I've never written my thoughts on this so here's my shot (manual chain of thought lmao):
1) Superintelligence is coming lot faster than everyone thinks if you believe scaling loss, time horizon expansion, etc.
Slowly, and then "all at once", as they say ;)
2) In an all out AI-race between superpowers, say US and China, there's no way around concentrated power because without centralization, you move too slowly and lose the war. Napoleon, Churchill, Stalin, Caesar, Roosevelt (4 terms lol), arguably every force winning a war has been near dictatorial or at least oligopolistic for a reason.
3) When humans have a lot of unchecked power, they almost always abuse it. ESPECIALLY if it's a position that people rotate out of (ex. President of the US)
4) The incentive to treat your people well is because power is very decentralized in democratic / capitalist systems. We're going from the government being dependent on the people to do all the country's labor to the people being dependent on the government for resources.
the incentive gets dissolved pretty quickly!
5) The reason we stuff animals in cages, skin minks alive for fur, have large scale slaughterhouses, etc. is being they are dumb and have no economic value aside from being cooked and eaten.
Now obviously as humans we care about each other, but when any group has 0 leverage, dependent, and economically useless, they aren't treated as well as their baseline (before).
I'm not saying we will become the animals of today, but idk, all I'm saying is that certainly in countries with existing dictatorships like Russia and China, I really doubt that life will get better for people. USA and developing world is more of a wild card, the former because we have a decent democracy (but again, power dynamics are double edged sword), and developing world might just win out because tech + biology + medicine is really underdeveloped
The senerio that's scary is that once ASI is controlled by the USG they can abuse their power (POTUS) to turn into a dictatorship and the people wont be able to fight back against ASI
Really thoughtful chain of reasoning. I especially like the way you separate the technological trajectory from the political incentives that follow.
To me, though, it leads to one more quession.
It’s not only whether the technology becomes inevitable.
It’s whether democratic institutions mature before the technology reaches its economic elbow.
If concentrated power becomes economically attractive before institutions adapt, your concern becomes much more plausible. If institutions mature first, the same technology could produce a very different outcome.
Cleanest version of this argument I've read and the ”r” framing is right: the crossing is about productive hours and teleoperation, not sticker price. Which is exactly why I think the denominator deserves a closer look.
The human side carries the full social cost: wage, taxes, Medicare, social security. The robot side carries hardware, teleop, downtime.
What's missing are the shared systems it depends on: the grid it draws on, the rare-earth supply under every actuator, and the data centers the models run in. One side's externalities are in, the other's aren't, so ”r” is comparing two things measured differently.
And there's a second gap, maybe the more interesting one. ”r” can't see work no human is paid for today: clearing the beach, rebuilding the archive, tasks with real public value and no paying customer. No wage means no denominator, so the ratio goes blind exactly where the public value is highest.
That's not a flaw in the math. It's simply a reminder that every ratio measures only the world its denominator can see.
It is factually correct that temperatures exceeding 100°F have occurred across all states historically, with notable spikes during the early 20th century, particularly in the 1930s. However, focusing solely on whether such thresholds have been reached overlooks the more critical dimension: the frequency, duration, and geographic scale of heat events. Modern climate data indicates that heat waves today are not just isolated incidents but are increasingly prolonged, synchronized across regions, and accompanied by higher nighttime temperatures, amplifying their economic and public health impact. For decision-makers, the issue is not about validating or dismissing past extremes, but understanding how systemic risk is evolving. The strategic priority should therefore shift from debating historical comparisons to building resilience against intensifying heat patterns, investing in adaptive infrastructure, and ensuring energy systems can withstand sustained demand surges. In this context, the discussion is not about fear or narrative framing, but about forward-looking risk management in an environment of increasing climate volatility.
1. "An AI good enough to do AI research designs a better AI". What is "better"? Can we say that what is better for humans is better for AI agents? The main driver for human's innovations during the whole history is that fact that we are mortal, our time is limited and we are aware of it. That's why all our innovations are targeted to improving our life and life of our successors.
What is "better" for AI driven research? Will AI have a willingness to improve in any sense at all? Nobody knows...
A part of me finds this all funny... everyone's thinking about making money but we're inventing the terminator in real time lol
Actually been getting more optimistic on humanity over time. Ironically enough
I've never written my thoughts on this so here's my shot (manual chain of thought lmao):
1) Superintelligence is coming lot faster than everyone thinks if you believe scaling loss, time horizon expansion, etc.
