On Monday, Google released a new version of TensorFlow that supports iOS for the first time. TensorFlow is Google’s machine learning software, which the company open-sourced last November and used to power AlphaGo. Engineers at Google have been actively working on a version of TensorFlow at least since then, CNET points out. The new release now lets iOS developer to build neural network features directly into apps, like the ability to parse sentence structure and / or recognize people and objects in photos.
Google used TensorFlow to build its Magenta project, which aims to advance machine generated art, and recently released a 90-second piano melody created solely by a neural network. This gives an idea of the sorts of things the software can be used to build. TensorFlow version 0.9 also adds a bunch of other features including Python 3.5 support, added support for processing on MacOS GPUs, and a bunch of bug fixes. It’s not the final version of the update, which doesn’t have a release date yet, 9to5Mac points out.