4 ms·
> A lot of exposés of LLM weaknesses involve abusing the ambiguities of English with questions that are impossible for humans to give a satisfactory answer to
by empath-nirvana 3y ago
> A lot of exposés of LLM weaknesses involve abusing the ambiguities of English with questions that are impossible for humans to give a satisfactory answer to
There's a general unstated assumption that a truly intelligent entity wouldn't make mistakes, which is sort of contradicted by, you know, the existence of humanity. I think making (certain kinds of) mistakes is a _sign_ of intelligence, rather than a sign that it's not.