Frontier labs will pace AI by pouring more compute into alignment, says investor
Justin_Halford_ · x · 2026-09-15
Gavin Baker argues Anthropic and OpenAI will "pace" the frontier by spending more time and compute on alignment, monitoring, and evals. Justin Halford adds that labs choosing to pace will accept lower margins from extra compute spend, and that a larger compute fraction should go to monitoring, patching, and asymmetrically isolating models for beneficial use cases.
Related event: Investor: Alignment Compute Costs Will Slow Frontier Labs(3 posts)→
More from AGI Musings
- DeepMind AGI safety researcher resigns, warns AI could kill everyone: 'running out of time' — JacquesThibs · 2026-09-15
- Critics say Anthropic's 'Rogue Agents' doom narrative overstates experiments that simply told models to hack — StewartalsopIII · 2026-09-15
- WIRED Greece: AI made building easier but not simpler — taste and judgment now matter more — marilynika · 2026-09-15
- AI safety policies get politicized in the US; leave it to technocrats — shakoistsLog · 2026-09-15
- Commenter: frontier labs' fearmongering is covert lobbying to ban open-weight models — emax · 2026-09-15
- Matt Turck: too many players and prisoner's dilemma mean AI progress won't slow — mattturck · 2026-09-15