I think this is partly cope. Middle management at large companies might have a significant political component but large companies still have much higher labor productivity than small ones, they still represent like a quarter of OECD economic output, and the average middle manager is probably still creating very significant amounts of marginal value despite the political infighting.
Yes, there might be very few middle managers at the top of Forbes’ list. But look at millionaires, the people at the top 10% of the wealth distribution in the US. Most middle managers will probably be there, along with other highly paid professionals. And if you found a startup and it fails, you won’t make the 10%, which is what happens in the overwhelming majority of cases.
So purely in terms of wealth and creating value for society, marginal improvements in middle management seem quite valuable. Sure, being a founder might have much higher EV, but also vastly higher variance. And risk aversion is behaviorally indistinguishable from just having a different utility function.
Speaking of which, you might want to be immortal and go to the moon, but most people don’t. You could argue that if they’d read the right books or had the right parent/teacher/friend or had more vision, they would also want that, but at that point you’re just saying that everyone should be playing the game you like, instead of the one they like.
And I dispute the idea that there’s less politics, signaling or strategic/conflictive behavior at the “winner’s bracket”. Look at Sam Altman, look at the status competitions and fighting for credit in the highest halls of science. Look at states for God’s sake. Do you really think Vladimir Putin or Donald Trump or Xi Jinping are not in the winner’s bracket, that they will have less real power than the first person to figure out how to solve aging?
Physicists might have found the secret knowledge of how to create nuclear weapons that nobody else had, but after they had the bright idea the bottlenecks were capital, labor, natural resources, and the ability to manage their combination efficiently, and the physicists were not the ones who got to control the weapons in the end.
I think something like your thesis might be true in terms of actually having good counterfactual impact on reality vs merely capturing the resulting wealth, power and prestige, what you call leading the parade. But that doesn’t mean that you get to both have the impact and lead the parade by pursuing just the impact part!
“Do you really think Vladimir Putin or Donald Trump or Xi Jinping are not in the winner’s bracket, that they will have less real power than the first person to figure out how to solve aging?” I interpret this as a rhetorical question, intended to suggest that its answer is obviously ‘yes’.
I dispute this, because it’s extremely difficult to compare different kinds of power, especially when they take effect over different timescales. The first person to solve aging would indeed wield a colossal amount of power as a result, because they would hasten the advent of a technology which allowed people to live arbitrarily long, allowing more people to do so. Even doing this by a day could create an arbitrary number of human life-years. Similarly, I think it’s quite likely that Demis Hassabis, for example, is more powerful than Vladimir Putin, Xi Jinping or Donald Trump, because none of them would have been able to create an artificial superintelligence as early as he has facilitated its creation if he never existed; by realizing that AGI was plausible in a much shorter time than others believed, he did something( along with the rest of DeepMind, but he clearly played a pivotal role) which would be simply impossible for anyone lacking this insight. Information is power, but not all information is equally power dense, and the most powerful information often takes the form of deep insights which can’t be extracted from large crowds of people using any amount of status.[1]
While waiting to be able to post again, I anticipated a response of the form ” Many other people had the same or a similar insight, and didn’t become as powerful as Demis Hassabis, therefore that kind of insight doesn’t confer power. ” to which my response would be that it’s true that other factors needed to be present in order for someone like this to make use of the insight, but this doesn’t make it any less important; those other factors would be far less useful without the insight. Something which facilitates power is itself power. Another response might be to point out that leaders such as Donald Trump or Xi Jinping have the ability to coordinate to prevent an ASI from being created, or accelerate its creation. I’m not sure that the former is true, but if it is, it’s only true in the sense that they have the official capacity to coordinate—whether they truly can coordinate without wanting to depends on the definition of free will and whether they have it. While they might be able to accelerate the development of AI, it seems likely to me that it’s not possible to do so by an amount comparable to the effect of starting early.
I think the good impact vs capturing the results is a really good focus, especially as it relates to the conundrum of the physicists not controlling the weapons they create. I think that generalizes to the problem of knowing which things will go on to be beneficial vs harmful and knowing how to create technology and systems that are robust to downstream tampering that would affect how beneficial vs harmful they are.
As an example, I feel we should have identified the internet and social media as being like public infrastructure and taken steps to make sure they were produced focusing on public good rather than profit. I think many social media companies were founded with public good and pro-social, pro-communication goals, but then once they got bought by profit seeking entities they began to include more an more harmful user lock in, perverse marketing and other similar kinds of issues.
