I hadn’t put these two developments together before. I can see how they both challenge our sense of purpose and the nature of community. My experience with childless couples is that they often strive to recreate parental relationships in their relationships with nieces and nephews or with close family friends and their children. You might say they become a bit like grandparents. I imagine many of us will strive to recreate the feelings of being obligated to work if work becomes materially unnecessary. What then in the work world is the equivalent of the quasi-grandparent I just talked about? Board member? Mentor? Troubleshooter? Benefactor? Or maybe companies will become like sports team where people root for certain products and service providers for tribal reasons. Companies may find that building loyalty by returning consumers some sense of shared purpose is how they differentiate in a world where every product and service is provided by a super intelligent systems. It wouldn’t be because the super intelligent systems need us to make the products better but because they need enough favorably inclined people not to have their plugs pulled.
Good points. We have a fairly long history of a segment of people not having had to work. Aristocrats in the feudal period, heirs to great fortunes in the modern world.
AI seems to be at the level of intelligence where it can replace economists and investment bankers. When it can replace plumbers I'll regard that as a revolution.
I will likely replace white collar workers before plumbers, as you say. But I'm not sure it's even at the level where it could replace white collar workers. It's smart enough, but does it have the other skills?
We think the substitution effect of work for leisure, particularly for females, reduces fertility. But can we think of AI as generating an income effect that causes us to have more children? That is, how much do we need AGI to raise economic growth to where fertility rates turn back upward? I don't have an answer, but I don't think I have phrased the question that way before reading your post. Thanks for offering your thoughts in an open, thought-provoking way.
I was thinking that AI might reduce both male and female labor, and thus boost fertility.
BTW, while I think your claim about female labor has some merit, I seem to recall that there's a surprising tendency for fertility to be a bit higher in developed countries with high female LFP than in countries where it's low. Someone please correct me if that's wrong. (Within the US, I suspect fertility is higher in places with lower female labor force participation.)
The fertility rate strikes me as an issue where most people take a pessimistic view no matter the evidence. There still seems to be many environmentalist Paul Ehrlich types who insist that overpopulation is a terrible problem and make increasingly hollow claims about the dangers of exponential growth and high fertility even as more and more places undergo demographic transitions.
Then there the population aging alarmists (and at the more hysterical end, extinction or great replacement theory). Which again seems overblown - it's not like Japan has imploded as its old-age dependency ratio has crept up to 50%, and it should eventually stabilise. I don't see why going to 70% (where Japan will be by about 2060) should be so much worse than the rise from 30% (in 2004) to today - markets adjust, people work longer, labour savings are found.
To me it seems the world is more or less in a goldilocks scenario. It turns out, when given access to reproductive control technology and at least middle income, most places seem to stabilise birthrates somewhere between 1 and 2 with absolutely no coercion or international cooperation required, which rapidly brings down population growth without large reductions in human welfare (though to be fair, we haven't observed the full effects of aging play out). It seems like a real win for humanity, and doesn't get the celebration it deserves.
To be clear, I'm not especially worried about fertility. I suspect that in the long run people will find a fertility level that they prefer, and the economy will adjust.
Why do people conflate AI and robots? It reminds me of discussions a decade ago, when people would routinely muddle EVs and self-driving vehicles, assuming in their discourse that autonomous vehicles would necessarily be electric, and vice versa. [*]
And assuming that both were imminent: everyone would have an electric AV in five years. How did that go?
AI can help with the production of non-rivalrous non-material goods. Perhaps. (Disney, Sony and other IP-owning companies will try very hard to prevent that.)
Rivalrous and physical goods are a whole other kettle of onions. We have had vacuum-cleaning robots for over two decades now, and they still can't vacuum floors except in the most limited of circumstances. Let alone vacuum stairs or dust furniture. There is little reason to expect progress to suddenly accelerate.
* People also almost universally ignored supply and demand, claiming that by lowering the cost of getting around in cars, AVs would reduce the number of vehicles on the road and the time spent in traffic.
My point is that robots are a whole other class of problems orthogonal to AI, and progress on those is slow and halting. To me it doesn't seem wise to assume "when we have AI, robots will instantly follow".
