Database Reference
In-Depth Information
Gutierrez:
What's something someone starting out should strive to under-
stand deeply early on?
LeCun:
It depends where you want to go because there are a lot of dif-
ferent jobs in the context of data science or AI. People should really think
about what they want to do and then study those subjects. Right now the
hot topic is deep learning, and what that means is learning and understanding
classic work on neural nets, learning about optimization, learning about linear
algebra, and similar topics. This helps you learn the underlying mathematical
techniques and general concepts we confront every day.
Gutierrez:
What do you look for when hiring others?
LeCun:
I'm in a special situation because I'm building a research lab with
world-leading scientists and so I hire research scientists at the junior level,
mid-level, and senior level. What I look for is a track record in research, which
means a strong publication record, not necessarily lots of papers, but papers
with a particularly large impact that we know contain really interesting ideas.
A large number of people that we hire tend to have been on our radar screen
for a few years. Occasionally, someone shows up that wasn't on our radar, so
we are constantly looking for great people as well.
There is another category of people that we recruit, but frankly, it tends to be
more internal recruiting than external. We look for people with extraordinary
programming skills combined with a good knowledge of things like machine
learning or at least the ability to learn it really quickly. We're very fortunate
at Facebook AI Research that some of the people in the group are essentially
the most respected and top engineers at Facebook, which is amazing. These
people are just astonishingly good. They're making things possible that we
wouldn't otherwise be able to do and we couldn't have even approached.
So we look for mainly those two types of people—research scientists and
exceptional engineers. You need a very wide spectrum of expertise. And
don't forget, diversity of point of view is also a very important thing. You don't
want to just hire clones of the same person, because then they will all want to
explore the same things. You want some diversity.
Gutierrez:
Where do you see the biggest opportunities for data science?
LeCun:
If you are a scientist in an experimental science, particularly social sci-
ence, I think there's a huge amount of opportunities at the boundary between
the method side of data science and the domain science. This is going to
revolutionize a lot of areas of science and so it's a very exciting place to be,
particularly in social science. Other areas have already had a head start, like
genomics and biology. But neuroscience, in particular, and social science are
big areas of opportunity. If people are just starting out, I would suggest looking
there for big interesting and exciting problems to tackle. And of course, if you
are interested in methods, deep learning is where the action is.
