WEBVTT

1
00:00:00.080 --> 00:00:03.439
I can add something. The general
principle here is that any part of this

2
00:00:03.480 --> 00:00:07.000
could fail and the call will keep
driving. So you could have cameras fail,

3
00:00:07.040 --> 00:00:11.000
you could have power circuits fail,
you could have one of the Tesla

4
00:00:11.320 --> 00:00:15.599
full self driving computer chips fail,
car keeps driving. The probability of this

5
00:00:15.640 --> 00:00:20.679
computer failing is substantial, lower than
somebody losing consciousness. That's the key metric,

6
00:00:20.719 --> 00:00:23.239
at least in order of magnitude.
At first, it seems improbable.

7
00:00:23.320 --> 00:00:26.640
How could it be that Tesla,
who has never designed a chip before,

8
00:00:26.879 --> 00:00:30.079
would designed the best chip in the
world. But that is objectively what has

9
00:00:30.120 --> 00:00:34.039
occurred. Not best by a small
margin, best by a huge login.

10
00:00:34.159 --> 00:00:37.679
It's in the cars right now.
Old Tesla's being produced right now have this

11
00:00:37.759 --> 00:00:43.079
computer. We switched over from the
Nvidious solution for sn X about a month

12
00:00:43.119 --> 00:00:47.320
ago, and you switched over Model
three about ten days ago. All cars

13
00:00:47.320 --> 00:00:52.240
being produced have the have all the
hardware necessary, compute and otherwise for full

14
00:00:52.280 --> 00:00:56.159
self driving. I'll say that again. Old Tesla cars being produced right now

15
00:00:56.439 --> 00:01:00.840
have everything necessary for full self driving. All you need to do is improve

16
00:01:00.880 --> 00:01:04.239
the software and later today you will
drive the cars with the development version of

17
00:01:04.280 --> 00:01:08.840
the improved software and you will see
for yourselves. We finished this design like

18
00:01:08.879 --> 00:01:12.200
maybe one and a half two years
ago and began design of the next generation.

19
00:01:12.400 --> 00:01:15.239
We're not talking about the next generation
today, but we're about halfway through

20
00:01:15.239 --> 00:01:18.879
it. All the things that are
obvious for our next generation chip. We're

21
00:01:18.879 --> 00:01:23.319
doing the strategy here and started basically
three level three years ago was designable to

22
00:01:23.400 --> 00:01:27.959
computer that is fully optimized and aiming
for full self driving. Software that is

23
00:01:29.200 --> 00:01:32.920
designed to work specifically on that computer
and get the most out of that computer.

24
00:01:33.079 --> 00:01:36.879
So you have tailored hardware that is
a master of one trade. Self

25
00:01:36.920 --> 00:01:40.840
Driving and Video is a great company, but they have many customers and so

26
00:01:41.159 --> 00:01:45.319
as they apply their resources, they
need to generalize solution. We care about

27
00:01:45.319 --> 00:01:48.920
one thing, self driving, so
that it was designed to do that incredibly.

28
00:01:49.040 --> 00:01:53.040
The software is also designed to run
on that hardware incredibly well. And

29
00:01:53.079 --> 00:01:55.959
the combination of the software in the
hardware I think is unbeatable. We' we're

30
00:01:56.000 --> 00:02:00.319
going to explained to you today is
that lighter is a fulls errand and anyone

31
00:02:00.560 --> 00:02:05.760
relying on light R is doomed.
Doomed expensive sensors that are unnecessary. It's

32
00:02:05.799 --> 00:02:08.439
like having a whole bunch of expensive
epeties, but one appendix is bad.

33
00:02:08.439 --> 00:02:12.120
We only want to put a whole
bunch of them. That's ridiculous. You'll

34
00:02:12.120 --> 00:02:15.240
see if you're driving for an hour
in a city and you had a solution

35
00:02:15.319 --> 00:02:20.199
hypothetically that it was a kilo,
what you'd lose four miles a Model three,

36
00:02:20.439 --> 00:02:23.840
So if you're only going, say
twelve miles an hour, then that's

37
00:02:23.879 --> 00:02:27.719
there would be a twenty five percent
impact on range in city. It's basically

38
00:02:27.840 --> 00:02:30.120
power is the power. The power
of the system has a massive impact on

39
00:02:30.159 --> 00:02:34.719
city range, which is where we
think most of the robotaxi market will be.

