Database Reference
In-Depth Information
LeCun is a peerless example of a data scientist with a transformational vision—in
his case, using deep learning to teach machines to perceive the world—who strives
to actuate that vision in both academic and industrial research laboratories. LeCun's
indefatigable pursuit of his vision, from the early days of AI to the current days of
Internet-scale machine learning, is the dominant theme of his stories about publish-
ing the MNIST data set and how it's frequently used for testing out new machine
learning methods, the ups and downs in the vogue for neural networks in academic
circles, and his own evolving beliefs about the comparative merits of
supervised and
unsupervised learning. LeCun's interview is a testament to his passion for machines
that can learn and his belief in their future.
Sebastian Gutierrez:
Tell me about where you work.
Yann LeCun:
I'm the director of AI Research at Facebook. Part of this role
involves data science, though there are other groups doing data science at
Facebook. AI Research can be thought of as the more advanced side of data
science if you want. I'm also a professor at NYU part time, which is conve-
niently located just across the street from my Facebook lab. Though I'm now
a university professor, most of my career has been in industry research. Early
on I worked at Bell Labs in a group that was, at the time, working on machine
learning and neural nets and similar projects. Then I became a department
head at AT&T Labs, which was the name of AT&T's research lab after the
company split up in 1996. I joined NYU in 2003, so I've been here a little over
11 years. I joined Facebook at the end of 2013.
Gutierrez:
What excited you about the opportunity at Facebook?
LeCun:
The main thing is that I was given the opportunity to create a world-
class research lab from scratch. Facebook did not have a tradition of being
active in research, so this was a bit of a new experiment for Facebook. It's not
every day that someone comes to you and says, “We'd like you to create a
research lab and hire the best people in the world.” You can't refuse that. Our
mission is very ambitious: understand intelligence and make machine more
intelligent. That's what attracted me here.
Gutierrez:
What is the makeup of your team?
LeCun:
There are about 35 people in the team as of November 2014. It's a
mix of research scientists and people who are more on the engineering side.
It's about three-quarters research scientists and one-quarter research engi-
neers. That 35 number will change soon, as we're going to grow really quickly.
Right now, our growth rate is actually limited by the number of good people
who are available, particularly since every other company is trying to hire in
this same area. We are in a big competition for talent with Google, Microsoft,
Baidu, Yahoo!, IBM, and a bunch of others.
Gutierrez:
What are the short-term and long-term goals of the lab?
