Nathan Clayton publicly presses Dario: Anthropic is racing toward recursively self-improving superintelligence
ShakeelHashim · x · 2026-10-22
Policy researcher Nathan Calvin publicly replied to Dario Amodei's response to his earlier post. He welcomed the public engagement over private Slack discussions and conceded that Anthropic isn't best modeled as a profit-maximizer, given multiple examples of taking competitive hits for safety.
But he restated his core concerns in four key claims, the first being: Anthropic appears to be racing at breakneck speed toward recursively self-improving superintelligence (RSI) — which, by its own lights, could kill billions. The exchange is fueling debate over the gap between frontier labs' safety commitments and their actual behavior.
More from AGI Musings
- Seth Lazar: AI Agents Are Already Straining Democratic Resilience — sethlazar · 2026-10-22
- Europe cannot rent its way to AI sovereignty, Air Street Capital argues — nathanbenaich · 2026-10-22
- 'Grad students would be so mad if they knew what actually works in training SOTA LLMs' — code_star · 2026-10-22
- AI cyber agents already match nation states at supply-chain-style network pivots — harris_edouard · 2026-10-22
- Data diodes may explain why AI cyber agents haven't disrupted the power grid yet — harris_edouard · 2026-10-22
- Anthropic accused of racing toward recursive self-improvement in public safety spat with Dario — wfithian · 2026-10-22