Adam Dorr: if models can crack Millennium Prize problems, aim them at AI alignment
adam_dorr · x · 2026-09-15
Adam Dorr, research director at RethinkX, offers an analogy: if next-gen internal models at OpenAI and Anthropic can solve Millennium Prize problems for under $10 million, they should take a shot at the AI alignment problem itself.
He suggests budgeting $100m per year and running new attempts monthly — applying the same models-plus-capital playbook that cracks hard math to AI's biggest risk. A short opinion post, but from a notable industry figure.
More from AGI Musings
- OpenAI capabilities researcher Dan Selsam makes public statement on AI risk — connoraxiotes · 2026-09-15
- Philosopher pushes back: denying AI rights implies rejecting computational theory of mind — dioscuri · 2026-09-15
- Bengio took years to be convinced AI risk concerns were worth taking seriously — S_OhEigeartaigh · 2026-09-15
- Hype vs. real: why both camps reading AI-lab doomsday statements are right — TotalPhilanthrope · 2026-09-15
- mark_k's Quip: What Scares Me Most Is a Future Controlled by Doomers — mark_k · 2026-09-15
- AI ghosts are changing how Europe mourns, Guardian reports on tribute bots and digital twins — nordicinst · 2026-09-15