CMU Mind Readers
Research on machine learning — that is, machines that improve with experience — has made astounding progress over the past decade or so. Such research is very mathematical and often theoretical in nature, which makes it hard for the general public (including policymakers thinking about funding priorities and high school students trying to decide a course of study in college) to understand what this is all about. And, of course, the name “machine learning” is, unfortunately, pretty inscrutable. Who knows how many people have heard the term and assumed that “machine learning” is about online courseware and other electronic learning aids.
But it looks like the public understanding of machine learning is starting to turn a corner. In some recent encounters with high school students, for the first time some of the kids seemed to know that statistics and computer algorithms are at the heart of everyday technologies such as search engines. And now, tonight, a truly amazing marriage of machine learning with neuroscience research here at CMU will be featured on the CBS show 60 Minutes. The feature will demonstrate the ability of machine learning to do, quite literally, mind reading. Today’s Pittsburgh Post-Gazette also has a very nice front-page feature on this research by Tom Mitchell and Marcel Just. (Tom is the Head of the Machine Learning Department and a Professor in the Computer Science Department. Marcel is a Professor of Psychology and co-director of the Center for Cognitive Brain Imaging.) In a nutshell, by using machine learning to analyze fMRI brain scans, a computer is able to “see” what word a person (even a news reporter) is thinking of, with remarkable accuracy.
Here at CMU we take machine learning, along with the incorporation of statistical methods into computer science, for granted. Indeed, the topic is required for the PhD in Computer Science and wildly popular in our undergraduate program. But in fact in most of the non-CS world, people don’t even know of the field’s existence. Press features like the ones in today’s Sunday paper and 60 Minutes should, in time, bring machine learning out of the closet. Indeed, I’m reminded of a recent statement by Hal Varian (a machine learning researcher at Google and Berkeley): “Statistics will be one of the sexy jobs of the next decade.”
Peter Lee @ January 4, 2009
[…] Sunday I wrote about the computational neuroscience work by Marcel Just and Tom Mitchell. Well, the CBS 60 Minutes feature was really quite interesting. In the segment, one of the staff […]