[lovich]

When AI doesn’t hallucinate as a fundamental aspect of the technology, that might be a viable option.

As it is, throwing a statistical word guessing machine at legal issues is a nightmare. Look at the lawyers already sanctioned for using AI that cited fake cases[1]

[1] https://natlawreview.com/article/lawyers-sanctioned-citing-a...

[tim-tday]

Generative AI will always hallucinate because it is impossible to separate the fact lookup part from the generative part.

This problem has been obvious for years and the evidence shows that hallucination rates have been steady the whole time.