist societies invariably turned into dystopias (I was born in one), and many of the redeeming qualities were about... work.
So instead of dividing future spoils, I'd focus on things like empowering people and making sure comparative advantage will still work.
For example, a lot of current protectionism assumes available workforce is fixed and regulations uniform. E.g. minimum wage. Today we can quibble if its effects are mostly good or mostly bad. In 10 years it'll simply make half the population unemployable, UBI or not.
Trade is usually mutually beneficial. Free markets usually work. It seems very odd to argue that Nigeria should reject, even ban, foreign AI, without explaining what market failure that avoids. My instinct is that if there's some US-owned monopoly Final AI priced exorbitantly, Nigeria should let Nigerians use it as they please, because the populace can figure out when they are getting from it more than they are paying for it and vice versa. Perhaps the most valuable thing for them to do with that intelligence is put it towards developing their own national competitor, but banning it seems value-destructive.
The simple answer: “Final AI is, by fiat, really different.”
Longer answer: As I said in the opening, if AI only ever amounts to being the same as the fax machine or the Internet (i.e. it opens up new jobs and sufficient labor opportunities remain for humans in general), then none of this applies. We’re assuming that something like the “progressive disempowerment” thesis holds and that at some point in the future most people around the world will be essentially unemployable. Whatever jobs are left will have such a huge worker surplus available to them that the bargaining power of labor drops to basically zero. Assuming an American AI lab wins, the US can handle this by acquiring that lab and creating a dividend-paying SWF, but what about the rest of the world? How can they trade their way out of poverty now?
As Amol notes in his answers, countries that get rich do so by plugging in to global supply chains. Traditionally, this meant starting with manufacturing and then working your way up to services as your ability to compete (and standard of living) converges. The Final AI world – which is admittedly literally science fiction! – is one in which rich countries no longer need offshore services in general, whether it’s highly technical stuff or it’s simple backoffice IT work or call centers. The ladder for poorer countries to converge loses the important rungs. They can still do manufacturing or resource-extraction work, but if the vision is that this is all they’re able to do, with massive and still-expanding populations competing to sate whatever rich-world manufacturing demand remains, I think the future looks pretty damn dire for most of humanity. The contrast with wealthy countries, whose citizens might not be doing any work at all, makes this look even worse.
I’m not sure if you were onboard for the whole sovereign wealth fund part, but if you were, the most obvious alternative hope (assuming we agree that “leave most of humanity in the dust” is an unacceptable conclusion) would be something like a global dividend-paying wealth fund based on AI holdings. If that’s politically possible, I’m really not fussed either way. But most of these conversations just talk about the US government acquiring bits of OpenAI, and something like “American Final AI is created -> other countries distill the model and the nationalize their own AI infrastructure” strikes me as somewhat more likely than “an American SWF generously agrees to share their dividends with the world” (shoutout the USAID closure, I’m still malding about that).
I mean, it's nice that the article starts listing its assumptions, but then it forgets it's talkung about a small probability slice and acts lime this is the one problem to solve.
We do have a recent example of industrial revolution. I don't know how US is doing, but Europe is still split in productive urban centers and mostly unproductive rural. Their solution was to throw money at the problem - if you drive in a small european village, chances are it has better roads and utilities than a large city, even if they were built at 10-50x the cost per capita. And in eastern europe this is mixed with corruption and political capture.
Bigger problem is that it's been over half a century since the latest agricultural mecanisation, and the problem is still mitigated, not solved. There is a bit of agriculture, a bit of tourism, with better transport some villages are now becoming suburbs (especially in denser areas like Switzerland), but all in all people there are poor purposeless, and vote accordingly.
A likely path forward is the same, only with white colar workers.
I'm not sure I understand your point here. Are you saying that the essay should've set out a bunch of assumptions, and then gone on to answer things beyond those assumptions?
No, it's a fair article in itself, and makes its point well. I'm only pointing out a slighly different PoV - pouring money at the problem will only mask it.
Or, more generaly: building a society without work as a foundation is something we know we don't know how to do. Large scale socialist societies invariably turned into dystopias (I was born in one), and many of the redeeming qualities were about... work.
So instead of dividing future spoils, I'd focus on things like empowering people and making sure comparative advantage will still work.
For example, a lot of current protectionism assumes available workforce is fixed and regulations uniform. E.g. minimum wage. Today we can quibble if its effects are mostly good or mostly bad. In 10 years it'll simply make half the population unemployable, UBI or not.
Why?
Trade is usually mutually beneficial. Free markets usually work. It seems very odd to argue that Nigeria should reject, even ban, foreign AI, without explaining what market failure that avoids. My instinct is that if there's some US-owned monopoly Final AI priced exorbitantly, Nigeria should let Nigerians use it as they please, because the populace can figure out when they are getting from it more than they are paying for it and vice versa. Perhaps the most valuable thing for them to do with that intelligence is put it towards developing their own national competitor, but banning it seems value-destructive.
From Nikhil
The simple answer: “Final AI is, by fiat, really different.”
Longer answer: As I said in the opening, if AI only ever amounts to being the same as the fax machine or the Internet (i.e. it opens up new jobs and sufficient labor opportunities remain for humans in general), then none of this applies. We’re assuming that something like the “progressive disempowerment” thesis holds and that at some point in the future most people around the world will be essentially unemployable. Whatever jobs are left will have such a huge worker surplus available to them that the bargaining power of labor drops to basically zero. Assuming an American AI lab wins, the US can handle this by acquiring that lab and creating a dividend-paying SWF, but what about the rest of the world? How can they trade their way out of poverty now?
As Amol notes in his answers, countries that get rich do so by plugging in to global supply chains. Traditionally, this meant starting with manufacturing and then working your way up to services as your ability to compete (and standard of living) converges. The Final AI world – which is admittedly literally science fiction! – is one in which rich countries no longer need offshore services in general, whether it’s highly technical stuff or it’s simple backoffice IT work or call centers. The ladder for poorer countries to converge loses the important rungs. They can still do manufacturing or resource-extraction work, but if the vision is that this is all they’re able to do, with massive and still-expanding populations competing to sate whatever rich-world manufacturing demand remains, I think the future looks pretty damn dire for most of humanity. The contrast with wealthy countries, whose citizens might not be doing any work at all, makes this look even worse.
I’m not sure if you were onboard for the whole sovereign wealth fund part, but if you were, the most obvious alternative hope (assuming we agree that “leave most of humanity in the dust” is an unacceptable conclusion) would be something like a global dividend-paying wealth fund based on AI holdings. If that’s politically possible, I’m really not fussed either way. But most of these conversations just talk about the US government acquiring bits of OpenAI, and something like “American Final AI is created -> other countries distill the model and the nationalize their own AI infrastructure” strikes me as somewhat more likely than “an American SWF generously agrees to share their dividends with the world” (shoutout the USAID closure, I’m still malding about that).