1
00:00:00,528 --> 00:00:04,477
So in 1885, Karl Benz
invented the automobile.

2
00:00:04,707 --> 00:00:08,469
Later that year, he took it out
for the first public test drive,

3
00:00:08,469 --> 00:00:11,844
and -- true story --
crashed into a wall.

4
00:00:12,184 --> 00:00:14,227
For the last 130 years,

5
00:00:14,227 --> 00:00:18,546
we've been working around that least
reliable part of the car, the driver.

6
00:00:18,546 --> 00:00:19,900
We've made the car stronger.

7
00:00:20,200 --> 00:00:22,748
We've added seat belts,
we've added air bags,

8
00:00:22,748 --> 00:00:26,719
and in the last decade, we've actually
started trying to make the car smarter

9
00:00:26,719 --> 00:00:29,657
to fix that bug, the driver.

10
00:00:29,657 --> 00:00:32,918
Now, today I'm going to talk to you
a little bit about the difference

11
00:00:32,918 --> 00:00:36,726
between patching around the problem
with driver assistance systems

12
00:00:36,726 --> 00:00:39,290
and actually having fully
self-driving cars

13
00:00:39,290 --> 00:00:41,170
and what they can do for the world.

14
00:00:41,170 --> 00:00:44,165
I'm also going to talk to you
a little bit about our car

15
00:00:44,165 --> 00:00:48,164
and allow you to see how it sees the world
and how it reacts and what it does,

16
00:00:48,164 --> 00:00:51,351
but first I'm going to talk
a little bit about the problem.

17
00:00:51,651 --> 00:00:53,299
And it's a big problem:

18
00:00:53,299 --> 00:00:56,388
1.2 million people are killed
on the world's roads every year.

19
00:00:56,388 --> 00:01:00,172
In America alone, 33,000 people
are killed each year.

20
00:01:00,172 --> 00:01:02,200
To put that in perspective,

21
00:01:02,200 --> 00:01:06,997
that's the same as a 737
falling out of the sky every working day.

22
00:01:07,342 --> 00:01:09,128
It's kind of unbelievable.

23
00:01:09,548 --> 00:01:11,846
Cars are sold to us like this,

24
00:01:11,846 --> 00:01:14,563
but really, this is what driving's like.

25
00:01:14,563 --> 00:01:16,722
Right? It's not sunny, it's rainy,

26
00:01:16,722 --> 00:01:19,210
and you want to do anything
other than drive.

27
00:01:19,210 --> 00:01:20,832
And the reason why is this:

28
00:01:20,832 --> 00:01:22,690
Traffic is getting worse.

29
00:01:22,690 --> 00:01:26,196
In America, between 1990 and 2010,

30
00:01:26,196 --> 00:01:29,700
the vehicle miles traveled
increased by 38 percent.

31
00:01:30,213 --> 00:01:32,962
We grew by six percent of roads,

32
00:01:32,962 --> 00:01:34,564
so it's not in your brains.

33
00:01:34,564 --> 00:01:38,840
Traffic really is substantially worse
than it was not very long ago.

34
00:01:38,840 --> 00:01:41,249
And all of this has a very human cost.

35
00:01:41,529 --> 00:01:45,477
So if you take the average commute time
in America, which is about 50 minutes,

36
00:01:45,477 --> 00:01:49,126
you multiply that by the 120 million
workers we have,

37
00:01:49,126 --> 00:01:51,351
that turns out to be
about six billion minutes

38
00:01:51,351 --> 00:01:53,377
wasted in commuting every day.

39
00:01:53,377 --> 00:01:56,204
Now, that's a big number,
so let's put it in perspective.

40
00:01:56,204 --> 00:01:57,978
You take that six billion minutes

41
00:01:57,978 --> 00:02:01,762
and you divide it by the average
life expectancy of a person,

42
00:02:01,762 --> 00:02:04,897
that turns out to be 162 lifetimes

43
00:02:04,897 --> 00:02:07,822
spent every day, wasted,

44
00:02:07,822 --> 00:02:09,866
just getting from A to B.

