The Grill: Tom Mitchell

Computerworld - Tom Mitchell is head of the Machine Learning Department at Carnegie Mellon University, where he oversees 45 doctoral students and 20 faculty members. He has spent more than 30 years performing research in the machine learning field. He recently discussed how machine-learning algorithms that analyze MRI data can know what you're thinking, and why enacting strict privacy regulations to bring order to the "Wild West" atmosphere that pervades the data collection business could shut down real-time data mining, for better or worse.

As a researcher, you're known for your work in machine learning. What's important about your work? The question that defines computer science is, how can we get machines to perform different algorithms, and what algorithms can we

write? Machine learning is like that, but with a twist. Instead of hand-coding what the computer does, we train it. We show it examples. Machine learning has to do with how we build computer programs that improve with experience or find trends in historical data that make good predictions in the future. Face recognition, speech recognition and many other kinds of perceptual sensing problems end up being in the sweet spot for machine-learning algorithms.

What breakthroughs have you achieved? We applied machine learning to problems in neuroscience, looking at brain image data. We're starting to understand how the brain uses neural activity to represent the meaning of different words. We've trained a program that can look at functional MRI images of someone's brain activity and tell whether they're thinking about a house or hammer, for example.

So you can identify, based on a previous analysis of other people's brain image patterns, what object I might be thinking of? That is correct.

What are the practical applications of this ability to read people's minds? We're at the beginning of a revolution in psychology and neuroscience. Suddenly you can look inside the brain and turn what used to be fun philosophical questions into empirical science. We can look inside your brain when you see the color red, and we can look inside my brain when I see the color red, and we can ask, "Is it or is it not the same pattern of neural activity?" (continued on page 2)