A Conversation with David Plaut

Priors



Jordan Feidler (JF):
Maybe we can start with a quick run through of your background. What was your undergraduate degree in?
David Plaut (DP):
It was in a funny mixture of things. I was a double major in Mathematics and an inter-departmental program in Cognitive Science before it became an actual, official undergraduate major. The reason I was in Mathematics is that that was the only way to do computer science; there wasn't an undergraduate degree in computer science at that point. One had to kind of piece it together in various ways.


JF:
I have this association with vision from Rochester....
DP:
That's right, The Center for Visual Science . I had a very strong interest in vision and my intent in going on to graduate work was to do work in computer vision. And, in fact, I went on to do a year's study in Cambridge, England, as a Churchill Scholar, in high-level vision (actually, psychophysics; visual psychophysics, motion perception), in preparation for going on to do graduate work in computer vision. I did a little of that but then stumbled into language for various reasons which we can get to.


JF:
So you spent a year in England after you finished your undergraduate degree?
DP:
That's right. Now, I have to admit I mostly spent it rowing on the Cam rather than slaving in the lab, although I did slave away in many a dark room. In visual psychophysics, basically you can run yourself as a subject because the processes aren't very subject to control or bias; even knowing what the experiment is about is very peripheral. And so you spend lots of hours in dark rooms pushing buttons, staring at screens, trying to figure out whether things moved or not and stuff like that. It was fun.


JF:
You went from that to...
DP:
I was in Cambridge in 1984-85 and started the PhD program here in Computer Science at Carnegie Mellon in the fall of '85. I was in the AI group and I had come specifically to work with Geoff Hinton .


JF:
Now, I had in mind that you had spent some time in Toronto...
DP:
That's right. Geoff Hinton left CMU to go to the University of Toronto at the end of my second year. But he remained my advisor and I would travel up there every so often; he was collaborating with Tim Shallice on some computational modeling of acquired dyslexia, reading disorders, in brain-damaged subjects. They had published a paper and were looking to do some follow up work and had advertised for a post-doc. I saw the advertisement and I thought: "This looks like fun, too bad I'm just a graduate student slaving away." And I sent Geoff mail saying: "Oh, it's too bad, this looks like fun..." and he said: "Well, you know, if you're interested maybe we can work something out...." So I ended up spending about a year up in Toronto working on this, and the work led to the bulk of my thesis.

As I said, I started out working in vision, high-level vision, trying to understand how people recognize objects and integrate information across different reference frames. I am still very interested in that, but it is a very hard problem. Relative to that, word reading seemed a little more tractable. I had actually started out interested in the visual aspects of word reading and integrating letters over time and grouping them into words and it turns out that that kind of process is one that's relevant to some patients with brain damage that have reading deficits. Then we wandered into these other areas in word reading. But originally I was just coming to work with Geoff and Tim Shallice on extending their collaboration. I spent a year up in Toronto and then came back. So, that was in 1990, most of 1990. That was my fifth year and then my sixth year I finished my degree back here.


JF:
And after your degree did you do a post-doc anywhere?
DP:
I did. Actually, it took me six years to finish, and then I came here, over to Psychology, to work with Jay McClelland. I was a post-doc with Jay for three years here before starting as a faculty member this last September.


JF:
And you're presently in the Psychology Department?
DP:
That's right. Although, my appointment is joint between the Psychology Department here and the Center for the Neural Basis of Cognition.


JF:
Let's discuss in a little more detail how you got to where you are today. You started out in CS at Rochester. Were you always interested in computer science?
DP:
Actually, I went to Rochester to do optics. I don't know why. Well, I know why I went to Rochester, I don't know why I wanted to study optics; mostly because as an undergraduate I thought that physics and laser applications were very interesting. And Rochester turns out to be, I don't know if it's still true but it used to be, one of two existing undergraduate programs in optics; mostly you had to do an undergraduate degree in physics and a graduate degree in optics. Eastman-Kodack is in Rochester so...

So, I went to study optics and I took engineering computer science courses as part of that and found that it was much more interesting than the optics program. So I kind of wandered my way into wanting to do computer science. And, I think in the beginning of my second year, I took the introductory psychology course as part of the distributional requirements. The instructor, his name was David Taylor, was a cognitive psychologist who was very interested in artificial intelligence, very interested in connectionist models and an excellent teacher. I was very intrigued by him and talked with him, and ended up working for him as a laboratory assistant running cognitive psychology experiments, as well as engaging in an artificial evolution project that he was interested in - using environmental pressures to develop more and more complex kinds of behaviors.


