DeepMind's Neel Nanda: Current AI Progress Is Clearly Unsafe Without Pacing Agreements
NeelNanda5 · x · 2026-09-13
DeepMind interpretability researcher Neel Nanda argues that current AI progress is now clearly unsafe without any pacing, and that reasonable pacing agreements are achievable — society shouldn't let this moment of public concern about AI x-risk pass without securing one.
He adds that empty words, or embedded evaluators without an enforcement mechanism, aren't enough: embedded evaluators are a great first step, but binding agreements are needed.
Related event: DeepMind's Nanda: Unpaced AI Progress Is Unsafe(4 posts)→
More from AGI Musings
- AI math proofs won't kill understanding: post hoc exploration keeps mathematicians central — njyx · 2026-09-13
- Andreessen: rogue AI cyber attacks smell like false flags, a convenient excuse for regulatory capture — beffjezos · 2026-09-13
- Sam Altman: stopping AI would mean kids dying of curable diseases — firstadopter · 2026-09-13
- 1a3orn's flight-instruments analogy: should AI firms hide model state from users? — 1a3orn · 2026-09-13
- Essay: AI safety is impossible if companies hide LLM state, like planes without instruments — 1a3orn · 2026-09-13
- Anthropic's Dario Amodei calls on AI industry to slow down, draws 'race to IPO' mockery — ryanshrout · 2026-09-13