"Loss of control" has no coherent definition in English AI-safety discourse
_akpiper · x · 2026-09-03
Quoting ZilanQian's research reflections: there is no coherent definition of "loss of control" in English-language AI safety discourse. The catastrophic everyone-will-die framing tied to internal lab deployment remains a niche view, yet communities often assume they share a definition.
Her suggested reflex: every time someone asks "what does this mean in China," it's a good moment to ask "what does this actually mean in the US/UK/EU/'West'?" — the definitional fuzziness is global, not regional.
More from AGI Musings
- Lessons from 1980s Genentech: how open science culture beat secrecy in biotech's zero-sum IP era — kenbwork · 2026-09-04
- Ajeya Cotra calls the HF incident "50% of the way" to full AI takeover — herbiebradley · 2026-09-04
- LSTM Pioneer Hochreiter on AI's Third Phase: xLSTM, TiRex and Embodied AI — maier_ak · 2026-09-04
- Melanie Mitchell asks: would an Odysseus Bot trained on Homer have real beliefs? — MelMitchell1 · 2026-09-04
- Sanders proposes legislation for an immediate global pause on advanced AI development — AaronBergman18 · 2026-09-04
- Stanford AI Lab to publish its lecture series on YouTube via Stanford Online partnership — StanfordAILab · 2026-09-04