what makes you think that current direction of AI development would lead to making less mistakes than humans do, as opposed to repeating same miskates plus hallucinating more?