The artificial intelligence (AI) currently being developed is largely benevolent. It can mimic the way humans think, complete menial and repetitive tasks, and more. But that doesn’t prevent people from being afraid of AI, thinking it will take away jobs or eventually turn Terminator into a documentary.
Somebody thought AI wasn’t scary enough, and did something to change that. Researchers from MIT and Australia’s CSIRO have created AI that actively warps pictures into scary nightmare fuel.
But they then fed it just one picture of a zombie, and changed the code so the zombie face would have more weight in the images generated. Soon after, the algorithm was generating nightmare fuel like crazy.
Creating sleepless nights is not the main reason for the Nightmare Machine. It has a higher purpose: to tell machines how to comfort us. When it learns what images or things scare humans, AI can then be taught to do the opposite, and generate behavior that makes us feel safe.
“Just like a child, or an adult, by learning behavior that upsets humans, a machine can then be trained to avoid that behavior,” says Manuel Cebrian Ramos, of the CSIRO, in a statement. “So the same technology we are using in this silly project could actually be used to comfort, to invite humans to co-operate with machines.”
In fact, he believes that this is the best way to create benign AI, as opposed hard wiring rules. “Instead of [Asimov’s] top-down rules, which are always going to have loopholes, it’s better for a machine to learn bottom up,” says Cebrian.
Essentially, if machines can help us understand them, then it’s going to be easier to work with them. “If we perceive them as alien, as too different from us, then we will fight them, and I don’t like that, I like co-operation,” he says.