If they can't understand AI, how good do you think the legislation is going to be?
Our president, most of the members of our Senate and House of Representatives, and our Supreme Court, are basically old people with zero background even in technology, let alone AI specifically.
Why does anyone think there will be good legislation on AI coming from such people?
Sorry, Andy--I agree more with the original. Two major issues with your edits:
1. major improvements in living standards--but for whom? Elon, Sam, Dario, etc? The employees of Anthropic and OpenAI who are driving SF housing prices to insane levels? I think we will see even more amazing levels of inequality driven by AI than we have now.
2. capabilities to respond quickly to AI challenges? Remember, this is the US Congress and President we are talking about. They still haven't done anything about the harms to kids from social media, and that's been going on for well over a decade. If we wait--even if the ultimate outcomes are unclear now--we won't get any action at all, at least at the federal level.
Both your assumptions are just assumptions, and pure ideological scare-mongering at that.
1. The only reason to frame "improvements in living standards" as only applying to a few people is to scare the populace. H.L. Mencken was writing about people like you.
2. There has been lots of scare-mongering over children and social media, but no concrete evidence, and the sloppy evidence which does exists goes both ways.
"Levels of inequality" don't really bother me. For instance, Bill Gates is many tens of thousands of times more wealthy than I am. But I'm comfortably retired, so that doesn't bother me at all.
I think the much more important issue is the income/wealth/well-being of the bottom, say 20 percent, of the population. Will that actually go down? Or will it go up, but just not as much as we'd like?
The original is the precautionary principle, and it's everything that's wrong with preemptive regulation and governance in general.
Your version is an improvement only in what it would have government do. Better yet is for government to just butt out until actual problems show up and are not corrected by markets.
Government regulation locks in yesterday's status quo, and robs society of the flexibility which had been behind all progress since Ooogh dared to play with fire.
Your edit resonates with me, Andrew. Reading the original statement, I had a sense of déjà vu: around 2000, as the internet spread like wildfire, we heard nearly identical warnings about disruption, displacement, and the urgent need to steer. The threats and the scale rhyme. What is genuinely different this time is the pace at which governments, industries and people must adapt. That argues for building capabilities to respond, as you put it, not for governing upstream what we do not yet understand.
I hear you. When asked to sign (or not!) yesterday, I wrote back with quite a similar edit. It was frozen, given that many are much more glass-half-empty. I thought that since the words opportunity and transformation were in there quite explicitly, that I could sign and maintain my focus on progress and upside without needless harm. I'd sign yours, with the exception that the past is no predictor of future success, specifically (as Tom says here as well) wrt distribution and pace of change. Negative externalities could accrue far too quickly, and in such cases discontinuity might be a fancy word for war.
I agree with the adjustment and think in one sense it’s an improvement, but I do think that (unfortunately) because the original has a directionality and sort of moral valence, it reads as more concrete, whereas your less-negative edit feels less so. I don’t pretend to know exactly how to land on more solid “what to do about it” language, but that’s my initial thought!
But economists should not be leaning on moral valence and false clarity. They should be relying on facts and probabilities. What you’re saying unfortunately is that political scaremongering is more effective than rational statements.
This is great to read - nothing about our AI future is known for sure. Acting now to guard against risks that likely will never happen risks causing huge damage.
We need economic progress to solve the world's problems and tech progress is an essential part of that - that applies to Africans living on less than $2 a day as much as it does to the rich world middle classes. We don't get that unless we embrace new technology ideas and use them the right way.
And worse, if there are major downsides to AI, Governments will fail to act because they are too busy fighting ghosts. The lessons of past tech disruption show we have failed to manage the transition. Economic progress has been good but too many groups have been left behind. Learn and understand how the change is evolving and help those who need help when we can see where the real problems arise.
AI is just a technology. It does not create or contribute to growing income inequality, which is a consequence of how societies are organized. The impact of AI in the U.S. will be different from the impact in Sweden, China, or Iran.
To manage societal disruptions you need to steer not the AI - you need to steer societies. But isn’t that what democracies are supposed to be doing already? And how does “steering” AI in a handful of countries help if the rest of the world marches on with no such steering? Will the steering committee members advocate forbidding citizens of their countries from accessing models developed in the un-steered world?
