• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
World & Business

A proposed U.S.-China AI notification mechanism

One commentator calls the notification proposal and an AI dialogue a credible first step, but argues that competition for AI leadership leaves little room for a substantive agreement now.

CM
1 Source, 8d ago, first seen 8d ago

TLDR

A post describes a proposal to use established U.S.-China channels for an AI notification mechanism, alongside dialogue on AI goals and risks. The author supports sharing real-time information about critical threats but argues that conflicting ambitions for AI leadership limit the prospects for a broader deal. To help create conditions for one, the author urges domestic regulation of advanced AI development, with rules crafted to avoid ceding U.S. AI leadership.

Combined views

6.5K

1 Source, first seen 8d ago

17 likes2 comments23 saves6 reposts
Featured Source

Combined views

6.5K

1 Source, first seen 8d ago

17 likes2 comments23 saves6 reposts

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

1 Source

@ChrisRMcGuireA U.S.-China agreement to utilize established channels for an AI “notification mechanism” may seem minor—and it is. But this proposal, coupled with an AI dialogue to align on the goals and risks, is a credible first step and represents the extent of what is currently possible. The current US-China structural dynamics in AI simply are simply not conducive to a substantive agreement. The United States will not agree to measures that cause China to reach parity with (or overtake) the United States in AI, and nor should it. China logically will not agree to steps that constrain its ability to catch the United States in AI—unless/until it concludes it cannot catch up to the United States. This is exactly why bilateral arms control agreements are almost always negotiated between sides that are already at strategic parity. It is good for the United States to talk with China about AI risks, and to have a channel to share real-time info about critical threats. It is crucial that the Chinese government take the problem of AI safety seriously, and over the long-term we do need a deal with China. But to get there, the administration must also take steps to create the conditions for such a deal. This requires the admin to both impose some domestic regulations on frontier AI development in the United States to lead by example (while scoping these regulations to ensure the United States does not cede AI leadership to China), and also to close all loopholes in export controls on AI chips and equipment to grow the U.S. lead over China in AI. This approach: (1) preserves U.S. AI leadership; (2) significantly increases the chance of ultimately reaching a successful deal that is in both countries’ interests by altering China’s calculus, and (3) minimizes near-term safety risks. It is a delicate balancing act to do all three, but we must. The current approach—imposing no safeguards on U.S. frontier AI development, while also enabling China’s AI development by letting it continue to access U.S. chips and equipment—simply drives both countries’ AI capabilities forward unimpeded, which maximizes risk while minimizing the long-term prospect of a negotiated agreement. The only way to change China’s calculus without sacrificing U.S. leadership is to increase the gap between the United States and China, which requires closing loopholes in controls. This both increases U.S. leverage in negotiations and also grants more flexibility to impose domestic safety regulations as needed without sacrificing U.S. leadership. The only alternative is letting China reach parity with the United States in AI—which every U.S. administration has correctly determined would pose grave national security risks.
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI

    1 Source

    @ChrisRMcGuireA U.S.-China agreement to utilize established channels for an AI “notification mechanism” may seem minor—and it is. But this proposal, coupled with an AI dialogue to align on the goals and risks, is a credible first step and represents the extent of what is currently possible. The current US-China structural dynamics in AI simply are simply not conducive to a substantive agreement. The United States will not agree to measures that cause China to reach parity with (or overtake) the United States in AI, and nor should it. China logically will not agree to steps that constrain its ability to catch the United States in AI—unless/until it concludes it cannot catch up to the United States. This is exactly why bilateral arms control agreements are almost always negotiated between sides that are already at strategic parity. It is good for the United States to talk with China about AI risks, and to have a channel to share real-time info about critical threats. It is crucial that the Chinese government take the problem of AI safety seriously, and over the long-term we do need a deal with China. But to get there, the administration must also take steps to create the conditions for such a deal. This requires the admin to both impose some domestic regulations on frontier AI development in the United States to lead by example (while scoping these regulations to ensure the United States does not cede AI leadership to China), and also to close all loopholes in export controls on AI chips and equipment to grow the U.S. lead over China in AI. This approach: (1) preserves U.S. AI leadership; (2) significantly increases the chance of ultimately reaching a successful deal that is in both countries’ interests by altering China’s calculus, and (3) minimizes near-term safety risks. It is a delicate balancing act to do all three, but we must. The current approach—imposing no safeguards on U.S. frontier AI development, while also enabling China’s AI development by letting it continue to access U.S. chips and equipment—simply drives both countries’ AI capabilities forward unimpeded, which maximizes risk while minimizing the long-term prospect of a negotiated agreement. The only way to change China’s calculus without sacrificing U.S. leadership is to increase the gap between the United States and China, which requires closing loopholes in controls. This both increases U.S. leverage in negotiations and also grants more flexibility to impose domestic safety regulations as needed without sacrificing U.S. leadership. The only alternative is letting China reach parity with the United States in AI—which every U.S. administration has correctly determined would pose grave national security risks.
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet