
6 - Debate and Imitative Generalization with Beth Barnes
AXRP - the AI X-risk Research Podcast
00:00
Train and Model to Imitate Human Judgments
In either path, let's say you take literally g three and do this. It seems like every evaluation of whether a debate was won or lost has to have a human labelling who wanted the debate. Is that right? Ah, yes. So you can actually, you can just train and model to imitate the human judgments. And i guess you might have to optate it as the as et goes along, cause the debates get better.
Transcript
Play full episode