On November 9th, Google announced that they’re open-sourcing their machine-learning system, TensorFlow, saying that they “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. And that, in turn, will accelerate research on machine learning, in the end making technology work better for everyone.”
Google, of course, isn’t the first company to open-source their system. Plenty of other brands have done this before. Epic Games, for example, open sources projects and even rewards contributors working on “innovative projects built in and around Unreal Engine 4” with their Unreal Dev Grant. For this, the company claims, there are no strings attached, and recipients can continue to own their IP, publish however they wish, and use grant funds without any restrictions or obligations to Epic.
“They recognize the value of contributions and of helping developers create,” explains Chaotic Moon Creative Technologist Matthew Murray. “If more people can build—if more people are using the engine and being successful—it makes the company more successful.”
But by providing access to TensorFlow—which Google claims is twice as fast and more flexible than DistBelief, their existing machine-learning infrastructure—there are some pretty big implications for the future of the tech as a whole.
However, before we get into that, let’s play some catchup. For those unfamiliar with the platform, TensorFlow is described here as “a library of pre-built portions of neural network code with easy-to-use tools to customize them deeply, and add to them with as much flexibility as possible.”
And for those unfamiliar with machine learning as a whole? Well, Wired says it best:
With deep learning, you teach systems to perform tasks such as recognizing images, identifying spoken words, and even understanding natural language by feeding data into vast neural networks connected machines that approximate the web of neurons within the human brain. If you feed photos of cats into a neural net, you can teach it to recognize cats. If [you] feed it conversational data, you can teach it to carry on conversations.
Now, at first glance, open-sourcing this system and sharing the code seems like a relatively altruistic play—rather than limiting access to this valuable tool and keeping it to themselves, they’re opening it up to anyone with interest in the technology. However, this isn’t just for the benefit of researches who want to implement the platform. Neither is it simply sweet news for nerds (no offense) who have the desire to toy with it in their free time. In all actuality, open-sourcing TensorFlow has the potential to advance the technology of machine learning as a whole and, on a more what’s-in-it-for-us note, improve the specific system itself.
“Progression speeds up when you open the system up,” Murray says. “When you have a machine that’s supposed to be learning, the fastest way to make it grow and learn is to open it up to more resources. They built it for a specific purpose, but by letting people use it for different purposes and experiment, they can find bugs, make improvements and better the system as a whole.”
And by giving way their algorithms and code, the secret software sauce, Google is also able to get what they really want: more data.
As Lukas Biewald—CEO of CrowdFlower, a company that assists online companies in dealing with data—told Wired, for AI the real value isn’t in the software or algorithms but the data that’s making it smarter. And by open-sourcing this platform and increasing the amount of data that’s being poured into the system, we’re making artificial intelligence more…well, intelligent…than ever.
For example—and this is the LAST Wired reference, we swear—as the publication so aptly puts it: “To teach a system to recognize a cat, you need an awful lot of machines and an awful lot of cat photos.”
“I think it’s great that big companies are open-sourcing their machine-learning and AI platforms,” says Chaotic Moon Creative Technologist Phillippe Moore. “It’s this whole idea that shared knowledge benefits everyone.“
And it’s not a decision that only matters to Google or to those really interested in the tech nuts and bolts. Open-sourcing TensorFlow is a move that’s set to affect anyone who uses certain Google services as well. (And, let’s be honest, that’s most of us.) For example, the people who use Smart Reply and Google Photos will benefit simply because the system that runs these platforms will continue to become more efficient and effective than ever.
After all, the more the system learns, the more it’s able to refine and recognize voice, images, etc. The result? More accuracy and hopefully fewer…um…humiliating mistakes.