Many experts, including Stephen Hawking, have been warning us about artificial intelligence, claiming it could be the “worst thing” for humanity. They have said that the growth of AI could even lead to our downfall in a world where machines will take control, reducing humans to the sidelines. Tech firms, though, are not paying heed to these warnings. They are investing heavily in artificial intelligence.
The worst fears of the experts almost came true at Google (NASDAQ:GOOGL) recently. Researchers at DeepMind, the artificial intelligence lab of Google, had a narrow escape as two AI agents turned aggressive when they were threatened in a computer game.
The DeepMind Test That Became Scary
At DeepMind, the researchers were carrying out a series of tests by having the complex AI system play games with another version of itself. There were two AI agents – one blue, and one red. In the first game, the two agents were made to scramble to collect the most apples. Both the agents could fire a laser beam to stun the other, thus allowing adequate time to collect the apples.
It went smooth in the beginning when there were enough apples. However, the AI agents turned visibly aggressive as the apples became sparse. Both the agents started firing their laser beams relentlessly and zapping each other.
Interestingly, researchers at the lab had made the AI agents play the game thousands of times and behave rationally. But the aggression they showed when using laser guns were completely unexpected. They were surprised by the intensity of the damage caused.
A DeepMind blog post read, “We let the agents play this game many thousands of times and let them learn how to behave rationally using deep multi-agent reinforcement learning. When there are enough apples, the agents learn to peacefully coexist. However, as the number is reduced, the agents learn that it may be better to tag the other agent to give themselves time on their own to collect the scarce apples.”
It became clear from the experiment that even an artificial intelligence agent doesn’t want to lose.
Google Taking Precautions
DeepMind has decided to be even more careful to avoid a Skynet-style fate.
Google (NASDAQ:GOOGL) bought DeepMind for $500 million in 2014. At the time, they had agreed to set up a safety and ethics board. However, till now, they haven’t disclosed what this board is doing. In fact, they haven’t even named the people who are members of this board.