I’m about your age so my worries about the future are the next twenty or twenty-five years. (smile) You and I may be lucky - the impact of population collapse may be mitigated by robots and smart devices that take care of us. The overall impact of AI may be slower than predicted because the primary bottleneck to productive change in much of the public and private sectors is not fast intelligence. For example, it won’t quickly revolutionize the building of houses or transportation systems in cities. My bigger worry over the next twenty years is the rapid collapse of trusted sources of information - replaced by the chaos of social media / smart devices / algorithms that keep us engaged in unproductive nonsense and feed tribalism, loss of empathy, and loss of community. We need more awareness and work on that one. I enjoyed your article - keep up the good work.
I agree that while I have written a few pieces on the fertility crisis, bemoaning a collapse in the human population is almost neomalthusiansim in its own right. Instead of seeing a population bomb outstripping food, we see a population collapse leaving an "empty planet."
I have, in the past, connected the topics of AI and fertility, noting that AI is emerging just as the supply of humans who can perform physical and cognitive labor may decline. This is a remarkable coincidence, no?
But was it a coincidence that the Haber-Bosch process was invented just as guano deposits ran dry in the 1910s? Was it also a coincidence that just as the population "bomb" was published, new dwarf varieties of wheat and rice doubled food output for hungry babies?
Perhaps AI is emerging because there are too few of us and the market is responding in kind. What this means for fertility, I don't know. But it may mean that the "empty planet" dystopia scenario won't come to pass either.
For this reason, I described AI as the potential Deux Ex Machina of the fertility crisis, which ironically translated to "God from the Machine." You may want to read this one: https://www.lianeon.org/p/god-from-the-machine
Thanks. That's a very interesting post. I'm not sure what to make of the coincidences that you mention--perhaps some of them were driven by market forces. If we get something like cheap fusion in the next decade (to address global warming), that would support your argument
Scott, I am still more worried about the ongoing resurgence of fascistoid nationalism, and being harmed by wetware that isn't even intelligent (pandemics). Oh and that asteroid.
Like you have said below and many have said before, AI will likely affect white collar jobs before blue collar. But even there, certain white collar activities - most notably, tech - will be affected more than others. I work at an independent government regulatory agency. Our decision-makers are appointed by government, which is itself necessarily (under law) made up of humans, who are themselves voted into office by humans. Human decision-makers tend to like human advisors and staff to boss around. Further, there is much trepidation in government agencies about the use of AI tools due to privacy and commercial confidentiality concerns. Therefore, AI might take a bit longer to make my role redundant than that of a computer engineer. Or so I hope, and if I get 10 years to save up a bit more, I won't mind being offered a payout.
As for where AI leads, one option is incredible growth, material abundance and leisure. Another not-mutually-exclusive option is a form of stagnation as AI becomes smarter at appealing to (pandering to? manipulating?) the emotions of consumers and voters in a way that serves certain ends, but reduces political, social and institutional stability. Already, the Twitter and TikTok 'algorithms' (AIs) have gotten very good at intensifying emotions and opinions. While I like to think I am pretty good at curating my feeds to find useful information and can overlook the dross and nonsense, a large proportion of the population can't handle it, just as many can't handle alcohol, drugs and gambling. Decision-making gets harder as more people have stronger opinions and a greater ability to complain or oppose, governments become more populist and interventionist and institutional environments become less predictable. When AIs can effortlessly facilitate 10,000 people objecting to a new apartment development (as compared to 10 or 100), or can persuasively justify introducing a new policy or the repeal of a previous policy, what progress can be made? The arguments and explanations and 'stakeholder management' may get a whole lot better. But what about the actual decisions that get made? Have my agency's decisions become better in light of the more advanced economics available today, the scope to scan and assimilate thousands more documents, and the greater ability to review and refine the documents we prepare? Do corporates pursue DEI or not? Do they pursue Net Zero ambitions or not? Do they strive for success, only to get targeted for regulation? We've just have 4 years of the AOC AI running the US, and now we have 4 years (?) of the Trump AI. What good stuff has and will come out of that? The newish British and Australian governments may struggle to get re-elected as a significant proportion of the electorate peels off to various special interest parties (AIs), making single-party government more difficult, like in Europe. In your previous post, you described Western Europe (correctly, in my view) as the most 'civilised' place in the world (although I think Australia and NZ are even more so). And what do we see in Europe? Intervention and stagnation. All that brainpower, lots of regulation, but not much progress. So I'm on the pessimistic side of things as well.
