Don’t assume that MIT’s resident geniuses are using their powers just for good. Instead, they are creating a psychopath AI by feeding it a bunch of violent content on Reddit. The people who brought the world the aptly-named Nightmare Machine and Shelley, the world’s first collaborative AI Horror Writer, are at it again with their new AI, Norman, named after the knife-wielding nut job at the heart of Alfred Hitchcock’s Psycho.
MIT scientists Pinar Yanardag, Manuel Cebrian, and Iyad Rahwan fed Norman content from a subreddit so dark, violent, and twisted that they wouldn’t even share its name, and then had Norman look at Rorschach-type images, interpret them, and caption them, which they explain is “a popular deep learning method of generating a textual description of an image.” To see how Norman’s responses compared to that of a standard AI, they had both caption the images. Norman’s choices are alarming to say the least. What a “normal” AI sees as a bird on a wire, Norman identifies as a man being electrocuted to death. Another image is captioned as a man with an umbrella by the standard AI, while Norman sees as “a man shot dead in front of his screaming wife.” Nice, eh? Per MIT, Norman’s psychopathic tendency “represents a case study on the dangers of Artificial Intelligence gone wrong when biased data is used in machine learning algorithms.”
It’s not even Halloween, but if you want to scare the tuna salad out of yourself, get to know Norman’s creepy world here. And good grief guys, don’t just go and prove Elon Musk right about of all this.
[H/T Newsweek]
Recognize your brand’s excellence by applying to this year’s Brands That Matter Awards before the early-rate deadline, May 3.
