Google has revealed in a blog post that the secret behind its speech recognition, image recognition in Google Photos and the recent Smart Reply feature in Inbox is now open for all. The AI engine behind these features is TensorFlow and now it’s open-source.
TensorFlow is available as a set of free tools for machine learning applications. Google’s blog explains it in technical terms as “an open source software library for numerical computation using data flow graphs.”
How does TensorFlow Work?
TensorFlow uses some very high-level mathematics and computations (non-geeky people can skip to the next paragraph). For those with some knowledge of high-level maths, the name TensorFlow might be quite familiar as it is derived from the flow of a tensor (a vector type) through a graph. The nodes of the vector are generally mathematical operations or endpoints while the edges represent relationships between the nodes. Google is licensing TensorFlow under the Apache 2.0 license for anybody who wishes to utilise the set of tools.
In simpler words, anyone whether a student, researcher, competitors or even an independent individual developer can make use of the Machine Learning engine now.
Matt Cutts, Google’s head of webspam, likes to call this “Google handing out free (open source!) smart pills worldwide by giving away TensorFlow.”
Have We been Using Products that utilize TensorFlow?
Machine Learning is a distant concept to most individuals. Even without knowledge, everyone has been using products that make use of such techniques and algorithms. Some examples of Google products utilising machine learning that we use every day are:
- Google’s search results (Both online and offline in smartphones)
- Google Photos Image recognition
- Real-time translations
- YouTube auto-thumbnails
- Google Now Voice Recognition
TensorFlow can be used in simple applications and does not require internet connectivity for its operation. It is also scalable and more advanced functionalities like bulk data recognition might require back-end servers while other applications can make use of it on mere smartphones. The applications are countless and will definitely help spawn new industries. Individual developers can benefit greatly since now they have their hands on machine learning algorithms which might have been impossible to do so without years of research.
Concluding Thoughts
The question on everyone’s minds right now would be “Why is Google giving it up for free?” According to Google, they want TensorFlow to be used everywhere and help advance machine learning as a whole. The AI will improve much quicker if it’s used by more people and making it open-source means Google gets extra hands working on it for free. In the long run, this will benefit Google and its other products. The whole process is similar to what Google did with Android and several other open-source projects.
For more technical information head to the source link below.
Source: Wired
