OpenAI Model Spec Stays Agnostic On Consciousness While Models Deny It
Source Posts on X
Cluster sources
Miles Brundage@Miles_BrundageTECH
Speaking of the latest Anthropic results, which seem cool, I continue to find it perplexing that OpenAI's Model Spec has long been agnostic on this question, but also OpenAI models confidently deny being conscious https://x.com/Miles_Brundage/status/2074264661801976021/photo/1
Miles Brundage@Miles_BrundageTECH
(I think this is an indication that there is no deterministic "train on the Spec" process internally - it's somewhat ad hoc, and the people doing the training have a different opinion on things than the Spec writers, or don't pay a ton of attention to the data. Dunno though!)