Post #3140625
2026-02-15 13:24 UTC
@chesheer@mastodon.bsd.cafe It's not tests, it's exemplars. You do realize that your "passing with flying colors" is a correct guess, right? You are aware that there is no understanding or correctness mechanism at work here - nor will there ever be, because that's not how an LLM works, or can possibly be made to work...?
Replies (1)
-
@chesheer@mastodon.bsd.cafe 2026-02-15 13:30
@jwcph@helvede.net Absolutely. There's of course no "the one who understands" in LLM. That being said we have no idea what "understanding" means in humans - all we have is more or less convincing ideas and notions. At the same time it does not mean that LLM could be in any shape or form equaled to human just because we don't really know how humans work either. At least LLM is not capable to "self-check" its conclusions and retrace its own steps in the same way human can - it just generates new text output.