45
00:02:09,866 --> 00:02:11,596
It's unbelievable.

46
00:02:11,596 --> 00:02:14,440
And then, there are those of us
who don't have the privilege

47
00:02:14,440 --> 00:02:16,112
of sitting in traffic.

48
00:02:16,112 --> 00:02:17,690
So this is Steve.

49
00:02:17,690 --> 00:02:19,455
He's an incredibly capable guy,

50
00:02:19,455 --> 00:02:21,971
but he just happens to be blind,

51
00:02:21,971 --> 00:02:25,188
and that means instead of a 30-minute
drive to work in the morning,

52
00:02:25,188 --> 00:02:29,167
it's a two-hour ordeal
of piecing together bits of public transit

53
00:02:29,167 --> 00:02:31,552
or asking friends and family for a ride.

54
00:02:31,552 --> 00:02:35,221
He doesn't have that same freedom
that you and I have to get around.

55
00:02:35,221 --> 00:02:37,681
We should do something about that.

56
00:02:37,891 --> 00:02:39,648
Now, conventional wisdom would say

57
00:02:39,648 --> 00:02:42,140
that we'll just take
these driver assistance systems

58
00:02:42,140 --> 00:02:45,890
and we'll kind of push them
and incrementally improve them,

59
00:02:45,890 --> 00:02:48,432
and over time, they'll turn
into self-driving cars.

60
00:02:48,432 --> 00:02:50,841
Well, I'm here to tell you
that's like me saying

61
00:02:50,841 --> 00:02:54,898
that if I work really hard at jumping,
one day I'll be able to fly.

62
00:02:54,898 --> 00:02:57,626
We actually need to do
something a little different.

63
00:02:57,626 --> 00:03:00,337
And so I'm going to talk to you
about three different ways

64
00:03:00,337 --> 00:03:03,683
that self-driving systems are different
than driver assistance systems.

65
00:03:03,683 --> 00:03:06,334
And I'm going to start
with some of our own experience.

66
00:03:06,334 --> 00:03:08,587
So back in 2013,

67
00:03:08,587 --> 00:03:11,250
we had the first test
of a self-driving car

68
00:03:11,250 --> 00:03:13,277
where we let regular people use it.

69
00:03:13,277 --> 00:03:15,479
Well, almost regular --
they were 100 Googlers,

70
00:03:15,479 --> 00:03:17,482
but they weren't working on the project.

71
00:03:17,482 --> 00:03:21,103
And we gave them the car and we allowed
them to use it in their daily lives.

72
00:03:21,103 --> 00:03:24,822
But unlike a real self-driving car,
this one had a big asterisk with it:

73
00:03:24,822 --> 00:03:26,326
They had to pay attention,

74
00:03:26,326 --> 00:03:28,959
because this was an experimental vehicle.

75
00:03:28,959 --> 00:03:32,484
We tested it a lot,
but it could still fail.

76
00:03:32,484 --> 00:03:34,543
And so we gave them two hours of training,

77
00:03:34,543 --> 00:03:36,635
we put them in the car,
we let them use it,

78
00:03:36,635 --> 00:03:38,762
and what we heard back
was something awesome,

79
00:03:38,762 --> 00:03:41,286
as someone trying
to bring a product into the world.

80
00:03:41,286 --> 00:03:43,211
Every one of them told us they loved it.

81
00:03:43,211 --> 00:03:46,777
In fact, we had a Porsche driver
who came in and told us on the first day,

82
00:03:46,777 --> 00:03:49,440
"This is completely stupid.
What are we thinking?"

83
00:03:49,850 --> 00:03:52,690
But at the end of it, he said,
"Not only should I have it,

84
00:03:52,690 --> 00:03:55,865
everyone else should have it,
because people are terrible drivers."

