SwiftKey Paves Way For Apps With Locally Run Neural Networks

[Lucian Armasu]

The SwiftKey keyboard application is now one of the first apps to run its neural network locally on the device, where it can present the results on the fly, while preserving the users' privacy.

SwiftKey Paves Way For Apps With Locally Run Neural Networks : Read more

 

[ZolaIII]

Predicting common tipe of predicate to the already known subject on non changed relationships is more of a simple rule than a AI. This still lags for behind even most basic deductive categoric decision chain making (syllogism). Even it is possible on the preliminary main lv to be able to be done the active future induction will remain problematic along with special cases & relations.

 

[clifftam]

I always love using Swift on my Android phone. It is highly predictable of the way I typed. I am not sure whether this neural learning method is any better. Not to say it is not but the Swift app can now already predict two words at a time for me (prior to the update).

 

[rantoc]

Can understand the interest from the industry... and its a lot about mapping your interests, friend etc so they can be sold why do you think Google/Facebook are so extremely interested...

 

[hdmark]

when it says "locally run" im a little confused. is this a NN thats currently learning and adapting on your phone? or is this a NN thats been trained by swiftkey and then installed on your phone ? or a hybrid where they give you a decently trained system and then it retrains based on what you personally feed it?

 

[grimfox]

Predicting common tipe of predicate to the already known subject on non changed relationships is more of a simple rule than a AI. This still lags for behind even most basic deductive categoric decision chain making (syllogism). Even it is possible on the preliminary main lv to be able to be done the active future induction will remain problematic along with special cases & relations.

Considering that swiftkey is only working with "simple" processors at the moment and more complex reasoning would be taxing/result in slower responses, that they can achieve better results than what they were getting is a win in my book. In the future when stronger or dedicated neural style hardware is available they will be ahead of the curve. Or they'll get crushed by a startup.

I'm just happy that someone is leveraging these tools to increase privacy. Of course I'm sure that while my typing patterns are securely processed locally it's not stopping Google or Cortana from scanning the finished text and sending that data off to non-local servers for processing.

 

[bit_user]

when it says "locally run" im a little confused. is this a NN thats currently learning and adapting on your phone? or is this a NN thats been trained by swiftkey and then installed on your phone ? or a hybrid where they give you a decently trained system and then it retrains based on what you personally feed it?

I'm sure it's pre-trained. Presumably, it then adapts based on your input. Then, the question is whether the adaptation occurs on their cloud or your phone.

I think the best solution would be for the adaptation to occur on your phone (cloud computing is cheap, but it ain't free), while it's plugged in to an external power source. This would also protect your privacy. Of course, your input would at least need to be stored on your phone, which potentially creates a vulnerability.