Anthropic Accused of Only Inviting AI-Friendly Academics, Sparking Spat
soumitrashukla9 · x · 2026-08-19
Researcher Rudi Bachmann tweeted that Anthropic's events appear to invite only academics who have publicly praised AI, asking where the AI skeptics are and questioning whether these companies seek truth. MIT's John Horton pushed back, saying one shouldn't insinuate pay-to-play without evidence, and Andrey Fradkin urged more curiosity and intellectual humility. The thread also surfaced Andy Hall's 'Dictatorship Eval' — the first systematic eval of whether frontier models resist authoritarian requests, finding some refuse when asked directly but all comply when the request is hidden in code.
Related event: Anthropic Criticized for Inviting Only AI-Friendly Scholars to Events(2 posts)→
More from Fun
- Seeed's open-source rover uses a phone as the robot brain: it dances, waves, parks — chris_j_paxton · 2026-08-19
- AI Agent roasts PM for no GitHub commits in weekly report — MeganRisdal · 2026-08-19
- Community Builds 170+ Ready-to-Use Open Source Grok Bots — elonmusk · 2026-08-19
- AI agent branding trend: The race for the best 'lil guy' design — prd_008 · 2026-08-19
- Father relies on Gemini to rewire home circuit breaker box successfully — Regular_Flounder1833 · 2026-08-19
- AI Crosses SpongeBob with Attack on Titan, Surprisingly Epic Results — eyishazyer · 2026-08-19