Saturday, October 13, 2007

Still Plodding Along

While my fellow denizens at UIUC are enjoying themselves at Reflections Projections, I myself am preparing for the Cornell University Glee Club Homecoming Concert tonight.

My research has been going slowly of late, as a number of other more pressing matters have been stealing my attention. And, of course, the web is always churning out new stuff. Notably for Cornell, Gun Sirer (who is also Ryan's advisor) was recently named one of Popular Science magazine's 6th Annual Brilliant Ten, and Jon Kleinberg was recently highlighted by Smithsonian Magazine in 37 Under 36: America's Young Innovators in the Arts and Sciences.

Paul Graham recently came out with another interesting read. It's probably true that a large fraction of people working in computer science are very focused towards web startups and whatever the next big thing coming from Google is. This certainly makes sense: Google and similar companies are highly visible, and web startups are relatively low-risk and potentially high-reward ventures. All you really need are a few good ideas, some good people and motivation.

In machine learning (like in many areas of computer science), the web is probably the largest problem domain. Everything from spam detection to search engines to online ad placement can be approached via machine learning techniques. In fact, my current research focus is geared towards designing machine learning algorithms for better search retrieval schemes. But what I love about research is not only the ability to break new ground, but also the chance to learn, collaborate and share insights across multiple disciplines. This is something that is typically harder to do in the industry.

The core contributions a company makes to society are the end products and services. On the other hand, research seeks to uncover the foundational principles which govern how we reason about the world. To be sure, most things one can reason about are indeed quite useless, so it's important to have practical considerations in mind. But as everyone knows, often just thinking about a problem in a different way can make a seemingly difficult task appear quite easy.

Michael Black recently visited Cornell. His work on computational neuroscience is quite inspirational and has been very well received. I encourage everyone to watch the video in the above link. After his talk, I asked myself whether, at the end of the day, would I be more satisfied knowing that my work helped Google improve its ad placement scheme, or that it helped a quadriplegic regain autonomy in his daily life? Both are certainly very valuable contributions. Fortunately, if I work hard enough, I might not have to make this choice. Hopefully, to some extent, I can get the best of both worlds.

No comments: