But to be clear, in this case the LLM has a full, direct, unobscured view of the chicken. A human, in that specific case -- i.e. looking at the same photo* -- w

But to be clear, in this case the LLM has a full, direct, unobscured view of the chicken. A human, in that specific case -- i.e. looking at the same photo* -- would not have trouble discerning and reporting the third leg. Perhaps if they were forced to scan the photo quickly and make a report, or were otherwise not really 'paying attention'/'taking it seriously', but the mere fact that LLMs fall into that regime far more than an 'serious employee' already shows that they fail in different ways than humans do.