Good discussion. Couple of things. It would be a mistake to attribute certain things to a broad strategy decision in Beijing to do something with respect to AI targeted at the US. Open source/weight models come to mind here. The move to open source was not hatched in Zhongnanhai but rather an organic bottom up decision, led by DeepSeek, then quickly adopted by most but not all of the other leading Chinese AI labs, that has sought to overcome the disadvantages that are inherent in China due to lack of unlimited access to compute, and a firm belief within some, but not all quarters of the China AI safety/security community that open source/weight models are the way to go. As I have said many times, the bureaucrats in Beijing were just as surprised as the tech bros in the Valley and investors on Wall Street with the success of DeepSeek. Also, thanks for validating what I have been saying, almost almost alone, that China is clearly not "AGI-pilled", and I have spent a lot of time in China looking for such a pill. I am less sure of the assertion that we must swallow the bitter lesson that we are locked into a structural conflict. As you note, this meme is frankly all based on questionable assumptions about the "end state", so there should be more room for development of collaboration. As you acknowledge, and I wrote about in great detail in a recent Cairo Review article, one major problem is Dario's assertion that "democractic AI" needs to win so that we can force regime change in China. This is clearly not only a contesable presumption, but also a very very dangerous one......
Great comments. As you mentioned, your voice sounds lonely in this environment where almost , almost everyone is high in the party and cheering on a fight. Somehow, the industry and government are self-hyping themselves into a good-vs-evil final battle. Private companies have real incentives to do this because trillions of dollars are flowing into AI investments/gamble. The government seems to be dupped or kidnapped. People often credit the Star War arm race accelerated the collapse of the Soviet Union. Could this AI buildup have the same effect on the US. We pour so much money in this while ignore the so much social and economic problems of our society, no guaranteed basic healthcare, drug addiction, poor education, etc. The contrast reflects with our ethics and morals (Also,our military also are killing people,maybe drug dealers, at the open sea.) If we have reasonable self-awarenes, are we trusting ourselves or think that other people should trust us to have the ULTIMATE power?
This is a nitpick to your main point, but I think the US is still a long way from being fully AGI pilled e.g. growth expectations and interest rates have hardly risen; and it would be possible to scale up compute a lot faster (e.g. most leading TSM nodes are still going to mobile phones).
The build out pace of nuclear reactors would cement the idea. Nullified environmental review and removal of other red tape would prove to me that the U.S. understands the bottlenecks.
Nuclear's basically a distraction from solar at this point, even if nuclear reactors could be deployed in five years, that would still be after AGI would be developed.
If this is THE definition time of humankind, the US,China and the rest of the world should work together to make sure that AI is not going evil. If we don't go evil, it doesn't matter who 'arrives' first. The distrust between the US and China could push each side to be evil because we assume each other are thinking evil. We don't know if this AI will be the 'end of history'. If it is not, the old rule will still apply: No country can be number one forever. We use your life experience to extraploit but our life time is too short in the the scale of history. There is not enough energy spent on that AI should benefit whole mankind. As you mentioned, the US AI lead by the private companies which are naturally profit-driven and competition-driven(just to win). There is some paradox in this: The "selfish" companies are working in something that could liberate us all or end us all. We have believed the invisible hand for so long. Are we willing to believe that the invisible hand that has been driven by self-interest will liberate us ALL and FOREVER THIS time?
Fabulous summary and insights from Dean Ball on AI's near future and the US/China AI race.
“unbounded, multi-dimensional, technological, scientific, and economic competition.”
"I find it intriguing that both countries seem to have converged on the strategies that best suit their respective strengths. Advanced AI is, at its core, software-as-a-service delivered through high-end semiconductors, cloud computing platforms, charismatic user interfaces, and enabled by clever financial and legal engineering. Every one of those things is America’s civilizational bread and butter. Embodied AI is, at its core, enabled by mass manufacturing excellence, thick trade networks, and other characteristics that fundamentally tilt in China’s advantage."
"The sad reality is that the current strategies of China and the U.S. are complementary."
For other perspectives and 'AI Future' summaries, check out Gary Marcus, Colin Lewis, Michael Spencer, and Luiza Jarovsky substacks.
I read the situation a bit differently. I think the Chinese leaders have not decided to support AI fully, like other hard techs they are supporting e.g. Robotics, Aviations, Spaces, Pharmaceuticals, Semiconductors, etc. They simply encourage, help remove roadblocks and nudge their tech big boys to support the open source direction to bring the cost down, enable AI to compliment every other techs and destroy the margins of American AI labs - so the AI bubbles will pop some days. This doesn’t cost them much money, and they must be happy seeing Chinese models almost catching up to the west in a very short time.