40
00:02:35.000 --> 00:02:37.639
So power is extremely important. We
don't want to spalok too much about

41
00:02:37.680 --> 00:02:40.680
the next generationship, but it's it'll
be at least, let's say, three

42
00:02:40.680 --> 00:02:45.840
times better than the current system.
And I think if if somebody started today

43
00:02:46.039 --> 00:02:49.319
and they were really good, they
might have something like what we have right

44
00:02:49.360 --> 00:02:53.080
now in three years, but in
two years we'll have something three times better.

45
00:02:53.319 --> 00:02:57.280
The thing that's I think a very
powerful sustainable advantage for us is the

46
00:02:57.319 --> 00:03:00.759
fleet. Nobody has the fleet.
Those weights are constantly updated and improved based

47
00:03:00.800 --> 00:03:05.719
on billions of miles driven. Tesla
has one hundred times small cars with the

48
00:03:05.719 --> 00:03:08.080
full self driving hardware that everyone else
combined we have by the end of this

49
00:03:08.159 --> 00:03:12.680
quarter, will have five hundred thousand
cars worth the full eight camera set up

50
00:03:12.719 --> 00:03:15.879
fall ultrasonic. Someone will still be
on hardware too, but we still have

51
00:03:15.960 --> 00:03:19.080
the data gathering ability, and then
by a year from now, we'll have

52
00:03:19.240 --> 00:03:23.960
over a million cars with full self
driving computer hardware everything. It's just a

53
00:03:23.960 --> 00:03:27.560
massive data advantage. It's similar to
how like the Google search engine has a

54
00:03:27.560 --> 00:03:31.919
massive advantage because people use it and
people are programming effectively programmed Google with the

55
00:03:32.000 --> 00:03:37.800
queries and the results. The simulator, we have quite a good simulation too,

56
00:03:38.080 --> 00:03:40.319
but it just does not capture the
long tail of weird things that happen

57
00:03:40.360 --> 00:03:44.400
in the real world. If the
simulation fully captured the real world, that

58
00:03:44.439 --> 00:03:46.120
would be proof that we're living in
a simulation. I think, yeah,

59
00:03:46.360 --> 00:03:50.960
it doesn't. I wish, but
simulations do not capture the real world.

60
00:03:51.039 --> 00:03:53.080
The real world is really weird and
messy. You need the cars on the

61
00:03:53.120 --> 00:03:58.560
road simulation. You're fundamentally grading your
own homework. If you know that you're

62
00:03:58.560 --> 00:04:00.439
going to simulate it, Okay,
you can definitely solve for it. But

63
00:04:00.479 --> 00:04:03.599
as Andrew saying, you don't know
what you don't know. The world is

64
00:04:03.879 --> 00:04:09.759
very weird and has millions of corner
cases. And if somebody can produce a

65
00:04:09.919 --> 00:04:14.520
self driving simulation that accurately matches reality, that in itself would be in a

66
00:04:14.879 --> 00:04:19.360
monumental achievement of human capability. They
can't. There's no way. Everyone's training

67
00:04:19.600 --> 00:04:24.360
the network all the time, is
what it amounts to. Whether autopoler is

68
00:04:24.439 --> 00:04:27.720
on orall. The network is being
trained. Every mile that's driven for the

69
00:04:27.720 --> 00:04:30.480
car that's hard or tour above is
training the network. The crazy thing is

70
00:04:30.519 --> 00:04:35.040
the network is predicting paths it can't
even see with incredibly high accuracy. It

71
00:04:35.040 --> 00:04:39.879
can't see around the corner, but
saying the probability of that curve is extremely

72
00:04:40.000 --> 00:04:42.639
high. So that's the path,
and it nails it. You will see

73
00:04:42.639 --> 00:04:45.720
that in the cars today. We're
going to turn on augmented vision so you

