The Web company, seeking to influence how people design, test, and run artificial-intelligence systems, is making its internal AI development software available for free.
The Alphabet Inc.-subsidiary is releasing a program called TensorFlow as freely available open-source software, it said Monday. It’s based on the same internal system Google has spent several years developing to support its AI software and other mathematically complex programs.
Artificial intelligence enables products such as personal assistants, including Google Now and Microsoft’s Cortana. It also lets Facebook and Google’s software automatically identify and tag the contents of images uploaded into their systems. It also makes entirely new product categories possible, ranging from self-driving cars from Tesla and Google, to new forms of entertainment in virtual reality applications being developed by Facebook for its virtual-reality system Oculus.
The release of TensorFlow means anyone can download and modify the development software that underpins RankBrain, the AI powering part of Google’s search engine and new features such as its Smart Reply e-mail tool, Google wrote in a blog post.
Companies ranging from Google to Facebook Inc. to Microsoft Corp. are seeking to influence AI development by staffing up research labs, publishing academic research papers, giving presentations at conferences and even guest lecturing at universities. They’re betting that by being open they can entice talented academics to work for them, while encouraging the wider community to work on new AI technologies. Other companies, such as Apple and Amazon have been more secretive in the past, but are seeking to be more open.
The companies are also trying guide AI development. Facebook has released free code for Torch, an AI development tool used by Google’s DeepMind subsidiary, AMD, Intel, and others. Twitter announced its first contribution to Torch on Friday. Microsoft, meanwhile, has published numerous AI programs and datasets as open source.
By releasing TensorFlow, Google aims to make the software it built to develop and run its own AI systems a part of the standard toolset used by researchers, said Jason Freidenfelds, a spokesman for Mountain View, California-based Google. It may also help Google identify potential talent.
Google has shared TensorFlow ahead of its release with a few people, he said. Christopher Manning, a professor of Linguistics and Computer Science at Stanford University, has along with two students been writing AI programs using TensorFlow. He says the system can perform operations much faster than other tools, and includes features that take a long time to program, like fine-tuned systems that work with graphical processing unit chips from companies such as Nvidia Corp.
The main factor that could make or break TensorFlow is Google’s involvement with the open-source community that adopts it.
“As a researcher, if a tool makes you faster that’s pretty compelling,” Manning said. “Based on the evidence available to me now I think it will be widely adopted.”