Bottom line: you see a problem in a society - fix the society. Don’t blame technology, however powerful that technology is or might be.
I think your question about "steering" in a "handful of societies" is hugely important. If AI is "steered" in the US, but not in China, I think that will mean that China's AI will become much more powerful than the AI developed in the US.
I think that would be a far worse outcome than if US AI is not "steered".
I agree more with your draft than the original, but think it can be further improved.
2. Like previous world-changing technologies, AI will drive economic growth, create new industries, and raise living standards. But it also brings new risks and harms, and we must expect that the benefits and disruptions will come quicker and more unevenly than in the past.
3. So economists, policymakers, and technology leaders must act now to ensure that the benefits of transformative AI are shared as widely as possible among all Americans, even as we build capabilities to understand AI and ameliorate or avoid rapid economic losses and social dislocations. We have learned much over the last quarter century that applies to AI as well, including the importance of fostering technological dynamism and resilience, preserving communities, protecting against extraordinary concentrations of economic and political power, and defending human agency, as well as the bodies and minds of our youth.
I agree your wording is better but I am left puzzled by what particular actions you would like to see flow from "3. So economists, policymakers and technology leaders must act now to understand the economics of transformative AI, and to build the capabilities needed to respond quickly and effectively to the challenges it will bring."
What would you like to see happen here?
How can they demonstrate their understanding?
If this is a rapidly developing general purpose technology what can they do to maintain and understand of what is a rapidly moving target as well as the secondary and tertiary effects as it diffuses into the economy?
YES! This is an improvement. The Chicken Little perspective is problematic.
I do think we have environmental impacts to consider. Robbing small towns of clean water and driving up the cost of electricity is something to be managed without trampling on humans.
Upstream governance may freeze the wrong future.
But downstream response assumes institutions can still catch a system moving faster than they can understand or legislate.
That may be the weaker assumption.
If they can't understand AI, how good do you think the legislation is going to be?
Our president, most of the members of our Senate and House of Representatives, and our Supreme Court, are basically old people with zero background even in technology, let alone AI specifically.
Why does anyone think there will be good legislation on AI coming from such people?
Sorry, Andy--I agree more with the original. Two major issues with your edits:
1. major improvements in living standards--but for whom? Elon, Sam, Dario, etc? The employees of Anthropic and OpenAI who are driving SF housing prices to insane levels? I think we will see even more amazing levels of inequality driven by AI than we have now.
2. capabilities to respond quickly to AI challenges? Remember, this is the US Congress and President we are talking about. They still haven't done anything about the harms to kids from social media, and that's been going on for well over a decade. If we wait--even if the ultimate outcomes are unclear now--we won't get any action at all, at least at the federal level.
Both your assumptions are just assumptions, and pure ideological scare-mongering at that.
1. The only reason to frame "improvements in living standards" as only applying to a few people is to scare the populace. H.L. Mencken was writing about people like you.
2. There has been lots of scare-mongering over children and social media, but no concrete evidence, and the sloppy evidence which does exists goes both ways.
"Levels of inequality" don't really bother me. For instance, Bill Gates is many tens of thousands of times more wealthy than I am. But I'm comfortably retired, so that doesn't bother me at all.
I think the much more important issue is the income/wealth/well-being of the bottom, say 20 percent, of the population. Will that actually go down? Or will it go up, but just not as much as we'd like?
Thomas Sowell might share your reaction to „steer AI“, pointing out (elegantly) the inevitable lack of competence or those willing to steer.
'willing' to steer? No, DESIRING to steer! And that is the problem.
The original is the precautionary principle, and it's everything that's wrong with preemptive regulation and governance in general.
Your version is an improvement only in what it would have government do. Better yet is for government to just butt out until actual problems show up and are not corrected by markets.
Government regulation locks in yesterday's status quo, and robs society of the flexibility which had been behind all progress since Ooogh dared to play with fire.
Your edit resonates with me, Andrew. Reading the original statement, I had a sense of déjà vu: around 2000, as the internet spread like wildfire, we heard nearly identical warnings about disruption, displacement, and the urgent need to steer. The threats and the scale rhyme. What is genuinely different this time is the pace at which governments, industries and people must adapt. That argues for building capabilities to respond, as you put it, not for governing upstream what we do not yet understand.