I almost mentioned Australia and New Zealand as two countries that were very unlikely to go to war with each other. I wonder if some of the pushback I got from people is due to the fact that most of us forget just how acceptable war used to be. Germany invades France in 1940? Japan bombs the US? Like, why would they do something so stupid? But that was the world when my dad was 20 years old.
I looked at a few comments on your post, but couldn't see much pushback on the claim that Western Europe is the most civilised place on Earth. Assuming it was there, I can think of several reasons why. First, people born after the early 1980s would have no memory of the Iron Curtain, and hence the notion of 'Western Europe' as a distinct political and cultural region. Central and eastern Europe has had its issues, from the wars in the former Yugoslavia in the 1990s to illiberalism in Hungary, to Russia-Ukraine today. Second, Europe seems to have had a lot more issues with the assimilation of Muslim refugees and migrants than the New World, probably due to its more rigid labour markets. Aside from the news about sexual harassment and assaults and the pimping of young women, there have been far more Islamist terrorist attacks in Europe than in the New World over the last 2 decades. All that gives the impression that Europe doesn't really have its domestic security house in order. (Frankly, the frequent gun massacres in the US give the same or worse impression to me.) Finally, while most western European cities are safe to walk around at night, they need to separate their soccer team supporters into different fenced areas. I presume this doesn't happen in the US and it certainly doesn't happen in Australia. That hardly seems civilised.
"Was "Event Horizon" instead of "Singularity" a conscious choice?"
No, I had forgotten that singularity is the more common expression. And as Gwern points out, I used the wrong term. But I assume that people know what I meant. Sort of like when an average person uses "light years" as a metaphor for vast periods of time, not distance. In context, the meaning is usually clear.
When it comes to science, I'm an "average person".
FWIW...I think you used the term "event horizon" accurately sometimes and should have used "singularity" a couple of other times. Sometimes you were talking simply about the limits of our knowledge and not what was creating the event horizon. I think it becomes even more interesting when you take the "event horizon" analogy further. We try to understand the nature of the singularity by studying the behavior of everything on our side of the event horizon because that's all the information we are capable of grasping, considering that we have never experienced the gravitational forces of a black hole and the fact that nothing (except a little Hawking radiation) escapes beyond the event horizon. How do we figure out what it's like to live in a world where human work is not needed for survival or even the creation of artistic artifacts? Well, we can look at how aristocrats lived and see what they did as you suggested above in your response to my comment.
Presumably. But as a terminological change, it's not an improvement because it's misleading where it's not irrelevant, compared to the original, which captures exactly what OP wants and needs to say.
An "event horizon" is nothing much; it is the 'point of no return' where all trajectories lead to the singularity. At the event horizon, nothing important might happen. (With a massive black hole, a human astronaut could pass the event horizon without being torn apart, even.) Everything looks normal, and you can predict everything that will happen for potentially arbitrarily long into the future... until you hit the singularity at the center. (Which might be trillions of years subjectively in your future.) And as a metaphor for the current moment, it is either too weak or too strong: if by 'possible trajectory' you mean business as usual always leads to the Singularity, then the 'event horizon' may have been centuries or millennia ago, and if you mean every trajectory, then we are probably nowhere near it (there are plenty of things that *could* stop or delay it, like a CCP invasion of Taiwan or if nothing else, a dino-killer asteroid).
A singularity, on the other hand, is the point at which everything changes and we have no idea what, if anything, happens at that point. There are many theories, but little knowledge. Everything from 'nothing' to 'you get spat out into a parallel universe' (white hole). No one can predict what happens after hitting the singularity.
