What the Heck is TensorFlow & Why Should You Care?


As technology continues to evolve, the applications of machine learning—and deep learning in particular—become increasingly fascinating. And those possibilities are even more exciting when a tech juggernaut opens up its code to everyone else.

Case in point: Google’s TensorFlow, an open source machine learning software library built with flexible architecture that allows programmers to compute anything that can be expressed in computational flow graph format. At its core, TensorFlow—which was built by the Google Brain team—applies deep learning to a lot of different areas. Google uses the technology for speech recognition, language translation, speed email replies, photo search functionality and more.

TensorFlow—which you can tinker with here—essentially enables researchers to try out any idea that might pop into their brains—even the craziest ones. When ideas are successful, they can be integrated directly into products without programmers having to rewrite code. The framework is speedy and versatile by design, allowing users to quickly shoot over ideas from their desktops to their mobile devices quickly.

Why open source?

According to Google, TensorFlow was open sourced to allow collaboration and communication between researchers plugging away from all corners of the globe. For example, a programmer in the United States might see what a peer in Belgium is working on, be inspired by that project, and take it even further. This naturally accelerates progress, as any programmer in the world can try his or her hand at deep learning (i.e., machine learning that attempts to mimic the way humans learn) by using what are arguably the best tools on the planet.

More accurately, Google likely open sourced TensorFlow to establish its approach to machine learning as the industry standard. Accelerating innovation and all that jazz? An added bonus.

TensorFlow is supposedly really powerful at perceptual understanding and language understanding, which means the technology has the capacity to do some pretty amazing things. So how exactly might TensorFlow affect your life one day? Let’s take a look.

Compose music with zero music skills!

A few days ago, Google announced Magenta, a project built on top of the TensorFlow framework that aims to create music and art. Coinciding with the announcement, Google released a 90-second melody created with Magenta, which sounds like a little ditty any b-grade pianist could create.

What if you’ve always wanted to be a musician but you never had any bona fide music chops? Let’s say you can’t sing to save your life, you could never figure out how to use your feet on a drum kit and you gave up playing guitar because your precious fingers hurt after a while.

Thanks to TensorFlow, the best composers of tomorrow may have never picked up a guitar or sat behind a keyboard. They might just be pretty skilled at coding.

Write poetry with zero writing skills!

TensorFlow has been used by a team of researchers in an attempt to make the AI more conversational by forcing it to read (and ostensibly learn from) thousands of romance novels and fantasy stories.

What popped out is quite interesting. But we’re not approaching Shakespeare just yet.

That’s okay. Google says we’re still a long way from machines being able to produce long, coherent narratives. In any case, suffice it to say TensorFlow is capable of writing clearer than an innumerable amount of individuals.

Build the right project on top of TensorFlow, and who knows? You may come up with volumes upon volumes of the most engaging poetry ever written.

Create art with zero art skills!

Earlier this year, Google demoed a mini robot that uses TensorFlow to sketch images. Take a picture on an Android phone, shoot it over to the robot, and voila: It’ll create a robotesque sketch of said photo.

You’ve always wanted to be an artist, but you never were able to graduate from drawing stick figures with smiley faces. But thanks to a little help from AI, you can use TensorFlow—or a program built on top of it—to create art.

Who knows whether the next Mona Lisa will be painted by a human, a robot or some combination thereof?

Classify images automatically!

Assuming you’re a fantastic artist, musician and writer, you may prefer using your humanity to create your works. But that doesn’t mean TensorFlow is useless to you.

For example, you can use TensorFlow to build an image classifier that can automatically sort, label and file images for you (or at least speed up the process; there may be some kinks you’ll have to iron out manually). A real estate agent might direct TensorFlow to sort pictures of homes by style and subsequently by color. Sure, there will be a home or two that the technology might not be able to classify. But for the most part, it’ll get it right.

Even if you can’t directly benefit from this kind of functionality, it’s not hard to see how tons of people, and organizations in particular, would.

Hire a multi-lingual virtual assistant!

There aren’t enough hours in the day, which is why virtual assistants are becoming more common. Thanks to TensorFlow, the virtual assistant on steroids may emerge sooner than you could imagine.

What if a virtual assistant were able to assist you all day long by learning about your daily routines, understanding patterns, recognizing images, and speaking multiple languages? According to one pundit, TensorFlow can make that happen.

Like any other piece of technology, TensorFlow is not without its critics. And quite frankly it’s scary to think about a world where technology creates all of our art. Still, the open source release of TensorFlow is exciting. No one can be sure where the technology will ultimately lead just yet. But if these early projects are any indication, the sky is truly the limit.

(Visited 831 times, 1 visits today)