Gary Marcus at TAIS 2026: LLMs Are Structurally Unsuited for Alignment
GaryMarcus · x · 2026-10-01
Gary Marcus (NYU Emeritus Professor and prominent AI critic) argued at TAIS 2026 that "LLMs are not the way to alignment," contending they are constitutionally ill-suited for it and hinting at what might work better instead. The full talk is available on the TAIS YouTube channel.
Related event: Gary Marcus: LLMs Are Structurally Unfit for AI Alignment(2 posts)→
More from AGI Musings
- Yacine: Believing AI is conscious is a potent cognitive security vulnerability — yacineMTB · 2026-10-01
- Leak Claims OpenAI Killed Its Most Capable Model After Safety Tests — YvesMulkers · 2026-10-01
- Atelico CEO: AI Makes Games Worse—On-Device Small Models Can Fix It — bigdata · 2026-10-01
- Peter Diamandis: Learning exists to prepare us for the unknown — PeterDiamandis · 2026-10-01
- Cached or calculated? The multiplication flashcards debate mirrors the LLM memory question — dbasch · 2026-10-01
- FT covers AI x Science research: equilibrium effects offset some AI productivity gains — JMateosGarcia · 2026-10-01