Database Reference
In-Depth Information
So if we look at future of data science, data science is not going away in the
sense that the science and the technology—as well as the engineering around
extracting knowledge from data—are going to be one of the big things of the
future that societies are going to be relying on. It's already the case to some
extent. The web relies on this already. But all society will rely on this. So this
is not a fad; it's not going away. If you say that data science is a fad, it's as if you
had said in 1962 that computer science was a fad. Look where we are now.
So my thinking around this phenomenon is that it will create—of course, it's
created an industry, which we all know—a demand for people who are edu-
cated in this area. And it's also creating a need for an academic discipline that
deals with this. This is something that some people aren't quite grasping at this
point. For instance, if you are a statistician, you say, “Well, that's just statistics.”
If you're a machine learning person, you say, “Well, that's just machine learn-
ing.” If you are a database person, you say, “Well, that's just a database, with a
bit of machine learning and statistics on top.” If you're an applied math person,
you say, “Well, all of these techniques and methods use applied math.”
All of those people are wrong. It's all of those things combined into one dis-
cipline: statistics plus applied math plus computation plus infrastructure plus
the application areas, which are the things that those methods can be applied
to, which require expertise. So techniques such as deep learning allow us to
reduce or minimize the amount of human expertise required to attack a new
problem, so that the machine does it by itself as much as possible. Of course,
at this point, there are always humans in the loop. And things like data visual-
ization make it easy for people to do things like this as long as there are still
humans in the loop. Eventually, however, those models will essentially just build
themselves.
I really believe in this concept of data science being a new academic discipline.
At NYU, we helped start this trend because we were early with the creation
of a Center for Data Science. We were also very early with the creation of
the master's degree in data science, which is and has been a big success. We
have had incredible support from the Moore-Sloan Data Science Environment
Initiative, which is a big program by the Moore and Sloan Foundations. This
initiative has grouped NYU, the University of Washington, and the University
of California, Berkeley, together with the purpose of establishing data science
as an academic discipline for the sciences.
Gutierrez:
What types of academic scientific disciplines will benefit?
LeCun:
The new way of doing genomics, astrophysics, neuroscience, and
social science is through analyzing massive amounts of data. That's going to
revolutionize all of these fields. In fact, it's already revolutionized genomics.
Genomics itself as a field is actually entirely dependent upon data analysis.
It didn't even exist before that. Genetics did, but not genomics.