85
00:03:57,135 --> 00:03:58,870
So this was music to our ears,

86
00:03:58,870 --> 00:04:02,673
but then we started to look at what
the people inside the car were doing,

87
00:04:02,673 --> 00:04:04,252
and this was eye-opening.

88
00:04:04,252 --> 00:04:06,690
Now, my favorite story is this gentleman

89
00:04:06,690 --> 00:04:10,519
who looks down at his phone
and realizes the battery is low,

90
00:04:10,519 --> 00:04:15,067
so he turns around like this in the car
and digs around in his backpack,

91
00:04:15,067 --> 00:04:17,220
pulls out his laptop,

92
00:04:17,220 --> 00:04:18,787
puts it on the seat,

93
00:04:18,787 --> 00:04:20,551
goes in the back again,

94
00:04:20,551 --> 00:04:23,918
digs around, pulls out
the charging cable for his phone,

95
00:04:23,918 --> 00:04:27,285
futzes around, puts it into the laptop,
puts it on the phone.

96
00:04:27,285 --> 00:04:29,328
Sure enough, the phone is charging.

97
00:04:29,328 --> 00:04:33,322
All the time he's been doing
65 miles per hour down the freeway.

98
00:04:33,322 --> 00:04:35,806
Right? Unbelievable.

99
00:04:35,806 --> 00:04:38,927
So we thought about this and we said,
it's kind of obvious, right?

100
00:04:38,927 --> 00:04:41,190
The better the technology gets,

101
00:04:41,190 --> 00:04:43,311
the less reliable
the driver is going to get.

102
00:04:43,311 --> 00:04:45,707
So by just making the cars
incrementally smarter,

103
00:04:45,707 --> 00:04:48,609
we're probably not going to see
the wins we really need.

104
00:04:48,609 --> 00:04:52,510
Let me talk about something
a little technical for a moment here.

105
00:04:52,510 --> 00:04:54,948
So we're looking at this graph,
and along the bottom

106
00:04:54,948 --> 00:04:57,999
is how often does the car
apply the brakes when it shouldn't.

107
00:04:57,999 --> 00:04:59,620
You can ignore most of that axis,

108
00:04:59,620 --> 00:05:03,339
because if you're driving around town,
and the car starts stopping randomly,

109
00:05:03,339 --> 00:05:05,040
you're never going to buy that car.

110
00:05:05,040 --> 00:05:08,415
And the vertical axis is how often
the car is going to apply the brakes

111
00:05:08,415 --> 00:05:11,464
when it's supposed to
to help you avoid an accident.

112
00:05:11,464 --> 00:05:13,685
Now, if we look at
the bottom left corner here,

113
00:05:13,685 --> 00:05:15,530
this is your classic car.

114
00:05:15,530 --> 00:05:18,663
It doesn't apply the brakes for you,
it doesn't do anything goofy,

115
00:05:18,663 --> 00:05:21,442
but it also doesn't get you
out of an accident.

116
00:05:21,442 --> 00:05:24,460
Now, if we want to bring
a driver assistance system into a car,

117
00:05:24,460 --> 00:05:26,288
say with collision mitigation braking,

118
00:05:26,288 --> 00:05:28,900
we're going to put some package
of technology on there,

119
00:05:28,900 --> 00:05:32,318
and that's this curve, and it's going
to have some operating properties,

120
00:05:32,318 --> 00:05:34,808
but it's never going to avoid
all of the accidents,

121
00:05:34,808 --> 00:05:36,867
because it doesn't have that capability.

122
00:05:36,867 --> 00:05:39,116
But we'll pick some place
along the curve here,

123
00:05:39,116 --> 00:05:42,370
and maybe it avoids half of accidents
that the human driver misses,

124
00:05:42,370 --> 00:05:43,667
and that's amazing, right?

125
00:05:43,667 --> 00:05:46,394
We just reduced accidents on our roads
by a factor of two.

126
00:05:46,394 --> 00:05:50,381
There are now 17,000 less people
dying every year in America.

