Can't find evidence of this. Resolves NO?

Does the endpoint need to invoke GPT-4 to satisfy the request, or is it sufficient that GPT-4 be used to generate training data for a lesser model?

Discussion bw Lex and Saltman about moderation tooling and AI 

 and this (and other stuff elsewhere in the report) from the systems card

A product that is to be sold for integration with other platforms for automated mods, filtering <harmful, sexual, etc> type of content. Based on GPT-4 

🗳️ Politics

Related questions

Related questions