The world of artificial intelligence and machine learning is growing by leaps and bounds. Midjourney's art AI recently won a blue ribbon in an art contest.
Google's chat AI wants a lawyer because it thinks it has become sentient, or at least some people think it has.
Are we on the verge of machines that so resemble human thought that they might as well be human? I don't know, but what I do know is that if the age of the machines is dawning, the last thing we want to do is train them to be psychopathic killers, which is exactly what some researchers at MIT just did.
They created an AI called Norman after Alfred Hitchcock's character Norman Bates in Psycho: you know, the one that dresses up like his dead mother and murders the girl in the shower.
Then the scientists fed Norman a steady stream of violent subreddits to "learn." When it came time to understand what Norman was learning, they gave the bot some Rorschach inkblot tests to interpret, and compared Norman's interpretations to those of AIs that did not hang out in the worst areas of Reddit.
Where the normal AIs saw a vase of flowers, Norman saw a man being shot to death. Where the normal AIs saw a couple standing together, Norman saw a pregnant woman falling to her death off a construction site…you get the idea.
The researchers said the AI
"suffered from extended exposure to the darkest corners of Reddit, [illustrating] the dangers of Artificial Intelligence gone wrong when biased data is used in machine learning algorithms."
And this was the purpose of the experiment. Algorithms are only as good as the information that is fed to them, which explains why AIs let loose on the internet often end up with some terrible ideas about people.
Almost all information is created by sinful people, it seems the machines, learning from us, reinforce and magnify that sinfulness the longer they study us.
P.S. Now check out our latest video 👇