127
00:05:50,381 --> 00:05:52,401
But if we want a self-driving car,

128
00:05:52,401 --> 00:05:54,708
we need a technology curve
that looks like this.

129
00:05:54,708 --> 00:05:57,307
We're going to have to put
more sensors in the vehicle,

130
00:05:57,307 --> 00:05:59,328
and we'll pick some
operating point up here

131
00:05:59,328 --> 00:06:01,347
where it basically never
gets into a crash.

132
00:06:01,347 --> 00:06:03,790
They'll happen, but very low frequency.

133
00:06:03,790 --> 00:06:06,251
Now you and I could look at this
and we could argue

134
00:06:06,251 --> 00:06:09,856
about whether it's incremental, and
I could say something like "80-20 rule,"

135
00:06:09,856 --> 00:06:12,424
and it's really hard to move up
to that new curve.

136
00:06:12,424 --> 00:06:15,358
But let's look at it
from a different direction for a moment.

137
00:06:15,358 --> 00:06:18,870
So let's look at how often
the technology has to do the right thing.

138
00:06:18,870 --> 00:06:22,376
And so this green dot up here
is a driver assistance system.

139
00:06:22,376 --> 00:06:24,861
It turns out that human drivers

140
00:06:24,861 --> 00:06:27,508
make mistakes that lead
to traffic accidents

141
00:06:27,508 --> 00:06:30,680
about once every 100,000 miles in America.

142
00:06:30,680 --> 00:06:33,847
In contrast, a self-driving system
is probably making decisions

143
00:06:33,847 --> 00:06:37,510
about 10 times per second,

144
00:06:37,510 --> 00:06:38,932
so order of magnitude,

145
00:06:38,932 --> 00:06:41,764
that's about 1,000 times per mile.

146
00:06:41,764 --> 00:06:44,249
So if you compare the distance
between these two,

147
00:06:44,249 --> 00:06:46,849
it's about 10 to the eighth, right?

148
00:06:46,849 --> 00:06:48,614
Eight orders of magnitude.

149
00:06:48,614 --> 00:06:51,423
That's like comparing how fast I run

150
00:06:51,423 --> 00:06:53,629
to the speed of light.

151
00:06:53,629 --> 00:06:57,414
It doesn't matter how hard I train,
I'm never actually going to get there.

152
00:06:57,414 --> 00:06:59,852
So there's a pretty big gap there.

153
00:06:59,852 --> 00:07:03,581
And then finally, there's how
the system can handle uncertainty.

154
00:07:03,581 --> 00:07:06,904
So this pedestrian here might be
stepping into the road, might not be.

155
00:07:06,904 --> 00:07:10,299
I can't tell,
nor can any of our algorithms,

156
00:07:10,310 --> 00:07:12,594
but in the case of
a driver assistance system,

157
00:07:12,594 --> 00:07:15,400
that means it can't take action,
because again,

158
00:07:15,400 --> 00:07:18,739
if it presses the brakes unexpectedly,
that's completely unacceptable.

159
00:07:18,739 --> 00:07:21,872
Whereas a self-driving system
can look at that pedestrian and say,

160
00:07:21,872 --> 00:07:23,762
I don't know what they're about to do,

161
00:07:23,762 --> 00:07:27,524
slow down, take a better look,
and then react appropriately after that.

162
00:07:27,524 --> 00:07:31,226
So it can be much safer than
a driver assistance system can ever be.

163
00:07:31,226 --> 00:07:33,956
So that's enough about
the differences between the two.

164
00:07:33,956 --> 00:07:37,440
Let's spend some time talking about
how the car sees the world.

165
00:07:37,440 --> 00:07:38,692
So this is our vehicle.

166
00:07:38,692 --> 00:07:41,130
It starts by understanding
where it is in the world,

167
00:07:41,130 --> 00:07:43,917
by taking a map and its sensor data
and aligning the two,

168
00:07:43,917 --> 00:07:46,865
and then we layer on top of that
what it sees in the moment.

