The New York branch of Microsoft Research, where computer scientists are brewing up new ideas in machine learning, prediction engines, and other fields, is getting space to bring in more collaborators. Though Microsoft maintains a massive office on the Avenue of the Americas, the research lab moved this month to a permanent new 22,000-sq.-ft. home on Sixth Avenue in the Flatiron District.
The new location puts the team in a geographically desirably spot to work more closely with research universities spread across the city, says Jennifer Chayes, co-founder of the New York lab and managing director at Microsoft Research New England in Cambridge, MA. It also offers breathing room to bring in additional postdoctoral scholars, interns, and other computer scientists.
The lab is being positioned to be the place for interdisciplinary data science. “We have some of the top [machine learning] researchers in the world and we’re hiring more,” Chayes says. The lab is also looking for hires in information retrieval, social science, and economics as they relate to data science and urban computing.
Scientists already on staff in New York include John Langford, architect of the widely-used open source, machine learning platform Vowpal Wabbit. Prediction engine research is spearheaded here by David Rothschild, Chayes says, who developed a platform that ran on Xbox Live for polling during the 2012 presidential election debates. He also created a separate methodology to predict the 2012 presidential election results. Teaming with in-house staff as well as collaborators at Columbia University and the University of Pennsylvania, Rothschild is developing new ways to do polling by working with various large samplings of data, such as users of Xbox Live, rather than traditional methods that take smaller samples that are supposed to represent a balance of the population, Chayes says.
By reweighing results from the samples, she says, Rothschild can detect small changes in who supports which candidate, a task that can be hard for polling companies that sample only a few hundred people, Chayes says. “That’s very different than doing market research based on a million responses,” she says, which Microsoft could potentially gather by contacting Xbox users and others.
Such new methods of polling might lead to new opportunities for Microsoft. “I see this as a new business for the company,” Chayes says. “Market research is a huge space where polling companies also dive in.”
In addition to prediction engines and machine learning, the staff at the New York lab also works on using data science in other fields such as computational social science. “It’s what underlies the interactions on social networks,” Chayes says.
The lab also teams with local university assets in such areas as information retrieval. “A lot of the data we get is text-based,” she says. “We’ve been working with Center for Urban Science and Progress (CUSP) at NYU and the city.” Data from the city, she says, is used to serve the population. As part of this collaboration, Claudio Silva, head of disciplines at CUSP and professor of computer science and engineering at NYU Poly, has even worked with Microsoft Research in China. Furthermore, the collaboration with CUSP has led to research in data privacy work for Microsoft teams in Cambridge and New York.
Microsoft established the New York lab in May 2012, first at the Avenue of the Americas office before moving last October to 102 Madison Avenue, where Perceptive Pixel, one of the company’s acquisitions, had been based. Chayes says even that space was not large enough for the growing New York research operation. The site also gives Microsoft labs in Redmond, WA, and Silicon Valley the chance to recruit local technology professionals who might not be ready to relocate. “Having a lab in the city now enables us to hire these people,” Chayes says.
