3 ms·
> I’d argue hallucinations are unexpected in LLMs by the large (non technical) number of users who use them directly, or indirectly though other services. Peop
by PittleyDunkin 2y ago
> I’d argue hallucinations are unexpected in LLMs by the large (non technical) number of users who use them directly, or indirectly though other services.
People also blithely trust other humans even against all evidence that they're trustworthy. Some things just aren't fixable.
- mdp2021 2y agoThe median individual is _not_ a model, and cannot represent the whole of the set. If the median is incompetent, the competent remain competent.