If you appreciate the speed and ability of Google products to do things like translate signs on the fly or find pictures in your Google Photos library, then you may be pleased to know that capability may start to spread throughout the computer systems and apps you use on a daily basis. The possibility for this is thanks to a decision by Google to open source their TensorFlow machine learning system.
Google is moving quickly into areas dominated by artificial intelligence, with machine learning being a big piece of that. They recently revealed their RankBrain AI engine is improving search results for users and they recently updated their Inbox app with a smart reply feature that draws on the TensorFlow code.
TensorFlow is Google's latest version of machine learning code, following their DistBelief system. When Google started building TensorFlow, one of their aims was to make it more scalable and to disconnect it from a heavy reliance on Google's infrastructure. The new codebase ended up being portable and is capable of running on hardware ranging from clusters of super powerful servers to a typical smartphone.
The freedom presented by these design decisions enabled Google to open-source the TensorFlow code. In the spirit of open-source code, Google hopes that access and use by researchers, engineers and even hobbyists will result in even better machine learning capabilities in the future. For end users, that means technology will work even better than it does now.
source: Google