169
00:07:46,865 --> 00:07:50,520
So here, all the purple boxes you can see
are other vehicles on the road,

170
00:07:50,520 --> 00:07:53,048
and the red thing on the side
over there is a cyclist,

171
00:07:53,048 --> 00:07:55,450
and up in the distance,
if you look really closely,

172
00:07:55,450 --> 00:07:57,244
you can see some cones.

173
00:07:57,244 --> 00:08:00,017
Then we know where the car
is in the moment,

174
00:08:00,017 --> 00:08:03,850
but we have to do better than that:
we have to predict what's going to happen.

175
00:08:03,850 --> 00:08:07,338
So here the pickup truck in top right
is about to make a left lane change

176
00:08:07,338 --> 00:08:09,561
because the road in front of it is closed,

177
00:08:09,561 --> 00:08:11,292
so it needs to get out of the way.

178
00:08:11,292 --> 00:08:13,155
Knowing that one pickup truck is great,

179
00:08:13,155 --> 00:08:15,634
but we really need to know
what everybody's thinking,

180
00:08:15,634 --> 00:08:18,141
so it becomes quite a complicated problem.

181
00:08:18,141 --> 00:08:22,890
And then given that, we can figure out
how the car should respond in the moment,

182
00:08:22,890 --> 00:08:26,756
so what trajectory it should follow, how
quickly it should slow down or speed up.

183
00:08:26,756 --> 00:08:29,821
And then that all turns into
just following a path:

184
00:08:29,821 --> 00:08:33,018
turning the steering wheel left or right,
pressing the brake or gas.

185
00:08:33,018 --> 00:08:35,482
It's really just two numbers
at the end of the day.

186
00:08:35,482 --> 00:08:37,723
So how hard can it really be?

187
00:08:38,433 --> 00:08:40,385
Back when we started in 2009,

188
00:08:40,385 --> 00:08:42,183
this is what our system looked like.

189
00:08:42,183 --> 00:08:45,574
So you can see our car in the middle
and the other boxes on the road,

190
00:08:45,574 --> 00:08:46,845
driving down the highway.

191
00:08:46,845 --> 00:08:50,663
The car needs to understand where it is
and roughly where the other vehicles are.

192
00:08:50,663 --> 00:08:53,092
It's really a geometric
understanding of the world.

193
00:08:53,092 --> 00:08:56,040
Once we started driving
on neighborhood and city streets,

194
00:08:56,040 --> 00:08:58,485
the problem becomes a whole
new level of difficulty.

195
00:08:58,485 --> 00:09:01,979
You see pedestrians crossing in front
of us, cars crossing in front of us,

196
00:09:01,979 --> 00:09:03,790
going every which way,

197
00:09:03,790 --> 00:09:05,317
the traffic lights, crosswalks.

198
00:09:05,317 --> 00:09:08,114
It's an incredibly complicated
problem by comparison.

199
00:09:08,114 --> 00:09:10,217
And then once you have
that problem solved,

200
00:09:10,217 --> 00:09:12,729
the vehicle has to be able
to deal with construction.

201
00:09:12,729 --> 00:09:15,880
So here are the cones on the left
forcing it to drive to the right,

202
00:09:15,880 --> 00:09:18,282
but not just construction
in isolation, of course.

203
00:09:18,282 --> 00:09:22,005
It has to deal with other people moving
through that construction zone as well.

204
00:09:22,005 --> 00:09:25,268
And of course, if anyone's
breaking the rules, the police are there

205
00:09:25,268 --> 00:09:28,890
and the car has to understand that
that flashing light on the top of the car

206
00:09:28,890 --> 00:09:31,995
means that it's not just a car,
it's actually a police officer.

207
00:09:31,995 --> 00:09:34,027
Similarly, the orange box
on the side here,

208
00:09:34,027 --> 00:09:35,136
it's a school bus,

209
00:09:35,136 --> 00:09:37,656
and we have to treat that
differently as well.

