Database Reference
In-Depth Information
LeCun:
I was born in 1960. So by the time I was nine, you had rockets flying
out in space, people landing on the moon, and
2001: A Space Odyssey
came
out, where you had space and intelligent computers in it. Science fiction was
the spirit of the time. I've always been interested in science. When I was a
kid, I thought and hoped I would become a scientist. I hesitated—not for a
very long time, unfortunately—between things like astrophysics, paleontology,
neuroscience, or AI. But I'm really an engineer. I got this from my dad who is a
mechanical engineer, I like to build stuff.
So what I thought about, as I thought about doing science, was: What are
the big scientific questions of our time? One question is: What is the uni-
verse made of? Which things like astrophysics and fundamental physics try
to answer. Another question is: What's life all about? Which biology and so
on try to answer. Another question is: How does the brain work? And this
question is a big, big scientific mystery. If you are a young scientist who has
not yet realized your limitation, you go for the big thing. And understanding
intelligence is a great big question.
As an engineer, I think of the brain as a very complex system. Intelligence is
something that is very abstract, which maybe can be modeled by mathemat-
ics, and so we can use an engineering approach to figuring out how the brain
works by trying to build intelligent machines to validate the designs or the
conceptual ideas that we have. A great deal of things have been said, some
very abstract, about how the brain actually works. But how do you know they
are right until you build a system that actually works? So at least there you
have most of the ingredients that are necessary. So that's the kind of scientific
question that has interest for me.
Of course, not only did I have to satisfy my desire to build stuff, I also had
to get jobs where I could develop good technology and do great work. It's
strange for me to say this, but it was never quite clear for me that I would ever
become an academic. I have—and maybe I should have become one earlier—
but industry research was kind of a perfect environment for me for a long
time. And so I'm sort of coming back to that now, though keeping a foot in
academia, and the two worlds are really complementary in this way I find. And
so I'm in this incredibly privileged situation where I can have one-and-a-half
feet in industry and half a foot in academia, which allows me to take advantage
of the complementarity between the two. In academia you can do things like
computational neuroscience and theory, while in industry you build ambitious
things that are difficult to do in academia.
Gutierrez:
What was the first data set that you worked with?
LeCun:
The first real data set I worked on was a medical data set that
I worked with while I was doing my PhD.The data set came from a medical study
of patients that come to an emergency room with abdominal pain. It turns out
that deciding whether or not to operate was a very difficult diagnosis to make
