In a show of engineering wizardry, last week Google rolled out an email assistant app to handle your inbox when you cannot. This week, Google open-sourced part of its powerful artificial intelligence (AI) engine, TensorFlow.
The email assistant, Smart Reply, is built using the software and capabilities of TensorFlow. Available for iOS and in the App store, Google created its Smart Reply app for people with a lot of email, and too little time to reply. Between here and there, mobile users often suffer inbox build-up. If you struggle through email that requires a fast response, Google Smart Reply may be just what you need.
Smart Reply functionality looks like this:
- Using AI neural networks, the Smart Reply app analyzes email inputs and outputs. With enough practice and learning, the algorithms atop the TensorFlow engine create a list of numbers, like a meme, called a vector.
- Once categorized, the vector is able to infer general topics. With enough training and data, Smart Reply understands a conversation about setting up a meeting, or other action.
- Using another neural network, potential outputs are analyzed and created—giving the mobile user pre-written options to respond to the email.
Google’s move to open source portions of TensorFlow kicks open the door for the rapid community development of AI capabilities and products.
A Google give-away?
On November 9, Google open sourced portions of TensorFlow, describing the AI engine as “our second-generation machine learning system… TensorFlow is general, flexible, portable, easy-to-use, and completely open source.”
Sharing on Google +, Vincent Vanhoucke, a Google tech manager writes, “TensorFlow is what we use every day in the Google Brain team, and while it’s still very early days and there are a ton of rough edges to be ironed out, I’m excited about the opportunity to build a community of researchers, developers and infrastructure providers around it.”
While Google is well-respected for its research advances, it is not commonly known for giving away highly marketable, proprietary software and hardware information. The answer seems to be that Google is giving away just enough information to spark innovation and creation—but not enough to lose its competitive advantage with AI.
According to Wired, Google is managing the release and use of select TensorFlow features as follows:
- While Google is sharing the TensorFlow code for use through an open licensing agreement, the TensorFlow development project remains housed at Google.
- The algorithms, and sample sets released by Google are sufficient for training models on public data sets, but not across machines. As noted by Skymind start-up entrepreneur, Chris Nicholson, “Google is still keeping an advantage. To build true enterprise applications, you need to analyze data at scale.”
From apps to autonomous robotics, artificial intelligence is a field primed for explosive growth. On November 5, car maker Toyota announced a $1 billion investment in an artificial intelligence and robotics technology research center in Silicon Valley.
States Jeff Dean, Senior Google Fellow, “Deep Learning has had a huge impact on computer science, making it possible to explore new frontiers of research and to develop amazingly useful products that millions of people use every day.”
For mobile apps, automobiles, and other adventures—the future of AI is wide open.