Critics call out AI labs for safety-washing: can't secure it, don't deploy it
gerardsans · x · 2026-09-20
In a debate with @Kantrowitz and @So8res, gerardsans argues AI risk is hidden behind a smokescreen of safety and alignment talk that gives labs more decency than they deserve, invoking the engineering principle "can't secure the system? Don't deploy it." He adds that betting against a probabilistic sampler means accepting the consequences of its errors, and blames the industry for keeping AI's probabilistic nature out of public discourse.
More from AGI Musings
- Frontier model outputs now need another model to read them, insider says — DavidSKrueger · 2026-09-20
- A single web search makes LLM inputs OOD, argues developer — menhguin · 2026-09-20
- RL-trained LLMs have functionally infinite search space, replies Jack Clark — menhguin · 2026-09-20
- Two types of daily AI users: becoming more slop or more human — StewartalsopIII · 2026-09-20
- Spotify reframes recommender systems for the agentic web in weekly IR papers roundup — _reachsumit · 2026-09-20
- AI moral status debate: does conceding machine suffering imply utilitarian math? — teortaxesTex · 2026-09-20