74
00:04:45.759 --> 00:04:48.560
can see the lane lines and the
path predictions of the cars overlaid on the

75
00:04:48.600 --> 00:04:54.360
video. The car is an inference
optimized computer. But we do have a

76
00:04:54.680 --> 00:04:57.839
major program at Tesla, which we
don't have enough time to talk about today,

77
00:04:57.879 --> 00:05:01.600
called Dojo. That's a superpower full
training computer. The goal dojo will

78
00:05:01.639 --> 00:05:04.959
be to be able to take in
vast amounts of data and train at a

79
00:05:05.040 --> 00:05:11.879
video level and do unsupervised massive training
of vast amounts of video with the dojo

80
00:05:11.959 --> 00:05:15.240
program dojo computer. That's for another
day. We're just being more conservative right

81
00:05:15.279 --> 00:05:19.839
now, and then as we gain
a higher higher confidence will allow users to

82
00:05:19.959 --> 00:05:23.800
select a more aggressive mode. That'll
be up to the user. But in

83
00:05:23.839 --> 00:05:26.920
the more aggressive modes, in trying
to merge in traffic, there is a

84
00:05:27.000 --> 00:05:30.079
slight no matter how many, there's
a slight chance of like vendor bender,

85
00:05:30.120 --> 00:05:33.399
not a serious accident, but you
basically will have a choice of do you

86
00:05:33.439 --> 00:05:38.519
want to have a non zero chance
of a vendor vender on freeway traffic,

87
00:05:38.560 --> 00:05:42.240
which is unfortunately the only way to
navigate it LA traffic. The car can

88
00:05:42.279 --> 00:05:46.079
operate if it's completely disconnected from the
fleet. It just it uploads the training

89
00:05:46.279 --> 00:05:50.079
that's better and better as the free
gets better and better. So simply if

90
00:05:50.079 --> 00:05:54.279
you're just connected from the fleet from
that point onwards, it would stop getting

91
00:05:54.279 --> 00:05:57.800
better, but it would still function
fine. The computer power in the full

92
00:05:57.839 --> 00:06:01.199
Sell driving computer is incredible, and
it maybe we should have mentioned that if

93
00:06:01.199 --> 00:06:05.600
it had never seen that road before, it would still have made those predictions

94
00:06:05.600 --> 00:06:09.040
provided it was a road in the
United States. You're all going to dump

95
00:06:09.079 --> 00:06:11.759
lighter. That's my prediction. Walk
my words. I should point out that

96
00:06:12.040 --> 00:06:15.120
I don't actually superheate lighter as much
as may sound. But at SpaceX,

97
00:06:15.560 --> 00:06:19.879
SpaceX Dragon uses light r to navigate
to the space station and dock. Not

98
00:06:19.959 --> 00:06:25.000
only that, we SpaceX developed its
own lighter from scratch to do that,

99
00:06:25.319 --> 00:06:29.759
and I bear ahead of that effort
personally because in that scenario, lighter makes

100
00:06:29.800 --> 00:06:33.079
sense. And in cars it's freaking
stupid. It's expensive and unnecessary, and

101
00:06:33.160 --> 00:06:36.480
as Andrew was saying, once you
saw a vision, it's a worthless so

102
00:06:36.480 --> 00:06:41.360
you have expensive hardware that's worthless on
the car that we do have a forward

103
00:06:41.439 --> 00:06:46.279
radar which is low cost and is
helpful, especially for occlusion situations. So

104
00:06:46.319 --> 00:06:48.759
if there's like fog or dust or
just snow, the radar can see through

105
00:06:48.800 --> 00:06:54.600
that. If you're going to use
active photon generation, don't use visible wavelength

106
00:06:54.759 --> 00:06:58.759
because once with passive optical you've taken
care of all visible way stuff. You

107
00:06:58.800 --> 00:07:00.399
want to if you want, I
want to use a wave link that is

108
00:07:00.399 --> 00:07:04.839
inclusion penetrating like radar. So liner
is just active photon generation in the visual

109
00:07:04.839 --> 00:07:08.959
spectrum. If you're going to do
active photon generation, do it outside the

110
00:07:09.000 --> 00:07:12.600
visual spectrum in the radar spectrum.
So like at three point eight millimeters versus

111
00:07:12.639 --> 00:07:15.240
four hundred one hundred antime, we're
going to have much better occlusion penetration.

