MIT scientists make the first implementation of psychopathism in Artificial Intelligence

Three scientists at MIT (Pinar Yanardag, Manuel Cebrian and Iyad Rahwan) have created the world’s first psychopath artificial intelligence, called Norman, with the purpose of demonstrate that AI cannot be unfair and biased unless such data is fed into it.

This AI psychopath is named after the fictional character Norman Bates, the title character from the Alfred Hitchcock classic Psycho.

Norman was created by training it on violent and gruesome content like images of people dying in various circumstances (from a not specified Reddit page) before showing it a series of Rorschach inkblot tests.


Rorschach test

The Rorschach test is a psychological test in which subjects’ perceptions of inkblots are recorded and then analyzed using psychological interpretation, complex algorithms, or both. Some psychologists use this test to examine a person’s personality characteristics and emotional functioning. It has been employed to detect underlying thought disorder, especially in cases where patients are reluctant to describe their thinking processes openly.
The test is named after its creator, Swiss psychologist Hermann Rorschach.
In the 1960s, the Rorschach was the most widely used projective test.

Wikipedia

What does Norman see?

MIT researchers compared Norman’s responses with the results of ‘ regular’ artificial intelligence for Rorschach inkblots that uses the COCO dataset to respond to the inkblots

Norman consistently saw horrifying and violent images in ten different inkblots where a standard AI saw a much more normal realistic picture, much like a human see.

More inkbloats and technical info on the project’s website.


References


Related posts

Comments

This site uses Akismet to reduce spam. Learn how your comment data is processed.