These faux questions always have a valid interpretation that the asker doesn't admit (for some reason). The model is then castigated for not making an opinionated choice
replyThat’s not what’s happening.
replyThe question is revealing that the model has a model of language but not of reality. It knows what words go together, but not real-world concepts.
This. LLMs are marketed on the false premise of all knowledge, intelligence and wisdom being possible to be encoded in language only.
replyAs a test, explaining away peculiar answers by imagining unlikely outlier scenarios is not the counter you seem to think it is.
replyFor most of them, we’d worry that a human answerer using maximum effort to produce the same outcome was having a stroke.
[dead]
replymaybe unreasoning.
replyalso, naysayers apparently DO have a compelling reason.