The Guide to Open Source

 

WHEN SOMEONE BUYS a new smartphone, often they're preoccupied with the camera specs or the size of the screen or its storage capabilities. It's easy to overlook one of the most foundational aspects of these sleek consumer gadgets: their operating systems. The world's most popular mobile operating system is Google's Android. It powers more than 86% of smartphones in the world. What's even more remarkable is that Android is based on the open source Linux operating system.

That means anyone can view the code at the heart of the vast majority of smartphones, modify it, and, more important, share it with anyone else. This openness enables collaboration. Unlike, say, Microsoft Windows, which was developed and is maintained by a single company, Linux is developed and maintained by more than 15,000 programmers around the world. These programmers might work for companies that compete with each other, or they might volunteer to create something new that’s then given away. For free. Gratis.

As crazy as that might sound, the open source way of building software is now embraced by the likes of IBM, which plans to pay $34 billion for open source company Red Hat, Microsoft, which paid $7.5 billion to acquire the code hosting and collaboration platform GitHub, and Walmart, which released its own open source software.

Open source is even seeing applications in the next iteration of technology: AI. Google open source its artificial intelligence engine, TensorFlow, in 2015, enabling companies and researchers to build applications using some of the same software the search giant used to create tools that search photos, recognize spoken words, and translate languages. Since then, DropBox has used TensorFlow to recognize text in scanned documents and photographs, Airbnb has used it to help categorize photos in its listings, and a company called Connecterra has used it to help dairy farmers analyze their cows' health.

Why would Google give away something so central to its business? Because it hoped outside developers would make the software better as they adapted it to their own needs. And they have: Google says more than 1,300 outsiders have worked on TensorFlow. By making it open source, Google helped TensorFlow become one of the standard frameworks for developing AI applications, which could bolster its cloud-hosted AI services. In addition to garnering outside help for a project, open source can provide valuable marketing, helping companies attract and retain technical talent.

Keep in mind that Google didn't give away the data that powers its AI applications. Just using TensorFlow won't magically allow you to build a search engine and advertising business that can compete with Google.

So Google stands to benefit, but why would an outsider contribute improvements to TensorFlow? Let's say a company makes its own version of TensorFlow with unique elements, but keeps those elements private. Over time, as Google made its own changes to TensorFlow, it might become harder for that other company to integrate its changes with the official version; also, the second company would miss out on improvements contributed by others.

In short, open source provides a way for companies to collaborate on technology that’s mutually beneficial.