Google this week open-sourced its TensorFlow scalable machine-learning system, which can run on a smartphone or across thousands of computers. According to a company blog post, “We use TensorFlow for everything from speech recognition in the Google app, to Smart Reply in Inbox, to search in Google Photos. It allows us to build and train neural nets up to five times faster than our first-generation system, so we can use it to improve our products much more quickly.”
With the system open-sourced, others—from hobbyists to academic researchers—can use it as well. As a bonus, the company reports, “TensorFlow is for more than just machine learning. It may be useful wherever researchers are trying to make sense of very complex data—everything from protein folding to crunching astronomy data.”
So why is Google giving away this technology? Alistair Barr in The Wall Street Journal explains it this way: “Google retains much of what makes its machine-learning effort special: massive piles of data, a powerful network of computers to run the software, and a big team of artificial-intelligence experts to tweak the algorithms.”
He quotes Nello Cristianini, a professor of artificial intelligence at the U.K.’s University of Bristol, as saying, “It’s not a suicidal idea to release this. Deep learning is not plug-and-play. It needs a lot of testing, tuning, and adapting.”
Barr also quotes Patrick Ehlen, chief scientist at machine-learning startup Loop AI Labs, as saying, “Google has a lot of problems and has teams that work on these. But there’s something to leveraging the wisdom of the crowd.”
