An insurmountable risk of using AI in medicine

22 октября, 2021 от Kinok Выкл

Look at the drawing with examples of various medical images.

Can you determine the patient’s racial affiliation?

Of this, it would seem not a terrible fact (after all, no one bothers the fact that the AI, for example, it is better than people in chess) it can be concluded about the serious danger of using AI in the tasks of medical imaging.

Logic here is such.

1) Recognition of medical images When making diagnostic and therapeutic solutions is one of the most massive and promising areas of applied use of machine learning systems (hereinafter referred to as AI).

2) attempts to use AI to make decisions in a wide range of applications based on judgments about people (eg, social, personnel, financial, police evaluating systems) were faced with a serious problem of racial bias, due to the data on which these AI was taught .

3) in context 1 and 2, new authoritative joint

His findings are as follows:

The results of this study in the general context of applied use AI for making decisions are almost like a sentence:

But after all, it is in such knowledge that inaccessible to us and the highest value of the AI. Otherwise, the AI will turn into an «intellectual excavator» that facilitates the process of «intellectual digging» in problems when making decisions.