When Google began to emphasize its artificial intelligence, neural networks, and deep learning technology, the idea that it would eventually be declared open source took root in the back of our minds. Early this week, that hunch became reality: With an announcement from Sundar Pichai himself, Google confirmed the opening of its TensorFlow platform, a critical piece of advanced services like Gmail, Google Now, and Google Photos.
Every word you choose to translate with Google Translate, every personalized request in Google Photos, every email unearthed from your Gmail inbox, and every semantic search in Google Search activates much more than it appears. Granted: Sometimes the results are pretty bad, and no one denies that in Mountain View. But what we should recognize is that the overall accuracy of these and other Google services rises over time. Many users on the Web even went so far as to say that Google was turning into Skynet. Well, what happens if 'Skynet' becomes open source…?
One of the keys behind all of Google's online services is TensorFlow, a software library designed for the development of artificial intelligence, training neural networks, and improvements in deep learning. Why is opening TensorFlow so important? The reason is simple: Google could have made a fortune with a licensing scheme for this technology. Obviously, the benefits that may come through the community with optimizations to its code are more important than a specific sum. Of course, this kind of announcement always carries an asterisk at the end, and in TensorFlow's case, Google will share only a part of the platform, and 'some' of the algorithms that run on it, in addition to being compatible with a single system (which can have multiple GPUs).
The TensorFlow project as open source software will be managed by Google through its official site (link below), and the chosen license is Apache 2.0, very common among Mountain View's developments. Another detail we can't overlook is the theoretical integration of TensorFlow with similar platforms, such as Torch and Caffe. In fact, that combination could give rise to a new platform that applies 'the best of both worlds', and to significant advances in artificial intelligence.
