Each summer, Microsoft Research labs around the world assign mentors, set up extra workspaces, and welcome a new crop of interns. The internship program is a significant component of Microsoft Research’s partnership with the academic world.
It provides an opportunity for students to work at a corporate research facility, interact with world-renowned researchers, and network with a diverse set of fellow interns. In return, the labs gain fresh perspectives, forge stronger ties with universities—and enjoy the energy and excitement of talented young scientists.
The diversity of research disciplines at each lab means interns are sure to join projects that require their domain expertise. What makes the summer special, though, are the relationships the interns build with researchers and their fellow interns—a key goal of the program.
Learning About Games
University of Cambridge student Abigail See has plans to pursue a Ph.D. in computer science after she completes her master’s program in mathematics. See always has been interested in game theory, and during her internship this year at Microsoft Research Cambridge, mentors Yoram Bachrach and Pushmeet Kohli, researchers with the Machine Learning and Perception group, gave her a chance to dive right in.
“We’re investigating a variation of ‘weighted-voting games’ in cooperative game theory,” See says. “A weighted-voting game models a system where agents, each in possession of a number of votes, can form a ‘winning coalition’ if that coalition holds enough votes. An example would be the formation of a coalition government from political parties that each has a certain number of elected representatives. Most models of weighted-voting games assume that all coalitions are possible—meaning that all parties are willing to work together. We are studying a model in which some parties are incompatible—just as in real life!”
“Perhaps the most challenging aspect of that internship for me was working on a large coding project with several other people,” See recalls. “Now, I’ve worked on two completely different projects and had the opportunity to learn from world-renowned scientists. Both experiences have been amazing introductions to the world of research. I love the vibrant atmosphere, especially the frequent lectures by visiting researchers, which are a good way to hear what’s going on in other areas.
“A Microsoft Research internship is a unique experience. You get an inside look at the world of research and the world of a technology giant.”
New Norms of Online Behavior
Ph.D. candidate in the Learning Sciences and Technology Design program at Stanford University’s School of Education, Matthews has been examining informal learning in online communities and how social-media practices alter our modes of acquiring and disseminating knowledge.
Working with her mentor, principal researcher Nancy Baym, and the Social Media Collective gave Matthews the opportunity to explore a different aspect of online communities. She has been able to study the evolving behavior and social norms that people must learn as they increasingly use online spaces for more aspects of their lives.
“The project asks what it means to behave professionally in a social-media world,” Matthews explains. “It studies how social media raises new and complex issues for the interactions between public figures and private citizens. I examined non-professional book reviewers and bloggers’ perception of their roles, as well as authors’ and agents’ roles, and how those perceptions defined expectations for literary professionals’ behavior toward private-citizen reviewers.”
Matthews makes it clear how much she appreciates the helpful environment at the lab.
“These are top researchers working on multiple, fascinating projects,” she says, “but they still make time to help out interns any way they can. The thoughtfulness and care they take whenever they engage with me can’t be stressed enough.
“To aspiring new interns, my advice is: Talk to people! There’s a wealth of resources to be tapped from both permanent and visiting scholars. Everyone is so welcoming that you shouldn’t be afraid to step out of your comfort zone or group to see what knowledge you can glean from them all.”
Faster Access to Massive Data Sets
“This is one of the most exciting projects I’ve ever worked on,” says Anna Choromanska. “It gives me the opportunity to get very deep into several different machine-learning techniques—and then borrow from those techniques and invent to create a completely new machine-learning tool with guaranteed good performance.”
Choromanska is a Ph.D. candidate in the Department of Electrical Engineering at Columbia University. Fascinated by machine learning, her main Ph.D. project is about splitting complicated optimization problems into much simpler sub-problems.
At Microsoft Research New York City, Choromanska works under the guidance of mentor John Langford, senior researcher, and postdoc researcher Alekh Agarwal. The project that has caught her imagination provides faster access to massive data sets through accurate prediction of data-set labels.
“Massive data sets are everywhere—think of biological data or web data,” she explains. “An example would be YouTube videos, where each video is an instance in the data set, and labels are their categories: news, music, sports, or cats—any one of a huge number of labels. The core challenge is to design algorithms that have been trained on labeled data sets to predict the labels of instances in huge unlabeled data sets and to do it very quickly and accurately.”
Her internship at Microsoft Research has given Choromanska plenty of challenges and hard work, but she already knows she will miss the lab when her internship is over.
“The culture of intellectual freedom and intensive work is compelling,” she says. “I was really impressed that interns have easy access to all the researchers. There is an open and friendly atmosphere. It’s truly amazing to sit in front of a whiteboard with the most famous researchers in your field and derive theorems with them. It’s like being part of the dream team.”