|
Look Who's Talking!
Tom
Mitchell is the director of the Center for
Automated Learning and Discovery (CALD)
and Fredkin Professor of AI and Learning. He’s Pennsylvania
born, loves to wind surf and ski. Here you can find out more about
his work, his life, and his cabin by the lake!
Can you tell me a little about yourself? Let’s
start with where you’re originally from
I’m from Pennsylvania. I was born in a little town called
Blossburg, Pennsylvania (which is in the middle of Northern-Central
Pennsylvania). But I grew up in up-state New York. I went to college
in Boston at MIT, went to graduate school in California at Stanford
and lived there for 5 years, and then I went and taught at Rutgers
University in New Jersey for 6-years. I finally came here in 1986.
So I’ve been kinda around…
So why did you choose Carnegie Mellon?
Oh, it was easy…I chose Carnegie Mellon because I was already
a professor at Rutgers University, and I came here on a 1-year visiting
professor position. I ended up having the best time! I team-taught
a course with Allan Newell and Geoff Hinton, which was great. So
at the end of the year when I was supposed to go back, we all agreed
that I should just stay here! It’s such a great environment
for doing research in AI or research in computer science. So for
me it was very easy.
Do you think the environment is very different
from MIT or Stanford?
I think it is. I’ve been a student at both of those places,
and there seem to be two big differences to me: one is that here,
there’s a very strong emphasis on experimental computer science—that
is to compliment the theory, to really be working on some computer
program that does something important in the world—and putting
those two together is a great thing that I think Carnegie Mellon
does better than anywhere. The second is that there’s this
wonderful lack of boundaries between disciplines. For example, right
now I’m collaborating with a professor in psychology, and
I’m directing the Center for Learning and Automated Discovery,
which is a huge collection of computer science and statistics faculty
who are together all the time. And there’s no bureaucratic
organizational issue at all at Carnegie Mellon. It’s kind
of expected that you’ll naturally work together—but
it’s actually kind of unique to Carnegie Mellon to be that
open.
Since your research focuses a lot on the automated
learning of computers, have you had a lot of background in (or do
you enjoy) the biological sciences a lot?
You know, it’s funny—I don’t have a lot of background
in biological sciences, but I’m learning more and more these
days. So I’m kind of a student right now, because in the last
couple of years I’ve moved from my earlier work, which was
based just on how we can get computers to learn, to applying that
to studying brain image data. So I get to learn some things about
neuroscience and I consider myself a student in that area right
now—and I’m fascinated by it.
How do you feel about the meshing of the two sciences:
Computer Science and Natural Sciences? What do you think the future
will bring with the combination of the two?
One area where you can already see big growth (and I think it’s
going to be even bigger in the coming 5-10 years) is that a lot
of natural science—biology, astronomy—is becoming very
data intensive…compared to a hundred years ago when Mendeleev
did his gene experiment with peas, now we have just huge data sets,
micro array data, and so forth, and for brain imaging experiments
we have terabytes of brain imaging data. So it’s not possible
for people to just sit down and look at a terabyte of data by hand.
So more and more you see computers being an integral part of empirical
science just because empirical science has huge data sets. Especially
in our Center for Automated Discovery and Learning, where we’re
specifically interested in finding what kind of computer algorithms
we can come up with that can find general trends in large, detailed
data sets. There’s a huge opportunity for building programs,
like the ones I work on, that learn the patterns of brain activation
that indicate you’re thinking of nouns instead of verbs, or
in the case of cell biology, we learn the patterns of protein appearance
or gene expression that indicate certain processes going on in the
cell. It’s really hard by hand to find those patterns, and
learning algorithms for finding those patterns—data mining
algorithms—are becoming increasingly important in empirical
science.
Learning the brain’s algorithms for doing
things is very difficult, and is not very well understood as yet.
Do you ever find it frustrating trying to get computers to learn
things that we ourselves don’t know the inner workings of?
That’s actually a very interesting observation—I actually
don’t get frustrated by that—why? [Laughs] I don’t
know!
Maybe it’s odd, but it’s true that much of the work
in machine learning—how to get computers to learn—has
been kind of unguided by anything we know about human learning.
It just grew up on its own—“ok, how would we engineer
this system to look at a lot of data and discover regularities?”—so
people engineered those instead of looking at how humans do it and
then trying to duplicate it. But recently, because I’ve been
looking at the brain, I’ve been starting to learn more about
what people know about human learning—and it’s very
different. For example, when we humans learn, a big part of what
determines whether we succeed or not is all about motivation. And
there’s nothing in machine learning algorithms that even remotely
corresponds to motivation. So it’s just a very different phenomenon…maybe
in 10 years we’ll understand it better, but right now, the
two are very different.
What’s
your favorite thing about Carnegie Mellon?
Wow…
I feel like it’s a place where you are encouraged to do whatever
you feel like doing, and you don’t have to worry about what
discipline you are in—here I’m in computer science and
I’ve decided I want to study the brain, and that’s fine.
There’s never any worry about whether you are allowed to work
on a certain problem—just do it!
Did you always want to teach?
No. It was a pure accident. I enjoy teaching though—I wrote
a textbook a few years ago even, because I started teaching this
machine learning stuff, and started noticing that it was enjoyable
to get the material all organized. But the thing that really made
me want to become a professor was that I was interested in the research.
And initially when I became a professor, I thought of teaching as
kind of interesting, but something that I just did on the side.
After I wrote the textbook, I decided that teaching was really a
major part of what I was doing—but I didn’t start out
that way!
What’s your favorite class to teach?
I love teaching the Machine Learning course, because that’s
the area I work in. But I also really enjoy picking a topic I don’t
quite understand, and then teaching a seminar on that, where I’m
running the seminar, but I’m learning as much as the students.
Those kinds of courses are fun.
Were you the one who actually ended up creating
the Machine Learning course here?
Yeah…that’s true! [Laughs]
Let’s shift gears a little now…onto
the more personal side…from all the experiences you’ve
had at Carnegie Mellon, what’s been your most cherished memory?
Oh, wow. There’s so many, it’s hard to say! Well, one
of my favorites really is when I first came here and taught that
course with Allan Newell and Geoff Hinton and we had the three of
us teaching this course, and it was on Architectures for Intelligent
Agents, which was an area that the three of us were working on from
different angles. We had the best bunch of students, and the course
was not at all a lecture course—we would just have one or
two papers each week that we all had to read—the instructors
and the students—and then we would come in and just talk about
it. That was one of my best memories of CMU because that was my
first exposure of how things go around here, and this opportunity
for free-wheeling, just very creative discussion.
What hobbies or passions do you have outside of
work?
I love to wind surf. I play a little music—some guitar; I
like to ski.
If you didn’t go into academics or administration,
what would you have done in another life?
It’s funny—sometimes I think I would have been a doctor…
When’s your birthday?
August 9th.
Are you married and do you have and children?
I’m married to a wonderful person who’s the head of
the Dewey decimal system—her official job title is, “Editor
of the Dewey Decimal System”—she makes up all the numbers
on the books… And I have two wonderful daughters, who are
both married. One of them is a PhD student, growing artificial arteries
for her thesis at Duke. The other one is a 5th grade math teacher
in an inner-city school in Washington, D.C.
What’s your favorite vacation spot?
My home! [Laughs]
Our house is in Maryland—I have an apartment in Pittsburgh,
but our house is in this place called Deep Creek Lake, which is
about a two-hour drive from here. We used to have a vacation cabin
down there and a house up here, but after our kids went off to college,
we decided that we like so much being at the Lake that we should
just make that our main residence. So I’m here in Pittsburgh
during the week, but on weekends, I’m at home!
What’s your favorite food?
My favorite restaurant in Pittsburgh these days is Eleven. It’s
downtown.
In general, I love good food—French food is my favorite.
Where’s your favorite place to go in da
‘Burgh?
I like to go to Café Zinho, which is in Shadyside. It’s
just a very small but very nice place to go. Eleven is a nice, expensive,
elegant place to go, but Café Zinho is my favorite ‘don’t
break the bank’ kind of restaurant. [Laughs]
What’s your favorite color?
Blue.
Complete this sentence: When the going gets tough,
I…
Run faster! [Laughs]
What motto or philosophy do you live life by?
Be honest…with yourself and others.
More Interviews
|