Contents
American scientists from MIT are trying to fight the problem of artificial intelligence racism using a non-trivial tool – chest x-rays
What’s going on
- Researchers at the Massachusetts Institute of Technology have taught artificial intelligence to determine a person’s race from x-rays.
- To do this, scientists first proved that the algorithm can predict race in several ways and based on different data.
- First, the AI model was taught to determine race on chest x-ray datasets. After – on other parts of the body: mammography, X-ray of the cervical spine and CT scan of the chest were used.
- The scientists also proposed models for the physical characteristics of patients: body habitus, breast density, diseases, the effects of stress, and even social prejudice. Comparing the results of the predictions on the two data, the researchers found that the model best determines the race of the chest X-ray.
- The results of the model are not affected by the color of the organ images. The algorithm takes into account all the data in the image.
What does it mean
The development of artificial intelligence is accompanied by ethical and legal issues. There are cases when AI behaved unethically and even violated the law.
In the United States, artificial intelligence is used in courts to calculate the likelihood of a crime. The investigative journalism project ProPublica found that AI marked black defendants as incorrigible criminals twice as often as white defendants, without considering other circumstances.
To resolve legal issues related to violations of laws by artificial intelligence, it is necessary to solve a number of problems in the legislation: to define a robot, to understand who is responsible for the violations of robots and who owns the rights to intellectual property created by AI.