Freeze AI capability, cut compute costs: Kokotajlo backs a 'zero existential risk' Plan S
wfithian · x · 2026-11-09
- Daniel Kokotajlo endorsed the idea of freezing AI capability at current levels while focusing solely on lowering compute costs, arguing it could be implemented almost immediately.
- The hard part is keeping the freeze: as trailing labs catch up to the frontier, they'd increasingly need to be stopped from surpassing it — though he thinks that's feasible too.
- He ties this to the "Plan S" branch of his AI 2040 scenario: revolutionize software development with zero existential risk.
Related event: AI community debates freezing frontier capabilities to cut compute costs(2 posts)→
More from AGI Musings
- Berkeley Dinner on the "AI Crisis" Features Talk by Prof. David Krueger — KatjaGrace · 2026-11-09
- Rethinking the Bitter Lesson: 'don't work on things that don't scale,' not 'scale is enough' — gowthami_s · 2026-11-09
- Agentic AI's datacenter GPU bet may be wrong—edge and small models could fill the gap — DavidLinthicum · 2026-11-09
- Claude-estimated CO2 per dollar: AI chatbot subscription emits just 0.009 kg, far below gas heating — AndyMasley · 2026-11-09
- Devs push back on 'AI will build AAA games in months' via 'The Door Problem' — _amirabs · 2026-11-09
- The churn era: customers now use Claude to slop up replacements for overpriced SaaS — joshgonsalves_ · 2026-11-09