Debate: can OpenAI's plan to have agents formally prove safety properties work?
zetalyrae · x · 2026-11-17
@bicellaria questions whether OpenAI's "The Plan" is to have agents formally prove safety properties despite nobody knowing which theorems need proving — arguing such a plan predictably won't work.
@tszzl counters that this is confused thinking: you can automate safety research while still verifying results, so human oversight isn't handed off; formal verification powers will only grow as we understand these systems better. zetalyrae amplifies the exchange, spotlighting the core dispute over formal methods as an AI-safety anchor.
More from AGI Musings
- “Stop aiming for 20% productivity — AI should end work-for-money altogether” — VraserX · 2026-11-17
- iamtrask: agent time horizon measures stability, not intelligence — don't conflate ASI signals — iamtrask · 2026-11-17
- Blogger: much of white-collar work was just "button clicking," now over as computers learned English — moonsandhues · 2026-11-17
- FT: Firms Investing Heaviest in AI Grew White-Collar Headcount 10.2% — Scobleizer · 2026-11-17
- AI wealth could drive tens to hundreds of billions a year in philanthropy, argues Nan Ransohoff — allisondman · 2026-11-17
- Panagram slop detection could collapse the anti-AI argument, argues MLStreetTalk — burny_tech · 2026-11-17