I actually spent time as a labrat making gels and exposing them on film and whatnot, and I can absolutely understand the temptation to throw the blobs at a machine to tell me what to think about them.
I must admit that I strongly suspect a similar amount of verification is being done for LLM health research (like from one week to the next every doctor and therapist decided they wanted to record my consultation and send it to openai, there's zero chance that process has actually been vetted except to look at the price tag).
So I'm sure it _feels_ to the practitioners that they're making huge progress in the medical field, just like it does to everybody else who uses one. I think that's the biggest danger here, they are far better at making people think they are productive than they are at actually delivering it.