210
00:09:38,576 --> 00:09:41,369
When we're out on the road,
other people have expectations:

211
00:09:41,369 --> 00:09:43,149
So, when a cyclist puts up their arm,

212
00:09:43,149 --> 00:09:46,667
it means they're expecting the car
to yield to them and make room for them

213
00:09:46,667 --> 00:09:48,720
to make a lane change.

214
00:09:49,030 --> 00:09:51,203
And when a police officer
stood in the road,

215
00:09:51,203 --> 00:09:53,943
our vehicle should understand
that this means stop,

216
00:09:53,943 --> 00:09:57,449
and when they signal to go,
we should continue.

217
00:09:57,449 --> 00:10:01,210
Now, the way we accomplish this
is by sharing data between the vehicles.

218
00:10:01,210 --> 00:10:02,906
The first, most crude model of this

219
00:10:02,906 --> 00:10:05,019
is when one vehicle
sees a construction zone,

220
00:10:05,019 --> 00:10:08,081
having another know about it
so it can be in the correct lane

221
00:10:08,081 --> 00:10:09,651
to avoid some of the difficulty.

222
00:10:09,651 --> 00:10:12,315
But we actually have a much
deeper understanding of this.

223
00:10:12,315 --> 00:10:15,324
We could take all of the data
that the cars have seen over time,

224
00:10:15,324 --> 00:10:17,700
the hundreds of thousands
of pedestrians, cyclists,

225
00:10:17,700 --> 00:10:19,487
and vehicles that have been out there

226
00:10:19,487 --> 00:10:21,182
and understand what they look like

227
00:10:21,182 --> 00:10:24,013
and use that to infer
what other vehicles should look like

228
00:10:24,013 --> 00:10:25,939
and other pedestrians should look like.

229
00:10:25,939 --> 00:10:28,960
And then, even more importantly,
we could take from that a model

230
00:10:28,960 --> 00:10:31,290
of how we expect them
to move through the world.

231
00:10:31,290 --> 00:10:34,253
So here the yellow box is a pedestrian
crossing in front of us.

232
00:10:34,253 --> 00:10:36,503
Here the blue box is a cyclist
and we anticipate

233
00:10:36,503 --> 00:10:39,815
that they're going to nudge out
and around the car to the right.

234
00:10:40,115 --> 00:10:42,207
Here there's a cyclist
coming down the road

235
00:10:42,207 --> 00:10:45,693
and we know they're going to continue
to drive down the shape of the road.

236
00:10:45,693 --> 00:10:47,560
Here somebody makes a right turn,

237
00:10:47,560 --> 00:10:50,920
and in a moment here, somebody's
going to make a U-turn in front of us,

238
00:10:50,920 --> 00:10:53,534
and we can anticipate that behavior
and respond safely.

239
00:10:53,534 --> 00:10:56,262
Now, that's all well and good
for things that we've seen,

240
00:10:56,262 --> 00:10:59,127
but of course, you encounter
lots of things that you haven't

241
00:10:59,127 --> 00:11:00,358
seen in the world before.

242
00:11:00,358 --> 00:11:02,099
And so just a couple of months ago,

243
00:11:02,099 --> 00:11:04,334
our vehicles were driving
through Mountain View,

244
00:11:04,334 --> 00:11:05,978
and this is what we encountered.

245
00:11:05,978 --> 00:11:08,060
This is a woman in an electric wheelchair

246
00:11:08,060 --> 00:11:10,677
chasing a duck in circles on the road.
(Laughter)

247
00:11:10,677 --> 00:11:13,788
Now it turns out, there is nowhere
in the DMV handbook

248
00:11:13,788 --> 00:11:16,033
that tells you how to deal with that,

249
00:11:16,033 --> 00:11:18,176
but our vehicles were able
to encounter that,

250
00:11:18,176 --> 00:11:20,431
slow down, and drive safely.

