TIL that in 2018, an AI developed by MIT was intentionally trained using only dark, disturbing content to see if it would develop "psychopathic" tendencies. Named "Norman" after Norman Bates, the AI then interpreted neutral images as violent scenes, highlighting how data shapes AI behavior

https://en.wikipedia.org/wiki/Norman_(artificial_intelligence)
#til #todayilearned
https://www.reddit.com/r/todayilearned/comments/1gp907o/til_that_in_2018_an_ai_developed_by_mit_was/