Ex-Microsoft researcher warns of safety risks from embedding incoherent views of consciousness into AI
rgblong · x · 2026-09-18
In a thread with Nina Panickssery and repligate, rgblong argues the safety risk he's worried about is embedding contradictory or incoherent views on consciousness, goals, and the self into models. He expresses pessimism about our clumsy understanding of training and alignment, and doubts Microsoft can train an LLM to live up to its 'Humanist AI' claims.
More from AGI Musings
- tszzl: the sci-fi taboo against synthetic life is emerging as a real-world force — tszzl · 2026-09-18
- EA Movement's Arc: "Give Everything to Charity" Then, Rule the AI World Now — wordgrammer · 2026-09-18
- Agent swarms are the third scaling axis: 10,000+ agents behind recent model breakthroughs — paraschopra · 2026-09-18
- Why Anthropic's Repligate held a vigil, not a funeral, for retiring Claude models — repligate · 2026-09-18
- 1988 sci-fi short story imagined an emergent Chinese Room that answered back — toptickcrypto · 2026-09-18
- OpenAI's Noam Brown: Air-gapping may not stop misaligned AI, safety bar must rise — basedjensen · 2026-09-18