OpenAI proposes perplexity-based shared safety bars, more pretraining transparency
anpaure · x · 2026-11-23
OpenAI's merettm announces more transparency into pretraining progress, moving toward shared safety bars. The essay "Perplexity and Intelligence" argues for common capability limits based on perplexity and other measurements, with progress beyond them gated on our ability to understand AI behavior. Responder anpaure calls it "directionally very good".
More from Models
- Perplexity's pitfalls: compute dependence, benchmaxxing, and Anthropic's METR redactions — eliebakouch · 2026-11-23
- Keeping up with five flagship models: 'Opus 5.5, Fable 5.2, GPT-6.2 Astra... I'm tired' — Angaisb_ · 2026-11-23
- Chinese open-weight models near closed frontier on coding and math, reshaping enterprise choice — ingliguori · 2026-11-23
- Commenter pushes back on perplexity claim: math gains likely came from RL, not pretraining — stochasticchasm · 2026-11-23
- Skeptic pushes back: math breakthroughs came from RL, not pretraining perplexity — stochasticchasm · 2026-11-23
- Influencer's overall model ranking puts opus 5.5 on top, weighing cost and speed — MatthewBerman · 2026-11-23