By Matt McFarland
The Pentagon’s high-tech research agency laid the groundwork for the Internet, stealth aircraft and self-driving cars. Now, it’s going big on artificial intelligence.
At its 60th anniversary conference on Friday, DARPA announced a $2 billion investment to push the frontier of AI forward.
“We think it’s a good time to seed the field of AI,” John Everett, the deputy director of DARPA’s Information Innovation Office, told CNNMoney. “We think we can accelerate two decades of progress into five years.”
Artificial intelligence, which lets machines perform tasks traditionally done by humans, is a trendy topic in technology and business circles. For example, Google recently delighted and alarmed observers when it showed how an AI system could call a restaurant and book a reservation while sounding entirely human.
Breakthroughs in the last decade have inspired companies to recruit top AI talent away from academia. Machines are now much more accurate at recognizing speech, understanding images and processing words, leading to products such as Amazon’s Alexa, Apple’s Siri, and Waymo’s self-driving vans.
The country’s biggest and most innovative companies rely on it to stay ahead of competitors. Waymo’s autonomous vehicles have driven more than 9 million miles on US roads thanks to artificial intelligence.
National governments, such as Canada, China, India and France, are prioritizing AI now too. They view artificial intelligence as essential to growing their economies in the 21st century. Most notably, China has said it wants to be the global leader by 2030.
DARPA’s investment will focus on creating systems with common sense, contextual awareness and better energy efficiency. Advances could help the government automate security clearances, accredit software systems and make AI systems that explain themselves.
But the industry has a hype problem, too. Machine learning relies on algorithms that learn from huge data sets. A computer might be shown millions of pictures of cats, and over time, it will recognize when another photo includes a cat. But these AI systems often require thousands of computer chips processing data for weeks before they learn something.
“Machine learning is remarkably inefficient,” Everett said. “It can do amazing things, but it’s also remarkable what it can’t do.”
For example, a next wave of artificial intelligence is required to support advanced in-home robots. Ask a robot to “pick up the living room,” and it won’t know what that means, Everett said. It will struggle identifying which items need to be picked up, and which don’t.
DARPA wants to embrace AI methods that are similar to how humans learn. And sometimes, a human can learn something from watching a single example.
The agency may ultimately invest even more money into the development of AI, Everett said.
“If we get positive results and they’re important, and they’re relevant to the military and national security, we’re not going to stop,” he said.