112
00:07:15.480 --> 00:07:18.639
And that's why we have a forward
radar, and then we also have twelve

113
00:07:18.720 --> 00:07:25.199
ultrasonics for near field information in addition
to the eight cameras and the forward radar.

114
00:07:25.319 --> 00:07:27.720
You only need the radar and the
forward direction because that's the only direction

115
00:07:27.839 --> 00:07:30.000
going real fast. So we've gone
over this multiple times, always sure we

116
00:07:30.040 --> 00:07:32.920
have the right sense of suite shure
we add anything more. No, we

117
00:07:32.959 --> 00:07:36.040
have four hundred and twenty five thousand
cars with hardware tuned beyond, which is

118
00:07:36.199 --> 00:07:40.959
means they've got all eight cameras right, the radar and ultrasonics, and they've

119
00:07:41.000 --> 00:07:45.160
got at least a video computer,
which is enough to essentially figure out what

120
00:07:45.199 --> 00:07:48.720
information is important what is not,
compress the information that is important to the

121
00:07:48.720 --> 00:07:53.160
most salent elements and uploaded to the
network for training. So it's a massive

122
00:07:53.240 --> 00:07:57.360
compression of real world data. I
suppose it could possibly be used for something

123
00:07:57.360 --> 00:08:00.959
besides self driving. We've been too
focused on self drive. So as we

124
00:08:01.000 --> 00:08:05.079
get that really nailed, maybe there's
going to be some other use for millions

125
00:08:05.120 --> 00:08:09.040
and then tens of millions of computers
with hardware three or full sub driving computer.

126
00:08:09.240 --> 00:08:11.279
Yeah, maybe there would be.
It could be. Maybe there's some

127
00:08:11.279 --> 00:08:13.519
sort of eight angle here as possible. There are a number of important signals

128
00:08:13.519 --> 00:08:16.720
as under saying lanelines are one of
those things, But one of the most

129
00:08:16.720 --> 00:08:20.360
important signals is drive space. So
what is drivable space and what is not

130
00:08:20.439 --> 00:08:24.879
drivable space? And what actually really
matters the most is drivable space more than

131
00:08:24.920 --> 00:08:28.839
lanelines. And the prediction of driveable
space is extremely good, and I think,

132
00:08:30.079 --> 00:08:33.200
especially after this upcoming winter, will
be incredible. It will be like,

133
00:08:33.240 --> 00:08:37.720
how could it possibly be that good? That's crazy. It's extremely important

134
00:08:37.759 --> 00:08:41.480
that things not be rigidly tied to
GPS because GPS eric can vary quite a

135
00:08:41.480 --> 00:08:48.279
bit, and the actual situation for
a road can vary quite a bit,

136
00:08:48.399 --> 00:08:52.000
so that reconstruction that could be a
detour, and if the car is using

137
00:08:52.039 --> 00:08:54.600
GPS as primary, this is a
real bad situation. Is asking for trouble.

138
00:08:54.720 --> 00:08:58.639
It's fine to use GPS for like
tips and tricks, so you can

139
00:08:58.759 --> 00:09:01.840
drive your home neighborhood better than a
neighborhood in some other country or some other

140
00:09:01.879 --> 00:09:05.360
part of the country. So you
know your own neighborhood well and you use

141
00:09:05.440 --> 00:09:09.759
like the knowledge of your neighborhood to
drive with more confidence, to maybe have

142
00:09:11.000 --> 00:09:16.279
counterintuitive shortcuts and that kind of thing. But the GPS overlay data should only

143
00:09:16.320 --> 00:09:18.759
be helpful, but never primary.
If it ever primary, it is a

144
00:09:18.759 --> 00:09:24.039
problem. High precision GPS maps and
lanes are a really bad idea. The