I think this is partly cope. Middle management at large companies might have a significant political component but large companies still have much higher labor productivity than small ones, they still represent like a quarter of OECD economic output, and the average middle manager is probably still creating very significant amounts of marginal value despite the political infighting.
Yes, there might be very few middle managers at the top of Forbes’ list. But look at millionaires, the people at the top 10% of the wealth distribution in the US. Most middle managers will probably be there, along with other highly paid professionals. And if you found a startup and it fails, you won’t make the 10%, which is what happens in the overwhelming majority of cases.
So purely in terms of wealth and creating value for society, marginal improvements in middle management seem quite valuable. Sure, being a founder might have much higher EV, but also vastly higher variance. And risk aversion is behaviorally indistinguishable from just having a different utility function.
Speaking of which, you might want to be immortal and go to the moon, but most people don’t. You could argue that if they’d read the right books or had the right parent/teacher/friend or had more vision, they would also want that, but at that point you’re just saying that everyone should be playing the game you like, instead of the one they like.
And I dispute the idea that there’s less politics, signaling or strategic/conflictive behavior at the “winner’s bracket”. Look at Sam Altman, look at the status competitions and fighting for credit in the highest halls of science. Look at states for God’s sake. Do you really think Vladimir Putin or Donald Trump or Xi Jinping are not in the winner’s bracket, that they will have less real power than the first person to figure out how to solve aging?
Physicists might have found the secret knowledge of how to create nuclear weapons that nobody else had, but after they had the bright idea the bottlenecks were capital, labor, natural resources, and the ability to manage their combination efficiently, and the physicists were not the ones who got to control the weapons in the end.
I think something like your thesis might be true in terms of actually having good counterfactual impact on reality vs merely capturing the resulting wealth, power and prestige, what you call leading the parade. But that doesn’t mean that you get to both have the impact and lead the parade by pursuing just the impact part!
“Do you really think Vladimir Putin or Donald Trump or Xi Jinping are not in the winner’s bracket, that they will have less real power than the first person to figure out how to solve aging?” I interpret this as a rhetorical question, intended to suggest that its answer is obviously ‘yes’.
I dispute this, because it’s extremely difficult to compare different kinds of power, especially when they take effect over different timescales. The first person to solve aging would indeed wield a colossal amount of power as a result, because they would hasten the advent of a technology which allowed people to live arbitrarily long, allowing more people to do so. Even doing this by a day could create an arbitrary number of human life-years. Similarly, I think it’s quite likely that Demis Hassabis, for example, is more powerful than Vladimir Putin, Xi Jinping or Donald Trump, because none of them would have been able to create an artificial superintelligence as early as he has facilitated its creation if he never existed; by realizing that AGI was plausible in a much shorter time than others believed, he did something( along with the rest of DeepMind, but he clearly played a pivotal role) which would be simply impossible for anyone lacking this insight. Information is power, but not all information is equally power dense, and the most powerful information often takes the form of deep insights which can’t be extracted from large crowds of people using any amount of status.[1]
While waiting to be able to post again, I anticipated a response of the form ” Many other people had the same or a similar insight, and didn’t become as powerful as Demis Hassabis, therefore that kind of insight doesn’t confer power. ” to which my response would be that it’s true that other factors needed to be present in order for someone like this to make use of the insight, but this doesn’t make it any less important; those other factors would be far less useful without the insight. Something which facilitates power is itself power. Another response might be to point out that leaders such as Donald Trump or Xi Jinping have the ability to coordinate to prevent an ASI from being created, or accelerate its creation. I’m not sure that the former is true, but if it is, it’s only true in the sense that they have the official capacity to coordinate—whether they truly can coordinate without wanting to depends on the definition of free will and whether they have it. While they might be able to accelerate the development of AI, it seems likely to me that it’s not possible to do so by an amount comparable to the effect of starting early.
I think the good impact vs capturing the results is a really good focus, especially as it relates to the conundrum of the physicists not controlling the weapons they create. I think that generalizes to the problem of knowing which things will go on to be beneficial vs harmful and knowing how to create technology and systems that are robust to downstream tampering that would affect how beneficial vs harmful they are.
As an example, I feel we should have identified the internet and social media as being like public infrastructure and taken steps to make sure they were produced focusing on public good rather than profit. I think many social media companies were founded with public good and pro-social, pro-communication goals, but then once they got bought by profit seeking entities they began to include more an more harmful user lock in, perverse marketing and other similar kinds of issues.