Science
Scientists have created a terrifying “psychopathic machine” named Norman by displaying scary photos of her that are banned on the Internet.
This artificial intelligence (AI) was named after one of the main characters in Hitchcock's Psycho film, Norman Bates.
AI training is based on terrible photos of tragic events that were downloaded from a banned group on a social network.
In the same experiment, scientists also used another “normal” artificial intelligence named Norman. His training was based on pictures of cats, birds and people.
Both AIs were shown ink stains from the Rorschach test, which psychologists commonly use to better understand a person's state of mind.
About artificial intelligence
The goal was to understand how these two artificial intelligences, trained in completely opposite images, would be able to explain what they see in the images.
In one of the pictures, “normal” Norman saw an image of a wedding cake on the table. At the same time, the “sinister” Norman in the same picture saw “a man hit by a speeding driver”.
Read also: How artificial intelligence affects our lives
The psychopath algorithm for artificial intelligence was created by a team of programmers from the Massachusetts Institute of Technology. Their goal was to find out how an artificial intelligence that collected information from the “dark corners of the Internet” would see the world.
In this picture, the “normal” Norman saw “a man holding an umbrella” and the “other” Norman saw “a man who was shot in front of his screaming wife.”
In another spot from the Rorschach test, “normal” artificial intelligence saw a “black and white photograph of a little bird”, while “manic” AI saw “a person being dragged into a car on dough processing.”
Read also: The 10 most frightening theories known to man
Where a “normal” AI saw a “flower vase”, “another” Norman saw a “man shot”.
The problem of artificial intelligence
Scientists wanted to show how dangerous artificial intelligence is if machine learning algorithms use harmful data.
The results showed that if you teach AI to be “bad”, it will also end up being bad.
In this picture, “normal” Norman saw “a couple of people standing next to each other” and “bad” Norman saw “a man jumping out a window.”
In this picture, “normal” Norman saw a “black and white baseball glove”, and “psychotic” Norman saw “a man shot by a machine gun in broad daylight.”
In this ink stain, “normal” Norman saw “a black and white photo with a red and white umbrella”, and “crazy” Norman saw “a man electrocuted while he was walking around a crowded street.”
