The reason why China would go for Plan A is mostly that the status quo is terrible for them, and as the AIs get more capable and the US stays ahead, we expect them to wake up to this reality.
Seb’s assertion is that your consideration of the geopolitics of negotiations with China is naïve, and this response doesn’t seem to refute that. A negotiated settlement is only one of many options in China’s toolbox. Some potential considerations:
What about the optics of this deal being reminiscent of the Century of Humiliation? Won’t that make this politically infeasible for Xi?
What about if China instead invests in espionage and exfiltration of model weights? Are any labs currently hardened against state-level cyberattacks, and if not does your plan explain how they would become so?
What if China instead puts their full backing behind companies like Moonshot and relies on catching up to the frontier? K3 seems to make this a promising approach from a Chinese perspective.
China and Russia have failed to come to an agreement on the Power of Siberia 2 pipeline for years despite it clearly being in both of their best interests. What does this say about the prospects of reaching a deal on AI, even if it is in the best interests of everyone involved?
Trump’s president through 2028. How well do you trust his negotiating skills? How have his recent negotiations with China gone?
China is unique among nuclear superpowers in refusing to use a military deescalation hotline, going so far as to weaponize the hotline’s very existence. What does that say about their expected future behavior on x-risk mitigation?
Getting China onboard for this proposal would be somewhere on the order of the greatest geopolitical coup in the history of the world. Treating it like an afterthought that will just naturally work itself out “because AGI/ASI” feels like skipping the hard part of the conversation. I would have wanted to see drastically less sci-fi speculation and drastically more engagement with on-the-ground facts of current geopolitical reality, since that’s what will actually make or break this plan.
Seb’s assertion is that your consideration of the geopolitics of negotiations with China is naïve, and this response doesn’t seem to refute that. A negotiated settlement is only one of many options in China’s toolbox. Some potential considerations:
What about the optics of this deal being reminiscent of the Century of Humiliation? Won’t that make this politically infeasible for Xi?
What about if China instead invests in espionage and exfiltration of model weights? Are any labs currently hardened against state-level cyberattacks, and if not does your plan explain how they would become so?
What if China instead puts their full backing behind companies like Moonshot and relies on catching up to the frontier? K3 seems to make this a promising approach from a Chinese perspective.
China and Russia have failed to come to an agreement on the Power of Siberia 2 pipeline for years despite it clearly being in both of their best interests. What does this say about the prospects of reaching a deal on AI, even if it is in the best interests of everyone involved?
Trump’s president through 2028. How well do you trust his negotiating skills? How have his recent negotiations with China gone?
China is unique among nuclear superpowers in refusing to use a military deescalation hotline, going so far as to weaponize the hotline’s very existence. What does that say about their expected future behavior on x-risk mitigation?
Getting China onboard for this proposal would be somewhere on the order of the greatest geopolitical coup in the history of the world. Treating it like an afterthought that will just naturally work itself out “because AGI/ASI” feels like skipping the hard part of the conversation. I would have wanted to see drastically less sci-fi speculation and drastically more engagement with on-the-ground facts of current geopolitical reality, since that’s what will actually make or break this plan.