And OP's problem is not that we have passed some point of no return which guarantees a certain far-distant future event, but that he can't *predict* what happens *after* that future event. Do you need to worry about government debt? Will fertility fix itself if everyone has robot nannies, chauffeurs, and teachers? Will the CCP stop being a problem? In other words, OP's problem is precisely why Vernor Vinge named the Singularity concept 'the Singularity' 32 years ago: https://edoras.sdsu.edu/~vinge/misc/singularity.html
> I think it's fair to call this event a singularity ("the Singularity" for the purposes of this paper). It is a point where our models must be discarded and a new reality rules. As we move closer and closer to this point, it will loom vaster and vaster over human affairs till the notion becomes a commonplace. Yet when it finally happens it may still be a great surprise and a greater unknown.
Neither Vinge nor you nor me nor OP can tell you for sure what that new reality is, after that Singularity point. That's what makes it a singularity in time, rather than an 'AI event horizon'. So OP should just call a spade a spade.
Running with your idea of an event horizon, I’d like to see people think more about how everyday laws and assumptions start to break down as we approach that limit.
Politically, the initial effects will be handled well enough, but I suspect that the system buckles before we reach the ability to radically reorder life under abundant AI. Who is out of work? Just how much unemployment can the system take? A country with 10% unemployment can muddle through, but not 50%. And how does this play out amongst the haves and have-nots internationally?
And for a world based on capitalism, what happens as we close in on that event horizon and the very notion of money starts to break down? The economy starts to resemble a single privately-owned resort that exists to make a profit, but all that money comes from the UBI of its customers, and UBI comes from taxes on the resort. It has the flavor of that old thought experiment where we imagine a world where one party can print unlimited money.
So, under radical AI, the only prediction I would make is that the laws of politics and economics must be radically different.
Yes, I agree that things would be quite different. Rather than 50% unemployment, we might first shift to a 20-hour work week. Real wages would be so high that 20 hours would be enough.
I’m unconvinced this will play out so well in practice. AI will decimate some sectors while leaving others completely untouched. A 20-hour work week is unlikely to help, in my opinion, although I’d love to be proven wrong. I suppose we will see where the great axe first falls and the general reaction.
Technological progress has always decimated some sectors more than others. I recall when an army of women worked as phone operators, jobs that were wiped out when electronic switching came on stream. The labor market can handle those things as long as new jobs open up.
Like you, I don't know anything. But I put the highest probability outcome (maybe a 40% chance?) that we have currently hit an AI plateau that will persist for our lifetimes. (I am 47.) Term papers (employing good grammar) are now much cheaper to produce, as are cute illustrations. But life will go on for the next generation or so of workers, pretty much as it has gone on for the previous three. Birth rates will continue to decline, but slowly, and migration will keep business as usual running in currently large cities in rich countries.
Yes, I think it's very possible that this is less transformative than many people (including me) assume. But I don't think "plateau" is the most likely outcome---more likely a slower rate of progress than the optimists assume. Hard to see how it doesn't gradually transform our economy, one robot at a time.
Agreed. But some of those changes in China are so dramatic, over such a short period of time, that they cannot possibly be explained by a changing number of women in that age group.
The scenario where robots do most everything seems to skip a big segment of the population - those the sell and distribute drugs and those that use them. If more activities are displaced by A{ automation will not a bigger percent of the population turn to drugs to give a portion of their lives meaning (if meaning can be found in the haze of drugs) and will not a bigger percent of the population seek innovation or artistic experiences through drugs. Where is the thinking on that segment of the population that is either currently left out or those that will be displaced?
I hadn’t put these two developments together before. I can see how they both challenge our sense of purpose and the nature of community. My experience with childless couples is that they often strive to recreate parental relationships in their relationships with nieces and nephews or with close family friends and their children. You might say they become a bit like grandparents. I imagine many of us will strive to recreate the feelings of being obligated to work if work becomes materially unnecessary. What then in the work world is the equivalent of the quasi-grandparent I just talked about? Board member? Mentor? Troubleshooter? Benefactor? Or maybe companies will become like sports team where people root for certain products and service providers for tribal reasons. Companies may find that building loyalty by returning consumers some sense of shared purpose is how they differentiate in a world where every product and service is provided by a super intelligent systems. It wouldn’t be because the super intelligent systems need us to make the products better but because they need enough favorably inclined people not to have their plugs pulled.
