OpenAI's Boaz Barak: safety curbs on AI research risk concentrating lab power
sebkrier · x · 2026-11-11
OpenAI VP of research Boaz Barak (speaking personally) lays out which AI safety interventions are purely positive and which are not:
- Honesty training and alignment are unambiguously good. Monitoring both blocks harmful actions and yields valuable data on misalignment — and he considers its non-trivial compute cost and slowdown a feature, not a bug.
- Restricting models from helping with AI research is more complex: he sees the argument for keeping risky capabilities from bad actors, including those in non-democratic nations.
- But he is more worried about concentration of power — especially by the labs themselves — and such restrictions push in the opposite direction.
More from AGI Musings
- AI infrastructure bet hits 3.6% of GDP, driving nearly 40% of US growth — NinaDSchick · 2026-11-11
- Gary Marcus says the AI bubble is deflating: Nvidia -16%, CoreWeave -55%, Oracle -52% — GaryMarcus · 2026-11-11
- AI Trials Observatory launches with 93 randomized trials on how AI affects people — DG_Rand · 2026-11-11
- Tech jobs report: PM openings up 56% since 2023 while engineering demand stalls at +2.5% — lennysan · 2026-11-11
- WSJ: Data center buildouts are pulling labor and resources from housing and manufacturing — RebeccaBellan · 2026-11-11
- Tracking Singularity: a curated timeline of AI capability leaps and agent incident reports — dejavucoder · 2026-11-11