CNET también está disponible en español.

Ir a español

Don't show this again

Christmas Gift Guide
Computers

Intel says its computer chips will accelerate AI revolution

Artificial intelligence lets phones understand your voice and Google spot your pets. Nervana chips will make AI 100 times faster, Intel says.

Intel CEO Brian Krzanich

Intel CEO Brian Krzanich speaks at a 2016 AI event.

Stephen Shankland/CNET

Intel might be an old-school computing company, but the chipmaker thinks the latest trends in artificial intelligence will keep it an important part of your high-tech life.

AI technology called machine learning today is instrumental to taking good photos, translating languages, recognizing your friends on Facebook, delivering search results, screening out spam and many other chores. It usually uses an approach called neural networks that works something like a human brain, not a sequence of if-this-then-that steps as in traditional computing.

Lots of companies, including Apple, Google, Qualcomm and Nvidia, are designing chips to accelerate this sort of work. But on Tuesday, Intel Chief Executive Brian Krzanich promised a major speed increase through its own chips, designed from the ground up for neural processing.

"We have a goal by 2020 to be 100 times better than the best AI systems out there today using this neural processing," Krzanich said at the Wall Street Journal D.Live conference.

Intel faces plenty of business challenges. The PC market is flat, Intel failed to secure a place inside mobile devices, and its server stronghold is under threat as Qualcomm tries challenging Intel's stronghold in data center technology. That's why Intel is betting on new directions like chips for self-driving cars, superfast 5G networking and artificial intelligence.

Intel will ship its Nervana chip for neural network-based artificial intelligence computing this year.

Intel will ship its Nervana chip for neural network-based artificial intelligence computing this year.

Intel

Its products in the AI market include the Movidius Myriad X chip, geared for running computer vision and other machine-learning tasks, and the Nervana Neural Network Processor, which is aimed at data centers that train AI systems so they can later be used in the real world. That training process, called inference, is extremely computationally intense, and Intel rival Nvidia leads the market today.

Intel said last year its Nervana chip should arrive in 2017, and Krzanich said Tuesday it will meet that deadline. "The potential benefits from self-learning chips are limitless as these types of devices can learn to perform the most complex cognitive tasks, such as interpreting critical cardiac rhythms, detecting anomalies to prevent cyberhacking and composing music," Intel said in a statement. Intel acquired the Nervana startup in 2016.

Facebook is evaluating the first Nervana chips, Krzanich said. "This is the first piece of silicon. We have a whole family of chips planned. Think of a new chip every year," Krzanich said. "They're helping us think about where is that going."

Krzanich is particularly excited about AI's uses in medicine.

"Where we are starting to see it break through that's really exciting is healthcare," for example using machine learning to detect Alzheimer's disease by looking at MRI scans, Krzanich said. "If it can be detected five to ten years earlier, we can profoundly affect people."

And Intel itself relies on AI and machine learning for its own chipmaking business.

"Intel is one of the largest manufacturers in the world. We produce about a million chips a day. For each one of those million chips, we take about 1.6 million pictures as it progresses through the production line," he said. When there's a defect, "We now use AI ... to go through those 1.6 million images. What used to take three weeks now takes a couple hours."

The Smartest Stuff: Innovators are thinking up new ways to make you, and the things around you, smarter.

iHate: CNET looks at how intolerance is taking over the internet.