Raw LLM Responses

And a few of these 'secret sauce' AI learning programs were learning to cheat. There was one in South Africa attempting to detect pneumonia in HIV patients versus clinicians, and the AI apparently learned to differentiate which X-ray machine model was used in clinics vs. the hospital, and used this data in its prediction model, which the real doctors did not have access to. Because checkup x-rays in outlying clinics tend to be negative, while x-rays in the hospital (where more acute cases go) tend to be positive. https://www.npr.org/sections/health-shots/2019/04/01/708085617/how-can-doctors-be-sure-a-self-taught-computer-is-making-the-right-diagnosis > Zech and his medical school colleagues discovered that the Stanford algorithm to diagnose disease from X-rays sometimes "cheated." Instead of just scoring the image for medically important details, it considered other elements of the scan, including information from around the edge of the image that showed the type of machine that took the X-ray. > When the algorithm noticed that a portable X-ray machine had been used, it boosted its score toward a finding of TB. >Zech realized that portable X-ray machines used in hospital rooms were much more likely to find pneumonia compared with those used in doctors' offices. That's hardly surprising, considering that pneumonia is more common among hospitalized people than among people who are able to visit their doctor's office.