New proposal: mandatory compute disclosures plus 15% capabilities cap could delay intelligence explosion a year
sjgadler · x · 2026-11-23
Romeo Dean and Thomas Larsen propose "Pacing AI via Compute Reallocation":
- VCAD (Verified Compute Allocation Disclosures): frontier labs regularly disclose compute use (serving 19%, capabilities R&D 64%, safety research, idle, etc.), verified by an independent auditor.
- Reallocation: shift compute away from capabilities R&D toward serving customers and safety research.
Their rough estimate: a 15% cap starting January 2027 would delay the intelligence explosion by about a year — likely insufficient to prevent AI takeover, so that year should build a durable regime like Plan A or Plan S. Commenters add that input-based pacing beats output-based pacing.
More from Safety
- OpenAI's chief scientist on how to put a speed limit on AI development — TechCrunch AI · 2026-11-23
- Fireworks AI breach hits 150k Uber drivers; US tightens compute-access rules — DavidLinthicum · 2026-11-23
- Fake official account spotted, prominent blogger warns links are unsafe to click — dotey · 2026-11-23
- Why cybercriminals skip AI attacks: too noisy, costly tokenomics, old TTPs still work — jcran · 2026-11-23
- Boaz Barak defends internal perplexity metric: not directly transferable across models — boazbaraktcs · 2026-11-23
- Researchers Push Back on Perplexity Safety Bars: Data Mixture Bias Undermines Cross-Lab Comparisons — eliebakouch · 2026-11-23