Diamandis: Labs should 100x alignment work and aim thousands of agents at it
PeterDiamandis · x · 2026-09-14
Peter Diamandis argues that if labs are truly concerned about AI risk, they should scale alignment research 10x–100x rather than call for slowdowns. His proposal: aim tens or hundreds of thousands of agents at the alignment problem, run the experiments, publish results, and make alignment the optimization function itself.
Related event: Diamandis: Scale Alignment Research 100x Instead of Slowing AI(2 posts)→
More from AGI Musings
- Andrew Ng: Don't stigmatize using AI to solve hard math problems — andrewgwils · 2026-09-14
- National survey of 10,000 Australians on AI, work and the nation's choices — TobyWalsh · 2026-09-14
- Dario Amodei: AI could double human lifespan and cure cancer within a few years — xiaosun86 · 2026-09-14
- Why AI Control Is Still a Working Hypothesis, Not a Finished Design — AryHHAry · 2026-09-14
- Hinton compresses his AGI timeline again: now 'maybe 10 years, maybe less' — Chris_Armstrong · 2026-09-14
- "I Now Suspect All Writing": Engineer on How AI Slop Broke Reading — cramforce · 2026-09-14