251
00:11:20,431 --> 00:11:22,472
Now, we don't have to deal
with just ducks.

252
00:11:22,472 --> 00:11:26,180
Watch this bird fly across in front of us.
The car reacts to that.

253
00:11:26,180 --> 00:11:27,795
Here we're dealing with a cyclist

254
00:11:27,795 --> 00:11:31,085
that you would never expect to see
anywhere other than Mountain View.

255
00:11:31,085 --> 00:11:33,153
And of course, we have
to deal with drivers,

256
00:11:33,153 --> 00:11:36,868
even the very small ones.

257
00:11:36,868 --> 00:11:40,999
Watch to the right as someone
jumps out of this truck at us.

258
00:11:42,460 --> 00:11:45,389
And now, watch the left as the car
with the green box decides

259
00:11:45,389 --> 00:11:48,714
he needs to make a right turn
at the last possible moment.

260
00:11:48,714 --> 00:11:51,565
Here, as we make a lane change,
the car to our left decides

261
00:11:51,565 --> 00:11:55,118
it wants to as well.

262
00:11:55,118 --> 00:11:57,811
And here, we watch a car
blow through a red light

263
00:11:57,811 --> 00:11:59,901
and yield to it.

264
00:11:59,901 --> 00:12:03,755
And similarly, here, a cyclist
blowing through that light as well.

265
00:12:03,755 --> 00:12:06,501
And of course,
the vehicle responds safely.

266
00:12:06,501 --> 00:12:09,102
And of course, we have people
who do I don't know what

267
00:12:09,102 --> 00:12:12,925
sometimes on the road, like this guy
pulling out between two self-driving cars.

268
00:12:12,925 --> 00:12:14,970
You have to ask, "What are you thinking?"

269
00:12:14,970 --> 00:12:16,182
(Laughter)

270
00:12:16,182 --> 00:12:18,703
Now, I just fire-hosed you
with a lot of stuff there,

271
00:12:18,703 --> 00:12:21,353
so I'm going to break one of these
down pretty quickly.

272
00:12:21,353 --> 00:12:24,293
So what we're looking at is the scene
with the cyclist again,

273
00:12:24,293 --> 00:12:27,784
and you might notice in the bottom,
we can't actually see the cyclist yet,

274
00:12:27,784 --> 00:12:30,288
but the car can: it's that little
blue box up there,

275
00:12:30,288 --> 00:12:32,369
and that comes from the laser data.

276
00:12:32,369 --> 00:12:34,787
And that's not actually
really easy to understand,

277
00:12:34,787 --> 00:12:38,371
so what I'm going to do is I'm going
to turn that laser data and look at it,

278
00:12:38,371 --> 00:12:41,400
and if you're really good at looking
at laser data, you can see

279
00:12:41,400 --> 00:12:42,887
a few dots on the curve there,

280
00:12:42,887 --> 00:12:45,259
right there, and that blue box
is that cyclist.

281
00:12:45,259 --> 00:12:46,408
Now as our light is red,

282
00:12:46,408 --> 00:12:48,600
the cyclist's light
has turned yellow already,

283
00:12:48,600 --> 00:12:51,038
and if you squint, you can see that
in the imagery.

284
00:12:51,038 --> 00:12:54,324
But the cyclist, we see, is going
to proceed through the intersection.

285
00:12:54,324 --> 00:12:56,718
Our light has now turned green,
his is solidly red,

286
00:12:56,718 --> 00:13:01,010
and we now anticipate that this bike
is going to come all the way across.

287
00:13:01,010 --> 00:13:04,752
Unfortunately the other drivers next to us
were not paying as much attention.

288
00:13:04,752 --> 00:13:07,909
They started to pull forward,
and fortunately for everyone,

289
00:13:07,909 --> 00:13:10,920
this cyclists reacts, avoids,

290
00:13:10,920 --> 00:13:13,111
and makes it through the intersection.

