AI Futures Project's AI 2040 Plan Criticized for Lacking Dispute Procedures
Mutual compute destruction is proposed, but how to resolve conflicts remains undefined.
The 'AI Futures Project' has released 'AI 2040: Plan A,' a normative scenario describing what should be done to manage many powerful AIs. The plan has two praised elements: limiting the intelligence gap between internal and external model deployments, and radical transparency about how AIs are trained. It also proposes a mutual assured compute destruction stalemate between the US and China, where each can destroy the other's compute infrastructure.
The primary objection from reviewer 1a3orn is that the plan contains no actual decision procedures for how nations would negotiate and resolve disputes over compute usage. Without such procedures, the mutual vulnerability framework is dangerously incomplete. For example, if a Chinese company discovers continual learning that could invalidate safety cases, there is no international process to decide whether to allow it. The plan details how regulators bind companies internally but offers nothing for intergovernmental conflict resolution, leaving a critical governance gap.
- AI 2040 proposes radical transparency and limiting the intelligence gap between AI companies and the public.
- The plan's centerpiece is mutual assured compute destruction between the US and China.
- Critics note the lack of international decision procedures for resolving compute-related disputes between nations.
Why It Matters
Real-world AI governance requires enforceable international agreements, not just hardware threats — or chaos will ensue.