Because we might as well get this robot apocalypse started, researchers at MIT have developed an artificial intelligence image-captioning psychopath named Norman, presumably after the Cheers character. "And not Norman Bates from Psycho?" Never heard of it. Some more info while I write these researchers an angry letter, probably filled with glitter so it gets everywhere and drives them crazy:
Norman is an AI that is trained to perform image captioning; a popular deep learning method of generating a textual description of an image. We trained Norman on image captions...dedicated to...the reality of death. Then, we compared Norman's responses with a standard image captioning neural network (trained on MSCOCO dataset) on Rorschach inkblots; a test that is used to detect underlying thought disorders.
Norman was not corrupted to make any sort of point about human psychology on the internet -- a neural network is a blank slate. It doesn't have any innate desires like a human. What Norman does address is the danger that artificial intelligence can become dangerously biased. With AI, you get out what you put in, so it's important that these platforms are trained to avoid bias
When reached for comment about the results of their little experiment, none of the researchers involved could be located, presumably because Norman was already wearing their skin like a tuxedo to become more human. And -- what's that? "A martini." Hoho, looks like he watched a couple James Bond movies too.
Keep going for the comparison of ten of Norman's ink-blot evaluations vs that of a 'normal' AI, none of which left me feeling warm and fuzzy inside.
Thanks to my dad, who went to Harvard and MIT, both in the same day.