Tip:
Highlight text to annotate it
X
So, Chris, let's start with you. You are Mr. Robotics; correct?
>>Chris Urmson: Yes. >>Jonathan Zittrain: And when you were ten
years old, how predictable was it that you might be going into robotics?
>>Chris Urmson: I don't know about robotics, but I think engineering was pretty clear,
I think. I literally -- on my parents' fridge right now, there's literally still a picture
there with the sweater vest and the glasses with tape in the middle. So I think it was
pretty preordained. >>Jonathan Zittrain: It was not Halloween.
>>Chris Urmson: It was not Halloween, sadly. >>Jonathan Zittrain: And you knew you wanted
to go into engineering and technology. What first caught your eye then about robotics
in particular. >>Chris Urmson: I actually didn't really think
about robotics until I was finishing up my undergraduate degree. And I had an opportunity
to go work in telecom. And I was walking around the engineering building, and I saw this poster
with this robot crawling into a volcano. And I thought, "***, that's awesome." You know,
I could go deal with telecom stuff or go do something cool and real. And I thought, that's
the thing to do. >>Jonathan Zittrain: And what have you done
since? >>Chris Urmson: I've been working in robotics
for, I don't know, 12, 13 years, something like that. I've been involved with a lot of
vehicles that drive themselves. And recently came to Google to lead the self-driving car
project here. >>Jonathan Zittrain: Got it. And I understand
you have a video that you can narrate for us a little bit about the project.
>>Chris Urmson: So this is a video showing our car driving itself in and around the Bay
Area in San Francisco. We built a fleet of seven self-driving Priuses.
Take a Prius, put crap on it. We take it out on the road and have it drive
around. Here is on the freeway. So we can do surface roads. We, of course, do that in
traffic. The sensors we have on the vehicle allow us to operate day and night.
>>Jonathan Zittrain: Was that planned? >>Chris Urmson: No. I tell you, that was an
exciting moment. The vehicles can do basically anything that
a person can do. So they can see all the way around them, they can make lane changes, we
can see pedestrians and track them, we can see traffic lights. And we have 360-degree
perception around the vehicle, so we can understand what traffic is doing in the neighborhood.
The vehicles can drive at freeway speeds. They can do it precisely. So here we go through
a toll booth. They can do it in urban areas. So this is
a resort town near San Francisco. All of this is the car driving. So it's seeing the pedestrians
reasoning about them, moving around. This is Lombard Street. It's a very scenic
place in San Francisco. It is also home to the dumbest pedestrians on Earth. They walk
in with the traffic. It's amazing. The cars are able to do, you know, pretty
much anything a person can. So here we are in a construction zone, merging with other
traffic. Then we're going to get back on the freeway here and have to deal with the fact
that it's a short merge and there's traffic whizzing by at 70 miles an hour.
>>Jonathan Zittrain: Does the car have an aggressiveness setting?
>>Chris Urmson: Doesn't yet, no. It's set to 11.
And then, of course, we do a lot of testing on the track as well. So we joked we're the
Prius drifting team sometimes. So part of how it works is that we use some
of the best maps in the world. So we've taken Google Maps and taken it to the next level.
We have 15-centimeter resolution models of the world, and we have lane-level models of
the world as well that kind of feed into the car how it drives.
And then in real time, we take laser data from the vehicle and we process about a million
and a half thousand measurements per second, and from that we cluster out where other people
are, where other vehicles are, and are able to track them, bicycles, and whatnot.
We've also done some work with radar. So here, this is a radar like you might have in your
expensive luxury car for adaptive cruise control. We've taken that, refined the algorithms,
and then used that to make the thing work. >>Jonathan Zittrain: Incredible.
[ Applause ] >>Jonathan Zittrain: So from your point of
view, is the aim here just to deliver a really cool car that drives well? Or is there a larger
strategy afoot? >>Chris Urmson: So, you know, our goal at
Google is to advance this technology. We think in the long term this has got a tremendous
societal benefit. Worldwide, there's something like 1.2 million people killed in traffic
accidents each year. If you look at the utilization of automobiles through the day, it's about
3% in the developed world. So you buy a car, it sits there 97% of the time. By developing
this technology, in the long term, we think we can reduce traffic accidents, we can give
people back their time when they're driving, and we can make better uses of the resources
we have. >>Jonathan Zittrain: Amazing.
>>Chris Urmson: Thank you. >>Jonathan Zittrain: And you started on the
robotics team at Carnegie-Mellon University. Can you tell us a little bit about the innovative
environment and any differences between academia and corporate zone.
>>Chris Urmson: Sure. I've worked at Google and at Carnegie-Mellon. Carnegie-Mellon was
fantastic. There's lots of brilliant people there. But it's a university. And so it's,
you know, somewhat resource constrained. When you come to Google, again, we have the same
caliber or higher caliber of people, and we effectively have, you know, unlimited resources.
So for an academic, it's magical. >>Jonathan Zittrain: So you're trying to recruit
all your former fellow academics, at least the good ones?
>>Chris Urmson: Yes. [ Laughter ]
>>Jonathan Zittrain: Amazing.