Technology > IT

A concept of an artificial intelligence algorithm. / Photo by: Getty Images

 

In what may be a first in the world, a psychopathic artificial intelligence algorithm named after the killer in the suspense thriller, Psycho was created by researchers from the Massachusetts Institute of Technology who exposed the algorithm to violent images found on Reddit. The researchers admitted that the algorithm had been exposed to the darkest corners of Reddit for an extended period of time. 

When the algorithm was subjected to a series of Rorschach inkblot tests, it interpreted the inkblots as either electrocutions or speeding car deaths. The same inkblots were perceived to be either umbrellas or wedding cakes by a standard AI, according to Benjamin Fearnow, writing for Newsweek magazine. 

Disturbing and creepy were some of the words used in describing some of the answers given by Norman the algorithm. It interpreted a Rorschach inkblot as a man being shot dead while the standard AI saw it as a closeup of a vase with flowers. Another inkblot was about a man being shot in front of his screaming wife, which the standard AI judged it as a person holding an umbrella in the air. The most demented reply that Norman the algorithm gave was that of a  pregnant woman falling from a construction site, while the standard AI deemed it as a tender moment of a couple standing together. 

The researchers said the aim of their research project is to prove that the method used in teaching a machine to learn an algorithm will influence its behavior afterward. They insist that algorithms should not be accused of being biased or unfair because the blame lies on the biased data that was fed into these algorithms. The research was done in the same lab that had produced other weird robots such as Shelley which can write horror stories and the Nightmare Machine which can produce frightening images.