Google Clips camera may be a few creepy, but it also appears pretty valuable for the best user; expanding computer learning to catch the genuine images of your children and pets automatically. But the answer to that functionality isn’t really Google’s AI courage; it also needs a specific processor developed by the Intel-owned chipmaker Movidius.
The chip in issue is the Myriad 2, which Movidius reports as a “visual processing unit” or VPU. (That’s as presented to a graphics processing unit, GPU; or primary processing unit, CPU.) The Myriad 2 is a processor tailor-made to manage computer perception tasks like object identification, and Movidius maintains it’s the “industry’s leading always-on vision processor.” It’s earlier shown up in Google’s Project Tango projects as well as DJI’s self-sufficient drones, and advocates to execute their onboard concept processing more effective.
Google has extended been involved in Movidius’ chips. As well as practicing their VPUs to control Project Tango, the search monster started on a combination with Movidius last year to promote how image identification controls on media like smartphones. With the release of the Clips camera, we have the best sample of the sorts of advantages these combinations bring.
Clips do it everything its AI processing on-media preferably than relying on an association to the cloud to look pictures for common faces. That’s great for customer secrecy (there’s no possibility of data being snaffled in-transit), but also improves the media’s battery life (because it doesn’t have to control an web connection at all points). These advantages are the primary result of practicing a specific chip like Movidius’ VPU.
More and more businesses are switching to technology like this to promote on-device AI. Just last month Apple revealed latest iPhones full with dedicated AI “neural powerhouse” processors, and Huawei dispensed off similar capacity with its new Kirin 970 chipset. On-device AI is the tomorrow, and specific silicon is supporting deliver it.