I hear you. When asked to sign (or not!) yesterday, I wrote back with quite a similar edit. It was frozen, given that many are much more glass-half-empty. I thought that since the words opportunity and transformation were in there quite explicitly, that I could sign and maintain my focus on progress and upside without needless harm. I'd sign yours, with the exception that the past is no predictor of future success, specifically (as Tom says here as well) wrt distribution and pace of change. Negative externalities could accrue far too quickly, and in such cases discontinuity might be a fancy word for war.
I agree with the adjustment and think in one sense it’s an improvement, but I do think that (unfortunately) because the original has a directionality and sort of moral valence, it reads as more concrete, whereas your less-negative edit feels less so. I don’t pretend to know exactly how to land on more solid “what to do about it” language, but that’s my initial thought!
But economists should not be leaning on moral valence and false clarity. They should be relying on facts and probabilities. What you’re saying unfortunately is that political scaremongering is more effective than rational statements.
And I definitely don’t mean to say that the moral valence OUGHT to be leveraged, only that, as you say, it is unfortunately more innately compelling.
This is great to read - nothing about our AI future is known for sure. Acting now to guard against risks that likely will never happen risks causing huge damage.
We need economic progress to solve the world's problems and tech progress is an essential part of that - that applies to Africans living on less than $2 a day as much as it does to the rich world middle classes. We don't get that unless we embrace new technology ideas and use them the right way.
And worse, if there are major downsides to AI, Governments will fail to act because they are too busy fighting ghosts. The lessons of past tech disruption show we have failed to manage the transition. Economic progress has been good but too many groups have been left behind. Learn and understand how the change is evolving and help those who need help when we can see where the real problems arise.
Well said.
Both statements are *extremely* vague. Why bother with them?
AI is just a technology. It does not create or contribute to growing income inequality, which is a consequence of how societies are organized. The impact of AI in the U.S. will be different from the impact in Sweden, China, or Iran.
To manage societal disruptions you need to steer not the AI - you need to steer societies. But isn’t that what democracies are supposed to be doing already? And how does “steering” AI in a handful of countries help if the rest of the world marches on with no such steering? Will the steering committee members advocate forbidding citizens of their countries from accessing models developed in the un-steered world?
Bottom line: you see a problem in a society - fix the society. Don’t blame technology, however powerful that technology is or might be.
I think your question about "steering" in a "handful of societies" is hugely important. If AI is "steered" in the US, but not in China, I think that will mean that China's AI will become much more powerful than the AI developed in the US.
I think that would be a far worse outcome than if US AI is not "steered".
I far prefer this version
I agree more with your draft than the original, but think it can be further improved.
2. Like previous world-changing technologies, AI will drive economic growth, create new industries, and raise living standards. But it also brings new risks and harms, and we must expect that the benefits and disruptions will come quicker and more unevenly than in the past.
3. So economists, policymakers, and technology leaders must act now to ensure that the benefits of transformative AI are shared as widely as possible among all Americans, even as we build capabilities to understand AI and ameliorate or avoid rapid economic losses and social dislocations. We have learned much over the last quarter century that applies to AI as well, including the importance of fostering technological dynamism and resilience, preserving communities, protecting against extraordinary concentrations of economic and political power, and defending human agency, as well as the bodies and minds of our youth.
I agree your wording is better but I am left puzzled by what particular actions you would like to see flow from "3. So economists, policymakers and technology leaders must act now to understand the economics of transformative AI, and to build the capabilities needed to respond quickly and effectively to the challenges it will bring."
What would you like to see happen here?
How can they demonstrate their understanding?
If this is a rapidly developing general purpose technology what can they do to maintain and understand of what is a rapidly moving target as well as the secondary and tertiary effects as it diffuses into the economy?
This statement feels incomplete.
The certainty you injected in the first two points does not marry best with a third point calling for flexibility on the basis of… uncertainty.
Definitely rather sign your #3 than the original.
YES! This is an improvement. The Chicken Little perspective is problematic.
I do think we have environmental impacts to consider. Robbing small towns of clean water and driving up the cost of electricity is something to be managed without trampling on humans.