Google Open-Sources TensorFlow

Google have been exploring the field of artificial intelligence for several years now, constantly improving the ways machines work for people. Recently, the company decided to share their intelligent resources with the world and move the boundaries of the research done in the field.

Namely, on November 9^th, the web magnate decided to open-source their advanced machine learning platform TensorFlow and enable researchers outside Google to expand its possibilities. The platform integrates into dozens of Google’s products and is primarily built for the purposes of conducting research in machine learning and deep neural networks. However, as noted on the website, TensorFlow may have a variety of different other applications, which is why Google decided to let external researchers work with it. This way, its possibilities could be tested in multiple different domains, whereupon the company would get some more specific results. In relation to this, Google CEO Sundar Pichai notes:

“We hope this will let the machine learning community—everyone from academic researchers, to engineers, to hobbyists—exchange ideas much more quickly, through working code rather than just research papers.”

A historical overview of Google’s search intelligence

Google has been working on the revolutionary ideas in machine learning for years now and this is how enabled us to talk to the Google app, read signs in different languages and a lot more. It all started back in 2011, when the company created a software framework DistBelief, within which the researchers developed two algorithms for large-scale distributed learning. These algorithms were used to improve speech recognition for the Google App, as well as for making the initial steps in assigning meaning to non-verbally expressed concepts.

In the following years, we’ve seen Google improving virtually all the aspects of their services and increasing their focus on developing machine-learning systems. Most recently, Google representatives talked about testing Thought Vectors, a project bound to change the way we perceive machines and their ability to comprehend more than text, speech and code. With the latest announcement of open-sourcing their TensorFlow, all this knowledge will become available to communities of researchers, who might find different ways to implement it.

Dissecting TensorFlow

TensorFlow represents a library of files that can use data flow graphs for numerical computation. Its flexible architecture allows different deployments and researchers can use it to create systems capable of not only reading and crunching huge volumes of data, but making decisions as well. By open-sourcing this library, Google practically gives power to researchers all over the world to adjust the system to their own use and further expand its functionality.

As suggested by one of Google’s top engineers working on TensorFlow, the company is currently interested in seeing the actual results of platform implementation and this was the primary reason for open-sourcing. By having multiple institutions working with TensorFlow, Google can get much more practical results, which would be more valuable for expanding the platform’s possibilities.

“The machine learning community has been really good at polishing ideas, and that’s a really good thing, but it’s not the same thing as polishing working code associated with research ideas.”