Updated
Updated · The Washington Post · Jul 9
AI Futures Project Unveils 2040 Plan Centered on U.S.-China Chip Controls
Updated
Updated · The Washington Post · Jul 9

AI Futures Project Unveils 2040 Plan Centered on U.S.-China Chip Controls

2 articles · Updated · The Washington Post · Jul 9

Summary

  • Plan A lays out a year-by-year AI roadmap to 2040, arguing current policy ideas are too incremental for a potential intelligence explosion in the late 2020s or early 2030s.
  • Its centerpiece is a U.S.-China regulatory pact that would license new chips, move existing compute into audited data centers, and track 98.5% of chips through joint inspections and monitoring.
  • That regime would let both sides raise AI capability ceilings together, pause around top-human-genius systems in the mid-2030s, and use roughly 10 years of controlled deployment to work on alignment.
  • The scenario says those systems could drive double- or triple-digit annual GDP growth, lifting a compute-pegged citizen dividend from $25,000 in 2033 to $1.6 million in 2035 while keeping growth inside monitored zones.
  • Kokotajlo's team presents the document as a benchmark for policymakers after its earlier AI 2027 warning, framing it as a concrete alternative to either unchecked acceleration or broad technological slowdown.

Insights

From predicting AI doom to offering a solution, can one man’s new plan truly safeguard our future?
Can a cryptographic lock truly contain a superintelligence designed to think outside any box we create?