If you ask a physicist what the shape of the earth is, and his response is “a sphere”, are you going to throw up your hands and say “well, we can’t trust physicists now. Maybe the flat earthers are right…” since it’s not technically a sphere (it’s an oblate spheroid), or do you still recognize that even though they both technically gave wrong answers, one is still clearly more qualified than the other?
Obviously, you can still make that distinction, and it’s the same with doctors, who until very recently were clearly more qualified to give medical advice than an AI language model. What were they supposed to be using to assess this shit over the past few years? The perfect AI doctor that doesn’t even exist yet?
It’s not the final goal though? The tweet was highlighting the rapid pace of improvement, not asserting that % of answers deemed correct by a doctor is the be-all end-all for determining correctness of medical advice.
Hell, we’ve had areas where AI algorithms outperform the pros at certain tasks (such as certain types of cancer detection) for years now. Obviously doctors aren’t the ultimate arbiter of truth.
Did he do something over the top or is this just you being generally fed up?