Good points. We have a fairly long history of a segment of people not having had to work. Aristocrats in the feudal period, heirs to great fortunes in the modern world.
AI seems to be at the level of intelligence where it can replace economists and investment bankers. When it can replace plumbers I'll regard that as a revolution.
I will likely replace white collar workers before plumbers, as you say. But I'm not sure it's even at the level where it could replace white collar workers. It's smart enough, but does it have the other skills?
We think the substitution effect of work for leisure, particularly for females, reduces fertility. But can we think of AI as generating an income effect that causes us to have more children? That is, how much do we need AGI to raise economic growth to where fertility rates turn back upward? I don't have an answer, but I don't think I have phrased the question that way before reading your post. Thanks for offering your thoughts in an open, thought-provoking way.
I was thinking that AI might reduce both male and female labor, and thus boost fertility.
BTW, while I think your claim about female labor has some merit, I seem to recall that there's a surprising tendency for fertility to be a bit higher in developed countries with high female LFP than in countries where it's low. Someone please correct me if that's wrong. (Within the US, I suspect fertility is higher in places with lower female labor force participation.)
The fertility rate strikes me as an issue where most people take a pessimistic view no matter the evidence. There still seems to be many environmentalist Paul Ehrlich types who insist that overpopulation is a terrible problem and make increasingly hollow claims about the dangers of exponential growth and high fertility even as more and more places undergo demographic transitions.
Then there the population aging alarmists (and at the more hysterical end, extinction or great replacement theory). Which again seems overblown - it's not like Japan has imploded as its old-age dependency ratio has crept up to 50%, and it should eventually stabilise. I don't see why going to 70% (where Japan will be by about 2060) should be so much worse than the rise from 30% (in 2004) to today - markets adjust, people work longer, labour savings are found.
To me it seems the world is more or less in a goldilocks scenario. It turns out, when given access to reproductive control technology and at least middle income, most places seem to stabilise birthrates somewhere between 1 and 2 with absolutely no coercion or international cooperation required, which rapidly brings down population growth without large reductions in human welfare (though to be fair, we haven't observed the full effects of aging play out). It seems like a real win for humanity, and doesn't get the celebration it deserves.
To be clear, I'm not especially worried about fertility. I suspect that in the long run people will find a fertility level that they prefer, and the economy will adjust.
Why do people conflate AI and robots? It reminds me of discussions a decade ago, when people would routinely muddle EVs and self-driving vehicles, assuming in their discourse that autonomous vehicles would necessarily be electric, and vice versa. [*]
And assuming that both were imminent: everyone would have an electric AV in five years. How did that go?
AI can help with the production of non-rivalrous non-material goods. Perhaps. (Disney, Sony and other IP-owning companies will try very hard to prevent that.)
Rivalrous and physical goods are a whole other kettle of onions. We have had vacuum-cleaning robots for over two decades now, and they still can't vacuum floors except in the most limited of circumstances. Let alone vacuum stairs or dust furniture. There is little reason to expect progress to suddenly accelerate.
* People also almost universally ignored supply and demand, claiming that by lowering the cost of getting around in cars, AVs would reduce the number of vehicles on the road and the time spent in traffic.
I should clarify that when I mentioned robots I meant human-like machines with high intelligence, not vacuum cleaners.
My point is that robots are a whole other class of problems orthogonal to AI, and progress on those is slow and halting. To me it doesn't seem wise to assume "when we have AI, robots will instantly follow".
It seems to me that recent progress in robots has been pretty fast. I'd expect human-type robots within 20 or 30 years, capable of many routine tasks.
I’m about your age so my worries about the future are the next twenty or twenty-five years. (smile) You and I may be lucky - the impact of population collapse may be mitigated by robots and smart devices that take care of us. The overall impact of AI may be slower than predicted because the primary bottleneck to productive change in much of the public and private sectors is not fast intelligence. For example, it won’t quickly revolutionize the building of houses or transportation systems in cities. My bigger worry over the next twenty years is the rapid collapse of trusted sources of information - replaced by the chaos of social media / smart devices / algorithms that keep us engaged in unproductive nonsense and feed tribalism, loss of empathy, and loss of community. We need more awareness and work on that one. I enjoyed your article - keep up the good work.
