As the new science fiction film "Ex Machina" puts it, artificial intelligence (AI) could be "the greatest scientific event in the history of man."
But tech giants like Bill Gates and Elon Musk have also warned that AI is one of the greatest existential risks to humanity if we don't research and develop it in a responsible way.
And according to Stuart Russell, a professor of computer science and engineering who appeared on an AI-themed episode of NPR's Science Friday, scientists really haven't been thinking about the safety concerns of AI at all until very recently.
That's because we didn't understand the nature of the problem, Russell said. AI development so far has been a race to create better and smarter models.
"We were focused on making machines smart because dumb machines are not very useful," Russell said. "It wasn't clear to most people in the field why really smart wouldn't be really good.”
Now more and more people are raising concerns about the future of AI development because these technologies are becoming increasingly sophisticated and more ubiquitous, Eric Horvitz, managing director of the Microsoft Research Lab, said during the episode. We plan our trips with GPS that came out of an AI algorithm, our smartphones can understand our speech, and Facebook can recognize our faces in the photos we post. AI is creeping into our lives and it's finally getting people thinking about where this is all going, Horvitz said.
And a big concern is what would happen if we create a robot that's smarter than us.
"The smarter machines get, the more careful we have to be to make sure that the objectives we give them are exactly aligned with what we want,” Russell said. "If you don't give them the right instructions, if you don't give them objectives that are perfectly aligned with what the human race wants, then you have a problem," Russell said.
That's not easy though because the human race isn't very good at defining what we want in the first place, Russell said. We have no idea what kind of loopholes we might leave in our instructions. And if we create a machine capable of learning so much that it eventually becomes smarter than us, then it can process more information than us, it can look farther ahead than us, and it can anticipate all of our counter moves, Russell said.
That's why it's crucial to start researching and testing out safety measures for AI.
Physicist Max Tegmark is trying to kickstart that research with the new Future of Life Institute.
"I'm a pretty cheerful guy, so to me the interesting question isn’t how worried we should be, but rather figuring out what we can do to help and then actually start doing it,” Tegmark said.
The goal isn't to hit the brakes on AI research and development. We just need to start doing it better and more responsibly, Tegmark said.
Musk recently donated $10 million to the Future of Life Institute to help fund AI safety research.
You can listen to full Science Friday episode below:
SEE ALSO: Who's set to make money from the coming artificial intelligence boom?