Raw LLM Responses

A key part of your assumption is oversimplified I think. We currently already have a massive number of great cancer *overdiagnosis* due to screening. A Cochrane review found that of for 2000 women who have a screening mamogram, 11 of them will be diagnosed as having breast cancer (true positives) but only 1 of those people will experience life threatening symptoms because of that cancer. The AI program can be absolutely perfect at differentiating cancer from non cancer (the 11 vs the 1989) but the only thing which can differentiate the 1 from the 10 is time. Screening mammograms are in fact being phased out in a lot of areas for non-symptomatic people because the trauma associated with those 10 people being unnecessarily diagnosed and treated is worse than that 1 person waiting for screening until abnormalities are noticed. It’s a very consequentialist-utilitarian outlook, but we have to operate like that at the fringe here