It's interesting to get into factual arguments with an LLM. If something's obvious, i.e., all over the training data, it might argue with you, but you can often get it to adopt and endorse all kinds of ridiculous ideas, especially if they're built-in assumptions in your prompt
Reposted from
ryan cooper
I've been studying a bit on the math behind LLMs and it is legit very interesting. but these things just do not have anything to do with truth bsky.app/profile/nucl...
Comments