In a way, this is a much easier techs for them to advance compared to high-end semiconductors. I don’t think they are worried about AI race at all.
I believe that AGI will be achieved, if we could, by smart logics and algorithms, not vast data, mass compute nor brute forces. And that could come from anywhere, including China. In fact, their cheap open source models may increase probability that it will happen there because many more people could try more new ideas in such environment.
I also think talent issue in this area are also overrated. With every million people, you will have more than a handful of geniuses. So, China has enough geniuses within their own to advance any techs they want to pursue, as long as they are genuinely trying and the process not get bog down by R&D politics and funding.
100% agree that the strategies are complementary - I argued in an essay 2 weeks ago that the differences stem from different mindsets, constraints & incentives but that the paths will eventualy converge: https://rajeshachanta.substack.com/p/spectacle-vs-scaffolding
The most valuable remaining factor for the US strategy is their 'distribution network': American companies have very much more robust presence 'on the ground' in almost every country than Chinese ones.
And lets not be candid: the US govt is quite pushy for their companies abroad, much more than China's govt.
Thats the real remaining edge that makes the US still win most procurement contracts, investement commitments, etc. And its not even mentioned in an otherwise very good essay.
Thanks for the post. It’s good to hear that folks in the current administration are considering the implications. I can heartily recommend Chris Miller’s Chip War as a way to think about the current AI competition (to borrow a line; quantity has a quality all its own). Much like the Cold War our goal should be to keep things from going kinetic in a big way and let our natural advantages slowly tip the scales. See the discussion by S. Paine in Foreign Affairs: https://www.foreignaffairs.com/authors/s-c-m-paine
That ultimately, we’ll need to contain China from building uncontrolled superintelligence, even if we manage to get there first, and so we’d do well to start investing more in control and verification in the meantime
AI Race strikes me just like Space Race and Arms Race - getting farther out faster and getting and maintaining an edge on the unknowably extensive frontier without any bounds on timeline or technological development.
Good discussion. Couple of things. It would be a mistake to attribute certain things to a broad strategy decision in Beijing to do something with respect to AI targeted at the US. Open source/weight models come to mind here. The move to open source was not hatched in Zhongnanhai but rather an organic bottom up decision, led by DeepSeek, then quickly adopted by most but not all of the other leading Chinese AI labs, that has sought to overcome the disadvantages that are inherent in China due to lack of unlimited access to compute, and a firm belief within some, but not all quarters of the China AI safety/security community that open source/weight models are the way to go. As I have said many times, the bureaucrats in Beijing were just as surprised as the tech bros in the Valley and investors on Wall Street with the success of DeepSeek. Also, thanks for validating what I have been saying, almost almost alone, that China is clearly not "AGI-pilled", and I have spent a lot of time in China looking for such a pill. I am less sure of the assertion that we must swallow the bitter lesson that we are locked into a structural conflict. As you note, this meme is frankly all based on questionable assumptions about the "end state", so there should be more room for development of collaboration. As you acknowledge, and I wrote about in great detail in a recent Cairo Review article, one major problem is Dario's assertion that "democractic AI" needs to win so that we can force regime change in China. This is clearly not only a contesable presumption, but also a very very dangerous one......
Great comments. As you mentioned, your voice sounds lonely in this environment where almost , almost everyone is high in the party and cheering on a fight. Somehow, the industry and government are self-hyping themselves into a good-vs-evil final battle. Private companies have real incentives to do this because trillions of dollars are flowing into AI investments/gamble. The government seems to be dupped or kidnapped. People often credit the Star War arm race accelerated the collapse of the Soviet Union. Could this AI buildup have the same effect on the US. We pour so much money in this while ignore the so much social and economic problems of our society, no guaranteed basic healthcare, drug addiction, poor education, etc. The contrast reflects with our ethics and morals (Also,our military also are killing people,maybe drug dealers, at the open sea.) If we have reasonable self-awarenes, are we trusting ourselves or think that other people should trust us to have the ULTIMATE power?
This is a nitpick to your main point, but I think the US is still a long way from being fully AGI pilled e.g. growth expectations and interest rates have hardly risen; and it would be possible to scale up compute a lot faster (e.g. most leading TSM nodes are still going to mobile phones).
The build out pace of nuclear reactors would cement the idea. Nullified environmental review and removal of other red tape would prove to me that the U.S. understands the bottlenecks.
Nuclear's basically a distraction from solar at this point, even if nuclear reactors could be deployed in five years, that would still be after AGI would be developed.
