[ad_1]
The Pentagon's state-of-the-art research agency has laid the foundation for the Internet, stealth aircraft and autonomous cars. Now, it goes a lot on artificial intelligence.
At its 60th anniversary conference on Friday, DARPA announced a $ 2 billion investment to advance the AI border.
"We think it's a good time to sow the field of AI," said CNNMoney John Everett, deputy director of DARPA's Office of Information Innovation. . "We believe we can accelerate two decades of progress in five years."
Artificial intelligence, which allows machines to perform tasks traditionally performed by humans, is a hot topic in tech 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 having a completely human sound.
Breakthroughs in the last decade have inspired companies to recruit the best AI talent outside academia. Machines are now much more accurate to recognize speech, understand images and process words, leading to products such as Amazon's Alexa, Apple's Siri and Waymo's standalone vans.
The largest and most innovative companies in the country rely on them to stay ahead of their competitors. Waymo's autonomous vehicles have traveled more than 9 million kilometers on American roads thanks to artificial intelligence.
National governments, such as Canada, China, India and France, now give priority to AI. They consider artificial intelligence as essential to the growth of their economies in the 21st century. China has declared that it wants to become the world leader by 2030.
The DARPA investment will focus on creating systems of common sense, contextual awareness and improved energy efficiency. Progress could help the government automate security clearances, accredit software systems, and make AI systems workable.
But the industry also has a problem of hype. Machine learning relies on algorithms that learn huge data sets. A computer can display millions of photos of cats and, over time, it will recognize when another photo includes a cat. But these AI systems often require thousands of computer processing chips for weeks before learning something.
"Machine learning is remarkably inefficient," said Everett. "He can do incredible things, but it's also remarkable what he can not do."
For example, a new wave of artificial intelligence is needed to support advanced home robots. Ask a robot to "pick up the living room" and he will not know what that means, Everett said. It will be difficult to identify which items need to be picked up and which ones are not.
DARPA wants to adopt AI methods similar to those of humans. And sometimes, a human can learn something by looking at a single example.
The agency could ultimately invest even more money in the development of AI, said Everett.
"If we get positive results and that they are important, and that they are relevant to military and national security, we will not stop," he said.
CNNMoney (Washington) First published on September 7, 2018: 12:21 ET
Source link
