How smart is your smartphone?
Until recently, not that smart, actually. Sure Siri can beatbox and Google can tell you what the fox says, but the phones themselves are not actually that smart. They're responding based on info they grab from the cloud.
For example, if you take a picture of an elephant, Android phones use Google's cloud computing to recognize it and categorize it as a picture of an elephant. But that process is taking place somewhere else.
Soon, that kind of machine learning will be able to take place right on your phone.
Google recently announced that it will be purchasing chips from Movidius, which makes low-power machine-learning processors. Machine learning is when devices can process and categorize external information -- like that picture of the elephant.
Before the processors required for machine learning were too large an energy drain for power-hungry smartphones, but Movidius' new chips purport to be very low energy, allowing the complex processes required for deep learning to take place -- without completely draining your phone's battery in minutes.
Google and Movidius have previously worked together on Project Tango, a bid to improve real time 3D mapping needed for virtual reality apps.
The benefit of putting machine learning processes directly on the phone is that the devices will be able to perform tasks that require deep learning without a strong Wifi signal -- a definite bonus if you've ever heard Siri say, "I'm sorry. I can't help you with that right now."
Speaking of Siri, Apple isn't resting on its laurels in the AI department, either. The company recently purchased an artificial intelligence company called Perceptio. Perceptio processes real-time machine-learning applications and information on mobile devices. The company also recently made two high-profile AI hires, cementing their move into the machine learning space.
The major difference between the two is that Google appears to be working towards smarter devices, rather than just smarter software -- although Apple may follow suit.
The ability to process machine learning operations on the device itself offers opportunities in many different applications, including categorizing images and audio. Google already took a major step towards this with its Google Translate app, which runs solely on the device rather than requiring an Internet connection to work (a real boon for travelers whose data plans may not follow them across borders).
With these advances, before long smartphones will be just as capable of running machine learning algorithms as any desktop computer, paving the way for smarter apps and truly smarter smartphones.