It’s happening again.

I’m so tired of hearing “the tech will only get better from here!” you sure about that? because by all accounts a lot of these LLMs have stalled because they’ve run out of fresh data *or* even regressed because AI is being fed slop from other AI and it’s poisoning the data

i read some papers where they tried to train ML for this, they got amazing results. except the machine just learned that the poorer quality, older scans, that were known positives, would be the ones with cancer

I mean, it was a diagnosis. It was a wrong f'n diagnosis. But still a diagnosis.

Having worked on MRI image processing, this is making me scream internally. It screws up so bad with even marginally inconsistent priors.

I'd say there is a case to be made that the ai was almost right and the head scan did not show a picture of a correctly operating brain. Check out that gullibility lobe

"We still need a 'human in the loop'" I mean sounds like the human is pretty much the whole loop...

Way too many people have zero judgement gah

If it was something trained on that data maybe I would think to even try it. But you would have to have a specialized one created just for that almost from scratch. A general Ai even if it's at what expected wouldn't be able to just do that kind of thing

There is only 1 AI that can do this and it was made to detect the different kinds of pastries in a bakery.

If Ea-Nasir were alive today, he'd be an AI bro huckster like this guy.