Database Reference
In-Depth Information
inside companies like Facebook, Google, and Microsoft. There are other play-
ers as well, of course, because it's a bit of a gold rush. It's not quite a land
grab because you can't stop other people from doing the same thing you are
doing, even though there are such things as patents. At Facebook we don't file
patents to sue other people; we occasionally file them so that other people
don't sue us.
Essentially, Facebook files patents so that we can say, “We came up with this
idea, perhaps at the same time that maybe you did, but you can't sue us for
using it.” Another way of protecting ourselves like this is to publish papers,
because that also establishes priority. So there is a bit of this interesting land
grab. What's been very surprising to me is the amount to which there is now
a sense in the industry that AI is going to revolutionize everything. AI all of a
sudden or machine learning, particularly deep learning, went from some sort
of obscure academic field of investigation to front and center at major suc-
cessful companies like Google, Facebook, Microsoft, IBM,Yahoo!, Baidu,Yandex,
and others. It's been a very recent, quick, and surprising phenomenon to me.
Gutierrez:
How have you found this sudden change and attention?
LeCun:
At some level it's not really a surprise, but it's not like you can count
on this kind of stuff happening. As a researcher, you do research and you
hope someone will pick it up. When it's Mark Zuckerberg picking it up, it's a
different kind of person than the usual colleagues in academia or industry. A
lot of people in Silicon Valley and technology are fascinated by the potential
impact of AI in the future. And that's what's driving a lot of this growth and
concentration in this academic field. I agree with this. Certainly, AI will have a
big impact on society. And it's probably true that companies that can position
themselves in the right way in AI are going to have some sort of position of
advantage going into the future.
That's why, for example, Google is going into robotics. They believe that robot-
ics is going to be the big thing over the next 10 years. Now, for some other
companies with less financial strength, it might seem premature. Maybe the
industry is not going to take off for the next 5 or 10 years, but Google has
enough resources to wait it out and make sure that they're ready or actually
help it happen. Facebook has a similar approach. Facebook's core business is
to facilitate communication between people and also facilitate communication
between people and the digital world. That's going to be mediated by AI. In
fact, it's already mediated by machine learning.
Gutierrez:
How do you know whether you're solving the right problem?
LeCun:
The question is different depending on whether we're talking about
research projects or more practical short-term things. It's always very dif-
ficult to know whether you're working in the right direction for a research
project. This relates to the vision thing I was talking about earlier, that if you
have some idea of where the world might be going, then it's easier to continue
