Google Clips Uses Intel’s Movidius Myriad 2 Vision Processing Unit

Google Clips Uses Intel’s Movidius Myriad 2 Vision Processing Unit

Google had a number of new hardware devices to show off at the company’s launch event earlier this week. We saw the official reveal of the two new Pixel smartphones, a smaller version of the Google Home and in a surprising reveal, an AI-based camera called Google Clips. It’s been confirmed that Google is using a vision processing unit from Intel called the Movidius Myriad 2 chip, that does all of the artificial intelligence processing directly on the device.

Google has been working on artificial intelligence via machine learning and neural networks for a couple of years now. We’ve seen their results trickle down into a number of their staple products and services including Gboard, Google Translate, Inbox by Gmail and more. SoCs from Qualcomm, Huawei and others are starting to incorporate some dedicated hardware for AI on their chips, but that’s not anywhere close to what Google is able to do in the cloud.

This is where most of Google’s machine learning computational work is done too – in the cloud. So it was a surprise to many when the company unveiled the Google Clips AI camera and told the world the recordings and photos would never leave the device itself unless the user agreed to upload it to Google Photos. This is even more impressive due to the artificial intelligence capabilities that the company has built into the Google Clips camera, on-device.

Many wondered how Google was able to do this all on the device itself and Intel has confirmed that it’s thanks to their Movidius Myriad 2 vision processing unit (VPU). Remi El-Ouazzane, vice president and general manager of Movidius at Intel New Technology Group, says it is remarkable to see what the team at Google has been able to do in such a small device like Google Clips. He feels “this intelligent camera truly represents the level of onboard intelligence we dreamed of when developing our Myriad VPU technology.”


Source: Intel Newsroom

Discuss This Story

You might also like