Estimated $20M per extra data point for AI swarm scaling curves, says Toby Ord
tobyordoxford · x · 2026-09-22
In a discussion on how AI agent-swarm capabilities scale with size, Oxford philosopher Toby Ord speculates that a frontier lab spent roughly $20 million running a 10K-agent swarm to solve a millennium prize math problem for the PR. Getting one more data point per swarm size would cost another $20M, and a full curve with reasonable confidence intervals perhaps $200M. He suspects labs already have curves for 2-, 4-, 8- ... 1,024-agent swarms on hard benchmarks — they just aren't published.
More from AGI Musings
- Business strategy debate: hide the AI behind humans, not humans behind AI — annetgriffin · 2026-11-12
- If superhuman AI researchers are imminent, will capability problems be easier than safety ones? — ronbodkin · 2026-11-12
- "AI is a terrible teacher": viral pushback against personalized AI tutoring hype — RachelVT42 · 2026-11-12
- TissueMicroscopyBench: 49 evals to test AI agents on biological microscopy — kenbwork · 2026-11-12
- OpenAI, Anthropic and DeepMind job posts signal PMs must write evals and code — aakashgupta · 2026-11-12
- Architecture study sparks monitoring concerns over a multi-hop planning spike — ChowdhuryNeil · 2026-11-12