Anyone who has attempted Google Photos would agree that this free photograph stockpiling and management service from Google is keen. It packs in different great features like advanced search, ability to categorize your photos by areas and dates, consequently make albums and videos in view of similitudes, and walk you down the memory lane by viewing you photographs of the same day several years ago. There are numerous things Google Photos can do that quite a while back would be machine-ly unimaginable. Google Photos is one of the many “brilliant” services from Google that utilizes a machine learning innovation called TensorFlow. The word learning demonstrates that the innovation will get smarter by time to the point that our present knowledge can’t envision.

What is TensorFlow?

TensorFlow is Google’s open-source and capable artificial intelligence software, which powers many services and activities from Google. It is the 2nd generation of a system for expansive scale machine learning executions, worked by the Google Brain group. This library of calculation succeeds DistBelief – the 1st generation

The technology signifies computation as stateful data flow graphs. What creates TensorFlow distinctive is its capacity to model computations on a tremendous scope of hardware, from buyer level cell phones to world-class multi-GPU servers. It can keep running on various GPUs and CPUs and assurances the scalability of machine learning among the different devices and contraptions without altering a lot of code.

TensorFlow began from Google’s have to instruct a PC system to mimic how a human brain works in learning and thinking. The system, known as neural networks, ought to have the capacity to perform on multidimensional data arrays referred as “tensors.” The end goal is to train the neural networks to identify and decode patterns and correlations.

In November 2015, Google made this technology open-source and permitted it to be adopted into a wide range of products and researches. Anybody, including researchers, engineers, and hobbyists, can help speed up the development of machine learning and take it to a larger amount in less time.

This move turned out being the correct one because there are such a variety of contributions from the independent developers to TensorFlow that they far surpass Google’s contributions. Wikipedia notices that “there are 1500 repositories on GitHub which say TensorFlow, of which 5 are from Google.” That being said, one of the discussions at Quora suspect that the released open-source code is the “cleaned-up” version from the one that Google utilizes as a part of its services.

How Does TenserFlow Work?

Utilizing the simple ordinary human language and a heavy simplification, we might see one side of TensorFlow as an advanced autonomous separating technology. At its heart, the technology is a tremendous software library of machine learning. It utilizes the database to support it “make decision”.

For instance, someone uploads a photograph to Google Photos. The technology will analyze all the details from the photo to its database and choose whether it’s a photo of an animal or human. At that point if it’s a human, it will try to determine the gender, age to all the way to who the individual is. The similar process is repeated for different objects in the photograph.

It also utilizes user’s data such as the identity of the individual in the image and the place where the image is taken, to enhance its library so that it can give well results in the future – both for the individual who uploaded the picture and for everybody else. Hence the term “learning”. But it does not stop at merely at knowing and learning data from pictures. There are so much that the technology can do with information from a picture. For instance, it can group pictures with related details such as the same person, the same location, the same date; perceive the pattern of faces to determine which family and friends the person in the picture belong to, and utilize the information to make videos of family vacation or animation from continuous shots.

That barely scratches the surface of how TensorFlow works, but I hope it can give you a general image of the technology. Also, utilizing only one instance can’t do justice to what it’s accomplished of.

And for all the Artificial Intelligence enthusiasts out there, it’s worth mentioning that Google previously created a computer chip technology optimized for machine learning and integrating TensorFlow into it. It’s called Tensor Processing Unit (TPU) ASIC chip.

Those who need to study more about TensorFlow can visit its tutorial page.

Applications of TensorFlow

We are at an initial stage of machine learning technology, so no one knows where it will take us. But there are a few early applications may give us look at what’s to come. As it starts from Google, it’s obvious that Google utilizes the technology for a large number of its services.

• More on Image Analysis

We have discussed the instance of utilizing the technology for image analysis in Google Pictures. But the image analysis application is also utilized in Google Maps’ Street View feature. For instance, TensorFlow is utilized to connect the image with the map coordinates and to consequently blur the license plate number of any car that is coincidentally included in the picture.

• Speech Recognition

Google is also utilizing TensorFlow for its voice assistant speech recognition software. The technology that permits users to speak out instructions is not different, but including the ever- developed library of TensorFlow into the mix may bring the feature up a few notches up. At present, the speech recognition technology recognizes over 80 dialects and variations.

• Dynamic Translation

Another instance of the “learning” some portion of machine learning technology is Google’s translation feature. Google permits its users to include new vocabularies and fix the errors in Google Translate. The ever developing data can be utilized to naturally detect the input language that other users need to translate. If the machine makes errors in language detection process, users can accurate them. And the machine will learn from those errors to improve its future performance. Furthermore, the cycle proceeds.

• Alpha Go

One fun instance of TensorFlow usage is Alpha Go. It is an application that is programmed to play Go. For those unfamiliar with Go, it’s an abstract board game for two players originated in China more than 5500 years ago, and it’s the oldest board game that is still continuously played now. While the rules are easy – to surround more territory than the opponent, the game is incredibly complex and, according to Wikipedia: “possesses more possibilities than the total number of atoms in the visible universe.”

So, it’s exciting what a learning machine technology can do with the infinite possibilities. In its matches against Lee Sedol – the 18-time Go world champion, Alpha Go won 4 out of 5 games and was given the honorary highest Go grandmaster rank.

• Magenta Project

Another exciting application of TensorFlow is the Magenta Project. It’s an ambitious project to make machine-generated art. One of the initial tangible results of the experiment is the 90 seconds piano melody. In the long term, Google wants to create more advanced machine-generated art through its Magenta project and build a community of artists around it.

On February 2016, Google also held an art exhibition and auction in San Fransisco showing off 29 PC generated – with a little help from human – bits of art. Six of the largest works were sold for as much as $8,000. The PC might still have a very long way to go before it can imitate a real artist, yet the amount of money people are willing to pay for the art demonstrate us how far the technology has gone.

Support for iOS

While we have already seen the abilities of TenserFlow on Android, with its latest version, TensorFlow finally includes supports for iOS gadgets. Since there are huge amounts of mobile applications accessible exclusively for iOS, or released first on iOS, it implies that we can expect more great mobile applications adopting machine learning in the near future. A similar thing can be said for the possibilities of wider adoptions and uses.