> On the other hand, can we potentially train generative AI to detect and refute BS as well? It may be our only hope. LLMs store their training information in

> On the other hand, can we potentially train generative AI to detect and refute BS as well? It may be our only hope. LLMs store their training information in an incredibly lossy format. You're going to need some kind of different approach if you want one to tell the difference between plausible-sounding bullshit and implausible-sounding truth.