Thanks, those are all plausible arguments.
Interesting piece Scott.
I agree that while I have written a few pieces on the fertility crisis, bemoaning a collapse in the human population is almost neomalthusiansim in its own right. Instead of seeing a population bomb outstripping food, we see a population collapse leaving an "empty planet."
I have, in the past, connected the topics of AI and fertility, noting that AI is emerging just as the supply of humans who can perform physical and cognitive labor may decline. This is a remarkable coincidence, no?
But was it a coincidence that the Haber-Bosch process was invented just as guano deposits ran dry in the 1910s? Was it also a coincidence that just as the population "bomb" was published, new dwarf varieties of wheat and rice doubled food output for hungry babies?
Perhaps AI is emerging because there are too few of us and the market is responding in kind. What this means for fertility, I don't know. But it may mean that the "empty planet" dystopia scenario won't come to pass either.
For this reason, I described AI as the potential Deux Ex Machina of the fertility crisis, which ironically translated to "God from the Machine." You may want to read this one: https://www.lianeon.org/p/god-from-the-machine
Thanks. That's a very interesting post. I'm not sure what to make of the coincidences that you mention--perhaps some of them were driven by market forces. If we get something like cheap fusion in the next decade (to address global warming), that would support your argument
Or possibly a renaissance in nuclear fission. Gen IV reactors are much safer and more efficient. Could be the near-term breakthrough we need.
Scott, I am still more worried about the ongoing resurgence of fascistoid nationalism, and being harmed by wetware that isn't even intelligent (pandemics). Oh and that asteroid.
I'm currently working on a post on authoritarian nationalism.
Like you have said below and many have said before, AI will likely affect white collar jobs before blue collar. But even there, certain white collar activities - most notably, tech - will be affected more than others. I work at an independent government regulatory agency. Our decision-makers are appointed by government, which is itself necessarily (under law) made up of humans, who are themselves voted into office by humans. Human decision-makers tend to like human advisors and staff to boss around. Further, there is much trepidation in government agencies about the use of AI tools due to privacy and commercial confidentiality concerns. Therefore, AI might take a bit longer to make my role redundant than that of a computer engineer. Or so I hope, and if I get 10 years to save up a bit more, I won't mind being offered a payout.
As for where AI leads, one option is incredible growth, material abundance and leisure. Another not-mutually-exclusive option is a form of stagnation as AI becomes smarter at appealing to (pandering to? manipulating?) the emotions of consumers and voters in a way that serves certain ends, but reduces political, social and institutional stability. Already, the Twitter and TikTok 'algorithms' (AIs) have gotten very good at intensifying emotions and opinions. While I like to think I am pretty good at curating my feeds to find useful information and can overlook the dross and nonsense, a large proportion of the population can't handle it, just as many can't handle alcohol, drugs and gambling. Decision-making gets harder as more people have stronger opinions and a greater ability to complain or oppose, governments become more populist and interventionist and institutional environments become less predictable. When AIs can effortlessly facilitate 10,000 people objecting to a new apartment development (as compared to 10 or 100), or can persuasively justify introducing a new policy or the repeal of a previous policy, what progress can be made? The arguments and explanations and 'stakeholder management' may get a whole lot better. But what about the actual decisions that get made? Have my agency's decisions become better in light of the more advanced economics available today, the scope to scan and assimilate thousands more documents, and the greater ability to review and refine the documents we prepare? Do corporates pursue DEI or not? Do they pursue Net Zero ambitions or not? Do they strive for success, only to get targeted for regulation? We've just have 4 years of the AOC AI running the US, and now we have 4 years (?) of the Trump AI. What good stuff has and will come out of that? The newish British and Australian governments may struggle to get re-elected as a significant proportion of the electorate peels off to various special interest parties (AIs), making single-party government more difficult, like in Europe. In your previous post, you described Western Europe (correctly, in my view) as the most 'civilised' place in the world (although I think Australia and NZ are even more so). And what do we see in Europe? Intervention and stagnation. All that brainpower, lots of regulation, but not much progress. So I'm on the pessimistic side of things as well.