Slowly, and then "all at once", as they say ;)
2) In an all out AI-race between superpowers, say US and China, there's no way around concentrated power because without centralization, you move too slowly and lose the war. Napoleon, Churchill, Stalin, Caesar, Roosevelt (4 terms lol), arguably every force winning a war has been near dictatorial or at least oligopolistic for a reason.
3) When humans have a lot of unchecked power, they almost always abuse it. ESPECIALLY if it's a position that people rotate out of (ex. President of the US)
4) The incentive to treat your people well is because power is very decentralized in democratic / capitalist systems. We're going from the government being dependent on the people to do all the country's labor to the people being dependent on the government for resources.
the incentive gets dissolved pretty quickly!
5) The reason we stuff animals in cages, skin minks alive for fur, have large scale slaughterhouses, etc. is being they are dumb and have no economic value aside from being cooked and eaten.
Now obviously as humans we care about each other, but when any group has 0 leverage, dependent, and economically useless, they aren't treated as well as their baseline (before).
I'm not saying we will become the animals of today, but idk, all I'm saying is that certainly in countries with existing dictatorships like Russia and China, I really doubt that life will get better for people. USA and developing world is more of a wild card, the former because we have a decent democracy (but again, power dynamics are double edged sword), and developing world might just win out because tech + biology + medicine is really underdeveloped
The senerio that's scary is that once ASI is controlled by the USG they can abuse their power (POTUS) to turn into a dictatorship and the people wont be able to fight back against ASI
Bingo
Really thoughtful chain of reasoning. I especially like the way you separate the technological trajectory from the political incentives that follow.
To me, though, it leads to one more quession.
It’s not only whether the technology becomes inevitable.
It’s whether democratic institutions mature before the technology reaches its economic elbow.
If concentrated power becomes economically attractive before institutions adapt, your concern becomes much more plausible. If institutions mature first, the same technology could produce a very different outcome.
Cleanest version of this argument I've read and the ”r” framing is right: the crossing is about productive hours and teleoperation, not sticker price. Which is exactly why I think the denominator deserves a closer look.
The human side carries the full social cost: wage, taxes, Medicare, social security. The robot side carries hardware, teleop, downtime.
What's missing are the shared systems it depends on: the grid it draws on, the rare-earth supply under every actuator, and the data centers the models run in. One side's externalities are in, the other's aren't, so ”r” is comparing two things measured differently.
And there's a second gap, maybe the more interesting one. ”r” can't see work no human is paid for today: clearing the beach, rebuilding the archive, tasks with real public value and no paying customer. No wage means no denominator, so the ratio goes blind exactly where the public value is highest.
That's not a flaw in the math. It's simply a reminder that every ratio measures only the world its denominator can see.
Excellent article. Never think about the analogy of digital AI and physical AI before. Always thought digital AI enabling on physical AI. Insightful
It is factually correct that temperatures exceeding 100°F have occurred across all states historically, with notable spikes during the early 20th century, particularly in the 1930s. However, focusing solely on whether such thresholds have been reached overlooks the more critical dimension: the frequency, duration, and geographic scale of heat events. Modern climate data indicates that heat waves today are not just isolated incidents but are increasingly prolonged, synchronized across regions, and accompanied by higher nighttime temperatures, amplifying their economic and public health impact. For decision-makers, the issue is not about validating or dismissing past extremes, but understanding how systemic risk is evolving. The strategic priority should therefore shift from debating historical comparisons to building resilience against intensifying heat patterns, investing in adaptive infrastructure, and ensuring energy systems can withstand sustained demand surges. In this context, the discussion is not about fear or narrative framing, but about forward-looking risk management in an environment of increasing climate volatility.
Excellent article.
Two remarks:
1. "An AI good enough to do AI research designs a better AI". What is "better"? Can we say that what is better for humans is better for AI agents? The main driver for human's innovations during the whole history is that fact that we are mortal, our time is limited and we are aware of it. That's why all our innovations are targeted to improving our life and life of our successors.
What is "better" for AI driven research? Will AI have a willingness to improve in any sense at all? Nobody knows...
2. They. All. Will. Need. Memory.