Google will include an additional processing chip in its future mobile devices (such as the Nexus line of smartphones and tablets) to enable "deep learning" applications without (necessarily) communicating to a central server:
Google has signed a deal with Movidius to include its Myriad 2 MA2450 processor in future devices. The search giant first worked with Movidius back in 2014 for its Project Tango devices, and it's now licensing the company's latest tech to "accelerate the adoption of deep learning within mobile devices."
[...] More recently, Google managed to cram a [neural] network into its Translate app, allowing users to convert the text in images on the fly. And SwiftKey also runs a small-scale network for word predictions in its SwiftKey Neural application. But all these applications require a large amount of processing power for what are relatively inane tasks. That's where Movidius' chip comes in.
The Myriad 2 MA2450 is referred to as a "vision processing unit." It's really got a single purpose: image recognition. The architecture has very little in common with a traditional CPU, and it's designed specifically to handle the myriad (get it) simultaneous processes involved in neural networks. As such, its power draw when, for example, recognising a face or an image, is much, much lower than doing the same task with a Snapdragon processor. As for how exactly will Google utilize the chips, that's something we're unlikely to know until it's ready to announce devices.
TechCrunch has some additional details about Movidius MA2150 and MA2450 chips. Or look at this product brief (PDF).
(Score: 2) by mhajicek on Sunday January 31 2016, @05:41AM
Get every customer to pay for a node in your global distributed neural net.
The spacelike surfaces of time foliations can have a cusp at the surface of discontinuity. - P. Hajicek
(Score: 2) by SubiculumHammer on Sunday January 31 2016, @06:06AM
Great so every app can profile me with deep learning on my battery.
(Score: 1, Insightful) by Anonymous Coward on Sunday January 31 2016, @10:44AM
It rhymes with "insidious".
(Score: 1, Insightful) by Anonymous Coward on Sunday January 31 2016, @01:45PM
"As for how exactly will Google utilize the chips, that's something we're unlikely to know"
Simple. It will allow everyone's phone to do the facial image recognition work for the NSA remotely which will vastly reduce the required data transfer bandwidth required for such spying which will make it that much easier to do without the bandwidth being noticed. "Smart"Phone indeed..
(Score: 4, Interesting) by Hyperturtle on Sunday January 31 2016, @05:28PM
Yes, I too would like to know how to disable this physically (although you didn't quite say that)-- just like how, theoretically, one could disable a TPM chip in any personal hardware one wishes to keep secured from remote entities with potential remote adminstrative capabilities.
Just think -- the technology can be licensable AND work with Windows Hello! Log into your windows whatever with your face, and get this to securely transmit it according to the EULA you agreed to, and the online-only privacy policy subject to change without notification to you. It's to streamline efficiencies and is done purely in the best interests of our consumers. Of course. You're the product, but the leash was free.
The chip will cost money of course, so you have to pay for the collar anyway. It'll seem to be a convenience to most people, though, because cloud safe security or something...
Most people still don't seem to recognize that when the personal data is stolen nowadays, it's not because they left their front door open on their PC or phone. It's the OS and the apps--no faceprint or fingerprint or password is going to protect anything except for keeping a kid brother from sending love notes to his older sister's contacts... or also protect people from recovering their own data when trying to read off a storage device like in the old days. Now, in many cases, it's encrypted or simply not stored locally any more. This won't usher in an era of privacy or security for anyone--not if it does more of the same to learn more about people and their friends.
It seems safe to believe that anything that hypes your security and privacy doesn't really mean "for the user" unless they specifically spell it out stating it as such and how it pertains to your data and what's done with it.
Anyway, I am sure the onhub internet of things will have one, and all their little sensors will tie into it and make educated guesses about the activities about all of the other google mobile products floating around the house, and in the vicinity of other onhubs in the neighborhood and in public places and stuff. A tinfoil travel bag is going to be about the best you can do to keep private and keep your phone on you if you drive around and don't want to be a record in all of the database entry points you drive past.
Of course that doesn't prevent other methods (cell tower/carrier specific, license plate photos, embedded car stuff), but it would prevent some of google's more invasive conveniences. Any privacy is better than none, I suppose. If anyone finds a way to opt out of being deep learned, I hope they share it.
(Score: 2) by Freeman on Monday February 01 2016, @08:51PM
The solution is simple. Vote with your wallet. Assuming, that's not an option for you for various reason, then that's the price you pay for doing what you are doing.
Joshua 1:9 "Be strong and of a good courage; be not afraid, neither be thou dismayed: for the Lord thy God is with thee"
(Score: 0) by Anonymous Coward on Sunday January 31 2016, @09:50PM
Funny how people used to believe Google's 'do no evil' line. Nowadays it's assumed the opposite will occur.
I guess that's what money does.
(Score: 0) by Anonymous Coward on Monday February 01 2016, @01:33AM
Nonsense. Google won't turn the data over to the NSA. They'll develop the profile, and lease access/queries based on the profile to the NSA. And to anyone else who is interested.
(Score: 2, Interesting) by Anonymous Coward on Sunday January 31 2016, @07:01PM
Given the wonderful capabilities of those Movidius chips, it sure would be nice if you could lay your hands on them directly, or at least a development version of a Project Tango device.
Yet this is apparently totally impossible -- at least for a EU citizen. Have any soylentils been able to test these chips/devices, and if so, what were/are your experiences with them?
(Score: 2) by darkfeline on Sunday January 31 2016, @10:24PM
The interesting thing about neural networks is that you don't actually know what's going on in them. You set up the network, you train it, and it works! But how exactly it works is a mystery.
So we deploy these chip on a global scale, and they work! Machine learning on the go! Except no one know what they're actually doing. You see where this is going, I hope.
Sufficient to say, I welcome our future deep learning overlords with open arms.
Join the SDF Public Access UNIX System today!