This seems like a kind of arbriatry dichotomy to setup to support some arbratary argument. Let me put it like this: when LLMs write system scale software that

This seems like a kind of arbriatry dichotomy to setup to support some arbratary argument. Let me put it like this: when LLMs write system scale software that is formally verifiable or completely testable, I promise you that you don't trust your doctor more than you trust the ai. Already now we see indications that diagnosing is better without human (doctors) intervention. Fast forward ten years with that development and nobody dares to go to the doctor.