If this is THE definition time of humankind, the US,China and the rest of the world should work together to make sure that AI is not going evil. If we don't go evil, it doesn't matter who 'arrives' first. The distrust between the US and China could push each side to be evil because we assume each other are thinking evil. We don't know if this AI will be the 'end of history'. If it is not, the old rule will still apply: No country can be number one forever. We use your life experience to extraploit but our life time is too short in the the scale of history. There is not enough energy spent on that AI should benefit whole mankind. As you mentioned, the US AI lead by the private companies which are naturally profit-driven and competition-driven(just to win). There is some paradox in this: The "selfish" companies are working in something that could liberate us all or end us all. We have believed the invisible hand for so long. Are we willing to believe that the invisible hand that has been driven by self-interest will liberate us ALL and FOREVER THIS time?
Insightful. To think we worry about the profit of it, when it poses an existential threat to humanity just boggles the mind.
Fabulous summary and insights from Dean Ball on AI's near future and the US/China AI race.
“unbounded, multi-dimensional, technological, scientific, and economic competition.”
"I find it intriguing that both countries seem to have converged on the strategies that best suit their respective strengths. Advanced AI is, at its core, software-as-a-service delivered through high-end semiconductors, cloud computing platforms, charismatic user interfaces, and enabled by clever financial and legal engineering. Every one of those things is America’s civilizational bread and butter. Embodied AI is, at its core, enabled by mass manufacturing excellence, thick trade networks, and other characteristics that fundamentally tilt in China’s advantage."
"The sad reality is that the current strategies of China and the U.S. are complementary."
For other perspectives and 'AI Future' summaries, check out Gary Marcus, Colin Lewis, Michael Spencer, and Luiza Jarovsky substacks.
I read the situation a bit differently. I think the Chinese leaders have not decided to support AI fully, like other hard techs they are supporting e.g. Robotics, Aviations, Spaces, Pharmaceuticals, Semiconductors, etc. They simply encourage, help remove roadblocks and nudge their tech big boys to support the open source direction to bring the cost down, enable AI to compliment every other techs and destroy the margins of American AI labs - so the AI bubbles will pop some days. This doesn’t cost them much money, and they must be happy seeing Chinese models almost catching up to the west in a very short time.
In a way, this is a much easier techs for them to advance compared to high-end semiconductors. I don’t think they are worried about AI race at all.
I believe that AGI will be achieved, if we could, by smart logics and algorithms, not vast data, mass compute nor brute forces. And that could come from anywhere, including China. In fact, their cheap open source models may increase probability that it will happen there because many more people could try more new ideas in such environment.
I also think talent issue in this area are also overrated. With every million people, you will have more than a handful of geniuses. So, China has enough geniuses within their own to advance any techs they want to pursue, as long as they are genuinely trying and the process not get bog down by R&D politics and funding.
100% agree that the strategies are complementary - I argued in an essay 2 weeks ago that the differences stem from different mindsets, constraints & incentives but that the paths will eventualy converge: https://rajeshachanta.substack.com/p/spectacle-vs-scaffolding
The most valuable remaining factor for the US strategy is their 'distribution network': American companies have very much more robust presence 'on the ground' in almost every country than Chinese ones.
And lets not be candid: the US govt is quite pushy for their companies abroad, much more than China's govt.
Thats the real remaining edge that makes the US still win most procurement contracts, investement commitments, etc. And its not even mentioned in an otherwise very good essay.
Thanks for the post. It’s good to hear that folks in the current administration are considering the implications. I can heartily recommend Chris Miller’s Chip War as a way to think about the current AI competition (to borrow a line; quantity has a quality all its own). Much like the Cold War our goal should be to keep things from going kinetic in a big way and let our natural advantages slowly tip the scales. See the discussion by S. Paine in Foreign Affairs: https://www.foreignaffairs.com/authors/s-c-m-paine
Thank you for saying this! Very much the competition is not a race, and I’m glad to hear that you’ve talked through those dynamics with relevant folks
My own take on the dynamic is here - https://open.substack.com/pub/stevenadler/p/contain-and-verify-the-endgame-of?r=4qacg&utm_medium=ios
That ultimately, we’ll need to contain China from building uncontrolled superintelligence, even if we manage to get there first, and so we’d do well to start investing more in control and verification in the meantime
AI Race strikes me just like Space Race and Arms Race - getting farther out faster and getting and maintaining an edge on the unknowably extensive frontier without any bounds on timeline or technological development.
I totally disagree with this, and would appreciate not being told what I do and do not believe “deep down.”