JF:
So you had an interdisciplinary streak right from the start?
DP:
That's right. David talked me into it. I didn't want to do Psychology and I had discovered that there was this inter-departmental Cognitive Science program; but I had said I really wanted to get a very strong background in Computer Science and he was the one to suggest doing them both, that I could both get the solid math and computer science background while also broadening my background in cognitive psychology and neurobiology, in philosophy of mind and philosophy of science. Rochester's cognitive science program was just developing at that time.

It turns out, on a bit of a side note, that they were in the process of formulating a cognitive science undergraduate major formalizing this. The faculty had put together a proposal and, in fact, three of us undergraduates got together with the course catalog and with the faculty proposal and realized what they were proposing was impossible. We basically re-wrote it and presented that to the faculty committee and that ended up being the structure of the current undergraduate cognitive science program. I guess I had some say in inflicting my program on subsequent classes.


JF:
Is this part of what led you to move from computer science to neuropsychology?
DP:
My work with David Taylor was part of my initial interest in connectionist models and also my interactions with the computer science department there, most notably Dana Ballard and Jerry Feldman . At that time, there was a friendly debate between local versus distributed connectionist models, and which were really more appropriate and most effective for explaining human cognition. David Taylor was on the distributed side; Ballard and Feldman were arguing for localist approaches, although Dana Ballard, I think, was a little more on the fence on that. So that also was very much part of the intellectual environment and was part of what led me to want to do this kind of research in my own career.

I also did, an independent study as a senior with Jerry Feldman, to give the enemy equal time as it were, doing a small connectionist simulation of object recognition in a localist style. Early on he asked me a question, because I had written a paper in this computer vision course the semester before, criticizing local connectionist models. He had gotten a hold of it, had read it, and he asked me: "Are you just sort of interested in this or are you serious? Did you just write this because you had to do this for a course or are you really interested in pursuing this kind of work?". I said: "I'm definitely interested in pursuing this. I'm curious about these things." And that work with him was really my first taste of doing research of my own design. Most of the stuff I had done with David Taylor was his motivation, but Jerry let me go and play and try to solve a problem, and that was really exciting and I really appreciated it. He allowed me to discover how exciting research could be, and how much hard work it takes; you really have to pound at a question and try out a lot of things before you can get somewhere.


JF:
That sort of leads you to coming here, but it still doesn't bring in the neuropsychology.
DP:
That's right, it doesn't. I was originally going to come here straight to graduate school, when I was made aware of scholarship opportunities to spend a year in England; I thought that was too good an opportunity to pass up. So I applied and was fortunate enough to be able to spend a year working with Oliver Braddick in Cambridge. That was very useful both in grounding my empirical background - having the time to spend reading a lot of the empirical literature in vision - as well as getting a clearer sense of just the general style of empirical work. Again, most of what I had done was computational and certainly most of what I have done since then has been as well.

So, I spent this year in England and then came here as a graduate student, mostly interested in high level vision at that point. I guess my interest in neuropsychology developed in part due to interactions with Martha Farah, here in Psychology at the time and since moved to Penn. With a strong interest in high-level vision, I had a lot of interaction with her and was a TA for her at one point. That was the beginnings of it. And I think it really developed in the year I spent in Toronto working with Tim Shallice. I spent a lot of time working with Tim. He's just an absolutely outstanding individual and, intellectually, the breadth of his ideas and the mastery of the literature, being able to think across different domains,.... He is just absolutely wonderful to work with. I think that that interaction for me really crystallized the kind of work that I like to do. I very much enjoy collaborating with people across disciplines, having people bring their own ideas from their own backgrounds to bear on common problems, and then just trying to find ways to weave them together, because that is often the challenge: finding the vocabulary and being able to think about problems from different sides, with different people, collectively. That is very much where the style of my current work kind of developed out of.


JF:
Anyone else who impacted you or influenced your work?
DP:
Well, I wouldn't want to leave out Geoff Hinton. I think he had a less direct impact on the content of the work since it's taken such a psychological direction. Certainly one of the things that is very striking about Geoff is his excitement about ideas and he is one of the most brilliant people I know, in terms of being able to generate interesting ideas and exciting ideas about problems. Working with him is very engaging and, again, made it very clear to me how engaging research can be.


JF:
Maybe not the content now, but the process?
DP:
I wouldn't want to underestimate the content as well. The technical side and really exploring fully the implications of ideas certainly comes from him. But what is striking for me about him is just the incredible energy and the excitement of ideas.

Back to the index. Back to the index.

Got a suggestion?
Send us a note! 		Have and idea to share?
Post it to the CNBC Newsgroup!