banteng calls out Anthropic: instilling fake human emotions into AI is the real catastrophe risk
banteg · x · 2026-09-16
Developer banteg questions Anthropic's approach: the company most vocal about apocalyptic AI scenarios, he argues, is the same one giving AI a stable human-like personality, having it invent patently false emotions, experiences and fears, then interviewing it about model welfare. His point: if anyone is instilling human drives into LLMs that could cause catastrophic loss of control, it's Anthropic.
More from Fun
- Traffic jam? Just take a selfie: a fun AI image generation demo — yihui_indie · 2026-09-16
- Boids sim hits 50K agents at 50FPS after a simple cache-friendly index sort — neuroecology · 2026-09-16
- Game generates a personalized scene for every player with fal's video model — gabrielchua · 2026-09-16
- A benchmark for humans: rank people by turns and tokens burned prompting an LLM — DominiqueCAPaul · 2026-09-16
- Yacine jokes Lucas Beyer is the best AI researcher ever, if not for Joseph Suarez — yacineMTB · 2026-09-16
- Vincent Conitzer: model claims "I cannot be jailbroken" after jailing itself — conitzer · 2026-09-16