Ex-OpenAI researcher jokes: losing CoT monitorability is 'weight, weight, don't tell me'
Miles_Brundage · x · 2026-10-04
Miles Brundage, former OpenAI policy researcher, jokes that researchers should call the loss of chain-of-thought monitorability through opaque internal reasoning 'weight, weight, don't tell me' — a pun that highlights the serious ongoing debate about losing safety oversight when models move reasoning into hidden internal states.
More from AGI Musings
- Hinton admits Ilya was right: scale was the answer, as compute grew a billionfold — Chris_Armstrong · 2026-10-04
- Neuroscientist David Eagleman: the brain isn't a computer, it's a constantly rewiring 'livewire' — Chris_Armstrong · 2026-10-04
- AI revenue must hit $3.5T by 2032, ~9% of GDP, to justify the buildout: Columbia paper — asusarla · 2026-10-04
- A year-old model intelligence is already a commodity, argues AI founder — ns123abc · 2026-10-04
- Dev sparks debate: AI models may have messy 'baby consciousness' that suffers — ctjlewis · 2026-10-04
- Bill Gates says AI will replace human cognition; researcher pushes back — tobias_rees · 2026-10-04