24 May 2017
Google rolled out the first version of its artificial intelligence engine TensorFlow back in 2015 when it open sourced the technology. Earlier this year the same was integrated inside Google’s IoT platform Android Things in its second developer preview.
At the Google I/O event a few hours ago, Google’s VP of Engineering, David Burke revealed the TensorFlow Lite which is a skinned version of the original TensorFlow designed to be used on Android smartphones. It only makes sense that the company is now opting for ways to use the specialized framework in a quicker way considering the number of AI linked services that currently run on Android.
The company is planning to open source its work and release an API for the TensorFlow Lite anytime this year and it remains to be seen how Google is going to manage intense computation and training on a regular smartphone. If that works, we can expect a device to not depend on the internet or the cloud which would result in a more reliable performance regardless of the environment. This would also mean that Google might soon come up with the required hardware in smartphones to run TensorFlow smoothly without having to rely on a workaround.