Google partners with Movidius to bring machine learning to future mobile devices
Google will buy low-power machine learning chips and software from Movidius, and contribute to the company's research.
A new partnership between Google and chip maker Movidius could change the way your phone looks at the world.
According to Movidius, Google will be licensing Movidius' latest flagship chip, the MA2450, along with the accompanying software stack.
The chip has 12 cores, and is able to perform low power, advanced computer vision processing. One possible use for this could be Google’s Project Tango. Google is partnering with Lenovo on a smartphone that would be able to sense where it is, how it's positioned, and what’s around it, which opens the door for many new software opportunities.
Movidius says it specializes in “low-power machine vision for connected devices” which makes us believe it’s be suited for phones or other mobile devices instead of Google’s driverless cars.
In essence, technology like facial and spatial recognition and being able to interpret and learn from this data is the type of thing Google would love to do with its forthcoming devices.
During a demonstration at CES, Google’s Johnny Lee showed off how you could measure a room using a prototype Project Tango tablet and then shop at Lowes for furniture that would fit.
Why this matters: Google is big on machine learning. In essence, the concept is that computers and software should get smarter over time—forming relationships between large sets of data and contextualizing information. An everyday example is Google Now, which anticipates your schedule, commute, and other details about you.
Google is going big into hardware, as it will surely have another Nexus phone this fall, and recently posted several jobs for a new VR division. This partnership includes Google agreeing to contribute to the Movidius neural network technology roadmap, which should benefit both companies.
For comprehensive coverage of the Android ecosystem, visit Greenbot.com.