Microsoft for the first time today talked about a fascinating new research initiative: machine teaching. The technology giant is looking for people who can teach computers how to perform machine learning, which can make software smarter over time.
Toward that end, Microsoft Research has assembled the Machine Teaching Group, comprised of people who are skilled in human-computer interaction, visualization, and other methods, under the leadership of researcher Patrice Simard, according to a blog post today from Next at Microsoft, a company-run website.
“Machine learning has proved so useful that it’s created a supply and demand problem: There just aren’t enough people with machine learning expertise to do all the projects businesses and organizations want,” Next at Microsoft’s Allison Linn wrote. “That’s prompted more efforts to make machine learning available to a broader group of people.”
The idea is to teach machines machine learning without requiring the human teachers to be especially technical. What the human teachers could have is domain knowledge. A chef, for example, would be quite useful for an application involving food, and from there, machines could become smarter based on initial input.
This is something that other tech companies are not doing — or if they are, they haven’t talked about it publicly.
It’s not clear how this work would be applied to Microsoft’s many applications across the board — although it is already being called on for the Language Understanding Intelligent Service, a new tool from Microsoft Research that falls under the Project Oxford suite, according to today’s article. Microsoft’s Face API, which lies at the core of Microsoft’s buzz-generating How Old Do I Look? app, is also part of Project Oxford.
Here’s a video summarizing the machine teaching project: