The IBM Watson supercomputer, which after a number of examinations received praise for the effective selection of ways to treat cancer, can still make mistakes. This conclusion was made by medical experts, analyzing the data on recommendations to patients. It turned out that the machine is mistaken regularly enough, and sometimes these errors can be fatal for the patient.
One example is the situation with a 65-year-old man who has been prescribed a drug by the AI that can in some cases cause uncontrolled bleeding. The patient had this moment indicated in the diagnostic map, however, the artificial intelligence did not take note of the information.
The problem, experts say, is that the medical engineers responsible for training AI have “fed” information to him about hypothetical patients, and not real cases. As a result, some experts, without restraining themselves in feedback from IBM, called the supercomputer a “piece of **** [fecal matter]”.
Despite all the difficulties, the recommendations of AI continue to be applied in practice, although the doctors and do not use them without looking back. And IBM is well aware that the system needs to be modified.