I almost mentioned Australia and New Zealand as two countries that were very unlikely to go to war with each other. I wonder if some of the pushback I got from people is due to the fact that most of us forget just how acceptable war used to be. Germany invades France in 1940? Japan bombs the US? Like, why would they do something so stupid? But that was the world when my dad was 20 years old.
I looked at a few comments on your post, but couldn't see much pushback on the claim that Western Europe is the most civilised place on Earth. Assuming it was there, I can think of several reasons why. First, people born after the early 1980s would have no memory of the Iron Curtain, and hence the notion of 'Western Europe' as a distinct political and cultural region. Central and eastern Europe has had its issues, from the wars in the former Yugoslavia in the 1990s to illiberalism in Hungary, to Russia-Ukraine today. Second, Europe seems to have had a lot more issues with the assimilation of Muslim refugees and migrants than the New World, probably due to its more rigid labour markets. Aside from the news about sexual harassment and assaults and the pimping of young women, there have been far more Islamist terrorist attacks in Europe than in the New World over the last 2 decades. All that gives the impression that Europe doesn't really have its domestic security house in order. (Frankly, the frequent gun massacres in the US give the same or worse impression to me.) Finally, while most western European cities are safe to walk around at night, they need to separate their soccer team supporters into different fenced areas. I presume this doesn't happen in the US and it certainly doesn't happen in Australia. That hardly seems civilised.
Good points. I guess it was the previous post where I made that claim.
I suppose I should have said "European Union", not Europe.
Was "Event Horizon" instead of "Singularity" a conscious choice?
"Was "Event Horizon" instead of "Singularity" a conscious choice?"
No, I had forgotten that singularity is the more common expression. And as Gwern points out, I used the wrong term. But I assume that people know what I meant. Sort of like when an average person uses "light years" as a metaphor for vast periods of time, not distance. In context, the meaning is usually clear.
When it comes to science, I'm an "average person".
FWIW...I think you used the term "event horizon" accurately sometimes and should have used "singularity" a couple of other times. Sometimes you were talking simply about the limits of our knowledge and not what was creating the event horizon. I think it becomes even more interesting when you take the "event horizon" analogy further. We try to understand the nature of the singularity by studying the behavior of everything on our side of the event horizon because that's all the information we are capable of grasping, considering that we have never experienced the gravitational forces of a black hole and the fact that nothing (except a little Hawking radiation) escapes beyond the event horizon. How do we figure out what it's like to live in a world where human work is not needed for survival or even the creation of artistic artifacts? Well, we can look at how aristocrats lived and see what they did as you suggested above in your response to my comment.
Thanks. Yes, I was thinking of the limits of our ability to foresee the future.
Presumably. But as a terminological change, it's not an improvement because it's misleading where it's not irrelevant, compared to the original, which captures exactly what OP wants and needs to say.
An "event horizon" is nothing much; it is the 'point of no return' where all trajectories lead to the singularity. At the event horizon, nothing important might happen. (With a massive black hole, a human astronaut could pass the event horizon without being torn apart, even.) Everything looks normal, and you can predict everything that will happen for potentially arbitrarily long into the future... until you hit the singularity at the center. (Which might be trillions of years subjectively in your future.) And as a metaphor for the current moment, it is either too weak or too strong: if by 'possible trajectory' you mean business as usual always leads to the Singularity, then the 'event horizon' may have been centuries or millennia ago, and if you mean every trajectory, then we are probably nowhere near it (there are plenty of things that *could* stop or delay it, like a CCP invasion of Taiwan or if nothing else, a dino-killer asteroid).
A singularity, on the other hand, is the point at which everything changes and we have no idea what, if anything, happens at that point. There are many theories, but little knowledge. Everything from 'nothing' to 'you get spat out into a parallel universe' (white hole). No one can predict what happens after hitting the singularity.
