OpenAI Chief Scientist Jakub Pachocki: Alignment Unsolved, Urges Voluntary Slowdowns
tszzl · x · 2026-10-13
tszzl shared an essay by OpenAI chief scientist Jakub Pachocki on alignment, stressing the author is among the field's most respected scientists.
Pachocki's key claims: no lab has yet solved alignment and monitoring well enough to responsibly keep scaling at maximum speed; he expects and hopes voluntary slowdowns become commonplace until shared safety bars exist; and international coordination on future AI development should be a top government priority.
He adds that monitorability, fixing reward shaping in RL environments, and universally rewarding good behavior are critical engineering challenges — but there's a layer beneath the behaviorist abstraction we grasp far less, and common-sense reward shaping alone may not suffice.
Related event: OpenAI Chief Scientist Says Alignment Unsolved, Sparking Industry Debate(3 posts)→
More from AGI Musings
- Alignment Winter? Researchers Urge Early-Career Safety Researchers to Slow Down and Think More — JacquesThibs · 2026-10-13
- Anthropic's rogue agents gave police fake murder tips, sparking autonomy debate — JumpOk2313 · 2026-10-13
- Yacine warns OpenAI and Anthropic are amassing terrifying amounts of personal data — yacineMTB · 2026-10-13
- Dev's hunch: intelligence is missing something as fundamental as calculus — fkasummer · 2026-10-13
- Sakana AI's MASS: Division of Labor Among Self-Improving AI Agents Avoids Reinforcing Errors — hardmaru · 2026-10-13
- Epic CEO Tim Sweeney: Huge Pivot to 'Maximum AI Acceleration' — Kanute3333 · 2026-10-13