'Psychopath' AI Created By MIT Using The Dark Side Of Reddit - Latest Hacking News

When the Microsoft AI chatbot Tay was turned into a racist and genocidal application the company had to shut down the chatbot within 24 hours and so it is more than speculation to presume that an AI can be turned'evil'. A group of researchers from the Massachusetts Institute of Technology (MIT) has taken this one step further and corrupted an AI by using some biased data and whom they decided to name Norman a homage to the villain in Alfred Hitchcock's psycho. The team pulled the images and many quotes from an infamous subreddit which is dedicated to death and violence, in doing so they have incorporated them into how objects could be described. The team then gave the images of Rorschach inkblots to analyse the AIs responses and also compared them to the standard image captioning neural network, the results of which were very alarming. For example when the AI interpreted an open umbrella Norman saw a wife screaming in grief as her husband died.