And OP's problem is not that we have passed some point of no return which guarantees a certain far-distant future event, but that he can't *predict* what happens *after* that future event. Do you need to worry about government debt? Will fertility fix itself if everyone has robot nannies, chauffeurs, and teachers? Will the CCP stop being a problem? In other words, OP's problem is precisely why Vernor Vinge named the Singularity concept 'the Singularity' 32 years ago: https://edoras.sdsu.edu/~vinge/misc/singularity.html
> I think it's fair to call this event a singularity ("the Singularity" for the purposes of this paper). It is a point where our models must be discarded and a new reality rules. As we move closer and closer to this point, it will loom vaster and vaster over human affairs till the notion becomes a commonplace. Yet when it finally happens it may still be a great surprise and a greater unknown.
Neither Vinge nor you nor me nor OP can tell you for sure what that new reality is, after that Singularity point. That's what makes it a singularity in time, rather than an 'AI event horizon'. So OP should just call a spade a spade.
Running with your idea of an event horizon, I’d like to see people think more about how everyday laws and assumptions start to break down as we approach that limit.
Politically, the initial effects will be handled well enough, but I suspect that the system buckles before we reach the ability to radically reorder life under abundant AI. Who is out of work? Just how much unemployment can the system take? A country with 10% unemployment can muddle through, but not 50%. And how does this play out amongst the haves and have-nots internationally?
And for a world based on capitalism, what happens as we close in on that event horizon and the very notion of money starts to break down? The economy starts to resemble a single privately-owned resort that exists to make a profit, but all that money comes from the UBI of its customers, and UBI comes from taxes on the resort. It has the flavor of that old thought experiment where we imagine a world where one party can print unlimited money.
So, under radical AI, the only prediction I would make is that the laws of politics and economics must be radically different.
Yes, I agree that things would be quite different. Rather than 50% unemployment, we might first shift to a 20-hour work week. Real wages would be so high that 20 hours would be enough.
I’m unconvinced this will play out so well in practice. AI will decimate some sectors while leaving others completely untouched. A 20-hour work week is unlikely to help, in my opinion, although I’d love to be proven wrong. I suppose we will see where the great axe first falls and the general reaction.
Technological progress has always decimated some sectors more than others. I recall when an army of women worked as phone operators, jobs that were wiped out when electronic switching came on stream. The labor market can handle those things as long as new jobs open up.
Yeah, who knows.
A guess is AI will boost productivity, same way computers and then the internet did. But that's it. It will make life better, raise output.
The assimilation of AI into the economy will happen over decades.
I wonder if I have enough years left to see it play out.
Like you, I don't know anything. But I put the highest probability outcome (maybe a 40% chance?) that we have currently hit an AI plateau that will persist for our lifetimes. (I am 47.) Term papers (employing good grammar) are now much cheaper to produce, as are cute illustrations. But life will go on for the next generation or so of workers, pretty much as it has gone on for the previous three. Birth rates will continue to decline, but slowly, and migration will keep business as usual running in currently large cities in rich countries.
Yes, I think it's very possible that this is less transformative than many people (including me) assume. But I don't think "plateau" is the most likely outcome---more likely a slower rate of progress than the optimists assume. Hard to see how it doesn't gradually transform our economy, one robot at a time.
I would think it would be better to use a birth rate that uses a denominator that is the number of women of child-bearing age.
Agreed. But some of those changes in China are so dramatic, over such a short period of time, that they cannot possibly be explained by a changing number of women in that age group.
The scenario where robots do most everything seems to skip a big segment of the population - those the sell and distribute drugs and those that use them. If more activities are displaced by A{ automation will not a bigger percent of the population turn to drugs to give a portion of their lives meaning (if meaning can be found in the haze of drugs) and will not a bigger percent of the population seek innovation or artistic experiences through drugs. Where is the thinking on that segment of the population that is either currently left out or those that will be displaced?
And will AI invent safe new psychoactive drugs that boost utility?
Prof. Sumner, your techno-optimism is a breath of fresh air! Here's a second helping from the brilliant Jim Pethokoukis:
https://fasterplease.substack.com/p/the-formula-for-the-future
https://fasterplease.substack.com/p/a-new-american-golden-age