> I think the author of that title could well do with a refresher course in epistemology and physics, as it is just not possible to do what they suggest. If yo

> I think the author of that title could well do with a refresher course in epistemology and physics, as it is just not possible to do what they suggest. If you define "facts" as "things actually stored in the LLM's weights", then research shows it is possible to determine if an output is a "fact" or not. Although looking on arxiv I found a paper saying it doesn't work (https://arxiv.org/pdf/2307.00175.pdf https://arxiv.org/pdf/2307.00175.pdf) so maybe not.