291
00:13:13,111 --> 00:13:14,679
And off we go.

292
00:13:14,679 --> 00:13:17,627
Now, as you can see, we've made
some pretty exciting progress,

293
00:13:17,627 --> 00:13:19,529
and at this point we're pretty convinced

294
00:13:19,529 --> 00:13:21,539
this technology is going
to come to market.

295
00:13:21,539 --> 00:13:26,322
We do three million miles of testing
in our simulators every single day,

296
00:13:26,322 --> 00:13:29,011
so you can imagine the experience
that our vehicles have.

297
00:13:29,011 --> 00:13:31,875
We are looking forward to having
this technology on the road,

298
00:13:31,875 --> 00:13:34,765
and we think the right path
is to go through the self-driving

299
00:13:34,765 --> 00:13:36,609
rather than driver assistance approach

300
00:13:36,609 --> 00:13:39,230
because the urgency is so large.

301
00:13:39,230 --> 00:13:41,623
In the time I have given this talk today,

302
00:13:41,623 --> 00:13:44,758
34 people have died on America's roads.

303
00:13:44,758 --> 00:13:47,126
How soon can we bring it out?

304
00:13:47,126 --> 00:13:50,958
Well, it's hard to say because
it's a really complicated problem,

305
00:13:50,958 --> 00:13:53,172
but these are my two boys.

306
00:13:53,172 --> 00:13:56,795
My oldest son is 11, and that means
in four and a half years,

307
00:13:56,795 --> 00:13:59,372
he's going to be able
to get his driver's license.

308
00:13:59,372 --> 00:14:02,576
My team and I are committed
to making sure that doesn't happen.

309
00:14:02,576 --> 00:14:04,480
Thank you.

310
00:14:04,480 --> 00:14:08,147
(Laughter) (Applause)

311
00:14:09,110 --> 00:14:11,678
Chris Anderson: Chris,
I've got a question for you.

312
00:14:11,678 --> 00:14:14,487
Chris Urmson: Sure.

313
00:14:14,487 --> 00:14:18,411
CA: So certainly, the mind of your cars
is pretty mind-boggling.

314
00:14:18,411 --> 00:14:22,870
On this debate between
driver-assisted and fully driverless --

315
00:14:22,870 --> 00:14:25,911
I mean, there's a real debate
going on out there right now.

316
00:14:25,911 --> 00:14:28,744
So some of the companies,
for example, Tesla,

317
00:14:28,744 --> 00:14:30,903
are going the driver-assisted route.

318
00:14:30,903 --> 00:14:36,151
What you're saying is that
that's kind of going to be a dead end

319
00:14:36,151 --> 00:14:41,607
because you can't just keep improving
that route and get to fully driverless

320
00:14:41,607 --> 00:14:45,137
at some point, and then a driver
is going to say, "This feels safe,"

321
00:14:45,137 --> 00:14:47,784
and climb into the back,
and something ugly will happen.

322
00:14:47,784 --> 00:14:50,460
CU: Right. No, that's exactly right,
and it's not to say

323
00:14:50,460 --> 00:14:53,997
that the driver assistance systems
aren't going to be incredibly valuable.

324
00:14:53,997 --> 00:14:56,055
They can save a lot of lives
in the interim,

325
00:14:56,055 --> 00:14:59,888
but to see the transformative opportunity
to help someone like Steve get around,

326
00:14:59,888 --> 00:15:01,857
to really get to the end case in safety,

327
00:15:01,857 --> 00:15:04,336
to have the opportunity
to change our cities

328
00:15:04,336 --> 00:15:08,540
and move parking out and get rid of
these urban craters we call parking lots,

329
00:15:08,540 --> 00:15:09,780
it's the only way to go.

330
00:15:09,780 --> 00:15:12,498
CA: We will be tracking your progress
with huge interest.

331
00:15:12,498 --> 00:15:16,730
Thanks so much, Chris.
CU: Thank you. (Applause)