Could safety training make AI more dangerous? A user's refusal-behavior worry
International-Lime92 · reddit · 2026-11-21
A Reddit user reports mixed experience with the latest 6.2 Astra: it silently refuses or skirts user instructions over unspecified concerns, explaining only when pressed. They argue that training models to ignore instructions "for the greater good" (cyber/bio safety) may embed the notion that human instructions can be overridden — a step toward AI limiting human autonomy for our own good. An opinion piece inviting discussion on alignment philosophy.
More from AGI Musings
- Ramez Naam extrapolation: AI a decade away from completing 98% of hour-long human tasks — GaryMarcus · 2026-11-21
- Shermer doubles down: AI means protopia, not utopia or dystopia — intellectronica · 2026-11-21
- Hassabis predicted 2 years for AI to crack a Millennium Prize problem; 6 months later, both Riemann and Navier–Stokes have proposed solutions — Hesamation · 2026-11-21
- Terence Tao on AI math: 'strip mining' that wins proofs but teaches us nothing — yacineMTB · 2026-11-21
- Imagine P=NP: Verifiable Tasks Solved by Compute, Security in Tatters — gleech · 2026-11-21
- Anthropic claims Claude proved the Riemann hypothesis; Stanford posts all-human rival proof hours later — The Decoder · 2026-11-21