Why did the doctor perform further said testing to verify the cancer? Because an AI predict it. And we prefer more false positives than false negatives, so we test the positive.

yeah, that’s not how medicine works. medicine works through diagnosis, testing, then treatment, especially with cancer. no doctor is going to trust software to diagnose a patient and then prescribe treatment blindly because it’s the doctors license on the line if the patient was treated without proper diagnosis.

Testing for medicine as poison will be done no matter if it was found by humans or not. Searching for potential medicine faster is a welcome in my book. Rather than finding being the bottleneck, I’d rather test be the bottleneck. It means we will have a potential answer than none at all.

this is the most brain numbing take. AI can generate 15 billion compounds with medical implications. out of those only 200 are viable. out of those 15 aren’t toxic to humans. problem is, it’s going to take 50 years to find those 200 and another 25 years for the 15. in the meantime all medical research has been dedicated to finding those 15 medications for 75 years and have completely ignored research into specific medicines to treat problems now. the biggest joke about those 15 medicines? they’re all “boner” pills because the model was trained on Pfizer data.

As for the astronomer case, it is true for every field. Cancer detection? Ideally, a doctor/medical technician feed the AI the data, and the doctor must also check the output of said AI. A simple X-ray scan with a marker marked as cancer will have a lot of parameters that the doctor could understand that a layman may not. Maybe it is the size, maybe it is the opacity, maybe it is the location, and many other things.

what’s your point? of course you need specialists to train the models, that’s besides the point I made. I’ll be blunt with you, AI is only as good as the data that built the model. we as humans often forget that we live in an interconnected world that has broad impacts on every topic. to build a model that only specializes in one topic will get you results that are biased towards the topic and whatever data you trained the model under. it will never see the world as a nuanced place and only see’s the world in what you tell it to see.

the examples you gave were poor. it’s not your fault because you only drank the koolaide that was put in front of you.

AI excels at pattern recognition, if you keep the focus small enough it can be a useful tool. what you’re doing however is “praising the hammer for building a house” when it was the team of carpenters, roofers, electricians, plumbers, drywallers, painters, etc that actually built the house.

Every single rebuttal that you did does not paint humans in a good light.

I have no idea why you would say that. everything I said painted humans in a human light. just like the poor examples you gave, you are human and are fallible. To believe you can live in a world where software is infallible is just foolish and naive.