This isn’t just about GPT, of note in the article, one example:

The AI assistant conducted a Breast Imaging Reporting and Data System (BI-RADS) assessment on each scan. Researchers knew beforehand which mammograms had cancer but set up the AI to provide an incorrect answer for a subset of the scans. When the AI provided an incorrect result, researchers found inexperienced and moderately experienced radiologists dropped their cancer-detecting accuracy from around 80% to about 22%. Very experienced radiologists’ accuracy dropped from nearly 80% to 45%.

In this case, researchers manually spoiled the results of a non-generative AI designed to highlight areas of interest. Being presented with incorrect information reduced the accuracy of the radiologist. This kind of bias/issue is important to highlight and is of critical importance when we talk about when and how to ethically introduce any form of computerized assistance in healthcare.