Yann LeCun on Autonomous Driving: Deep Learning is Obviously Part of the Solution | AI Podcast Clips

Yann LeCun on Autonomous Driving: Deep Learning is Obviously Part of the Solution | AI Podcast Clips


[Lex:] Elon Musk is confident
that large scale data and deep learning can solve
the autonomous driving problem. What are your thoughts on
the limits and possibilities of deep learning in this space? [Yann:] Well it’s obviously
part of the solution. I mean, I don’t think we’ll
ever have a self-driving system or at least not in the foreseeable future that does not use deep learning. Let me put it this way. So in the history of engineering particularly sort of AI-like systems, there’s generally a first phase where everything is built by hand, and then there is a second phase, and that was the case
for autonomous driving, you know 20, 30 years ago. There’s a phase where a little
bit of learning is used, but it is a lot of
engineering that’s involved in taking care of corner cases and putting limits, etc, because their learning
system is not perfect. Then as technology progresses, we end up relaying more and more learning. That’s the history of character recognition, it’s the history of speech recognition, computer vision, natural language processing. And I think the same is going to happen with autonomous driving that currently the methods that are closest to providing some level of autonomy, some
distant level of autonomy, where you don’t expect
a driver to do anything, is where you constrain the world. So you only run within
your 100 square kilometers, or square miles in Phoenix where the weather in nice
and the roads are wide, which is what Waymo is doing. You completely over-engineer the car with tons of lidars and
sophisticated sensors that are too expensive for consumer cars, but they’re fine if you just run a feat. And you engineer the hell
out of everything else, you map the entire world
so you have complete 3D model of everything, so the only thing that the perception system
needs to take care of is moving objects and construction and things that weren’t in your map. Then you can engineer a good SLAM system, all that stuff, right? So that’s kind of the current approach that’s closest to some level of autonomy, but I think eventually
the long- term solution is going to rely more and more on learning and possibly using a combination
of self-supervised learning and model-based reinforcement learning
or something like that.

5 thoughts to “Yann LeCun on Autonomous Driving: Deep Learning is Obviously Part of the Solution | AI Podcast Clips”

  1. Watch the full episode with Yann LeCun here: http://bit.ly/2NJiCov If you enjoy it, consider subscribing, sharing, and commenting.

  2. Without a fully fledged 5G infrastructure autonomous cars are not scalable. For scalability you must move away from reacting and focus on cooperation. Instead of your car reacting to changes (which requires petabytes of data to analyze very quickly) it will communicate and cooperate with all the vehicles, edge computing devices, wearable technology and other objects around it.

Leave a Reply

Your email address will not be published. Required fields are marked *