Dario claims Anthropic always ran a 'race to the top' — 2022 interview says he pledged not to race ahead
Miles_Brundage · x · 2026-10-21
Responding to safety researcher Nathan Calvin, Dario Amodei wrote that Anthropic's strategy has "always been" a "race to the top" — showing it's possible to build AI responsibly to raise the bar for the entire industry, adding that this strategy works best from the frontier.
Miles Brundage pushed back, citing a 2022 interview where Dario explicitly said Anthropic "shouldn't be racing ahead" and should build models only "reasonably close to state of the art" to enable safety work — framing Anthropic as a "good player and good citizen" attacking "some specific problems," with Daniela Amodei even discussing "disarming race dynamics" via standards bodies. The phrase "race to the top" never appeared in that interview, suggesting the narrative has shifted.
Related event: Amodei Responds to Safety Criticism Amid Backlash(3 posts)→
More from AGI Musings
- AIs that copy themselves enter natural selection — Hendrycks' warning revisited — ZeroStateReflex · 2026-10-21
- Carlota Perez: AI bubble compares to 1929 and 2008, not the 2000 Nasdaq crash — 2C_ornot2C · 2026-10-21
- Pedro Domingos: AI is the end of kabuki software — pmddomingos · 2026-10-21
- Hot take: home automation wins by standardizing spaces, not building robots — alysha_lobo · 2026-10-21
- Paras Chopra: Building tools that build tools is the highest-leverage way to accelerate civilizational magic — paraschopra · 2026-10-21
- Google VP blunt on machine consciousness: machines aren't conscious — neither are some people — docmilanfar · 2026-10-21