Dataset builder says most academics told her LLMs aren't real AI — and advised "don't"

BlancheMinerva · x · 2026-11-22

AI researcher Blanche Minerva recounts conversations with academics about LLMs: the vast majority said LLMs aren't real AI, the questions are uninteresting, and improving LLMs isn't their job. Even when she explained she builds LLM datasets and wants better ones, the advice was simply "don't." She also recalls trying to discuss constructive versions of their critiques — e.g., how to adjust training data so gender/profession bias mirrors BLS statistics — without getting substantive engagement.

Original post →

More from AGI Musings

AGI Musings channel →