This is the first psychopath AI in history. The MIT Media Lab’s dubbed it AI Norman – named for Norman Bates, the murderer from Alfred Hitchcock’s Psycho. And it is just scary as it sounds!
A team of MIT employees took a normal image-captioning AI – designed to look at pictures and provide a written description of what it sees – and fed it a steady stream of images from an unnamed Reddit board where people exclusively post horrifying, morbid images of murder and death.
Afterward, the team showed AI Norman a series of Rorschach inkblots, which are used by psychiatrists and psychoanalysts to judge a patient’s mental state.
The team then compared Norman’s captions to a normal AI that had not been traumatized with images of death and found a disturbing pattern.
For example, Norman captioned one inkblot “man is murdered by machine gun in broad daylight“, while the other AI captioned the same image, “a black and white photo of a baseball glove.”
Norman’s morbidity plays out again and again in the tests.
A colorful inkblot that looks like “a black and white photo of a red and white umbrella” to the vanilla AI look like “man gets electrocuted while attempting to cross busy street” to Norman.
“A close up of a wedding cake on a table“? Nope, that’s “man killed by speeding driver.”
It might seem disingenuous to give an AI like Norman nothing but grisly images for reference and act surprised when it sees nothing but murder everywhere it looks, but this little experiment shows how an AI’s machine learning process can internalize biases and end up with warped perceptions of the world.
Luckily for Norman, humans can counter-balance his morbid outlook by taking the Rorschach tests themselves and allowing him to learn from their answers.
You remember the brainwashing scene in A Clockwork Orange? Very close and terrifying, no?