145
00:09:24.080 --> 00:09:28.279
system becomes extremely brittle, so any
change like this might any change to the

146
00:09:28.320 --> 00:09:33.720
system makes it it can't adapt.
So if it locks onto GPS and high

147
00:09:33.759 --> 00:09:37.840
precision lane lines and does not allow
a vision override, In fact, vision

148
00:09:37.840 --> 00:09:41.960
should be the thing that does everything, and then like lane lines are a

149
00:09:41.000 --> 00:09:45.799
guideline, but they're not the main
thing. We briefly balked up the tree

150
00:09:45.879 --> 00:09:48.679
of high precision lane lines and then
realized that was a huge mistake and reversed

151
00:09:48.720 --> 00:09:52.799
it out. It's not good.
There's three steps to self driving, because

152
00:09:52.799 --> 00:09:56.120
it's being future complete. Then there's
being future complete to the degree that where

153
00:09:56.120 --> 00:09:58.399
we think that the person in the
car does not need to pay attention and

154
00:09:58.440 --> 00:10:03.279
then there's at a reliability level,
where we've also convinced regulators that is true.

155
00:10:03.399 --> 00:10:05.559
So it's there's three levels. We
expect to be set feature complete and

156
00:10:05.600 --> 00:10:09.759
self driving this year, and we
expect to be confident enough from our standpoint

157
00:10:09.919 --> 00:10:13.080
to say that we think people do
not need to touch the wheel look out

158
00:10:13.080 --> 00:10:18.120
of the window sometime probably around a
second quarter of next year, and then

159
00:10:18.240 --> 00:10:22.360
we start to expect to get regulatory
approval at least in some jurisdictions for that

160
00:10:22.639 --> 00:10:26.039
towards the end of next year.
That's roughly the timeline that I expect things

161
00:10:26.080 --> 00:10:30.080
to go on, and probably for
trucks, the tuning will be approved by

162
00:10:30.120 --> 00:10:33.200
regulators before anything else, and you
could have, maybe if you're a long

163
00:10:33.240 --> 00:10:37.279
haul doing long haul freight, you
could have one driver in the front and

164
00:10:37.279 --> 00:10:41.919
then have four semis trailing behind in
a platooning manner. And I think that

165
00:10:41.080 --> 00:10:45.639
probably the regulators will be quicker to
approve that than other things. If in

166
00:10:45.679 --> 00:10:48.840
order to have a self driving car
or robotaxi, you really need redundancy throughout

167
00:10:48.879 --> 00:10:52.480
the vehicle at the hardware level.
Starting in as it was October twenty sixteen,

168
00:10:52.679 --> 00:10:58.120
all cars made by Tesla have redundant
power steering, So we've redundant motors

169
00:10:58.120 --> 00:11:01.559
and the power steering, so any
one failure of if I'm the motor fails,

170
00:11:01.559 --> 00:11:03.759
the car can still steer. All
of the power and data lines have

171
00:11:03.840 --> 00:11:07.200
redundancy, so you can sever any
given power line or any data line and

172
00:11:07.200 --> 00:11:11.279
the call will keep driving the auxiliary
power system. Even if the main pack

173
00:11:11.440 --> 00:11:15.519
you lose complete power in the main
pack, the car is capable of steering

174
00:11:15.600 --> 00:11:18.879
and braking at using the auxiliary power
system, so you can completely lose the

175
00:11:18.919 --> 00:11:22.399
main pack and the car is safe. The whole system, if from a

176
00:11:22.440 --> 00:11:26.759
hardware standpoint, has been signed to
be a robot taxi since basically October twenty

177
00:11:26.759 --> 00:11:31.639
sixteen, so when we roll that
hardware autopowered version two, but we do

178
00:11:31.679 --> 00:11:35.720
not expect to upgrade cars made before
that. We think it would actually cost

179
00:11:35.759 --> 00:11:39.759
more to make a new car than
to upgrade the cars, just to give

180
00:11:39.759 --> 00:11:41.480
you a sense of how hard it
is to do this. Unless it's designed

181
00:11:41.480 --> 00:11:43.679
in, it's not worth it.
So we've gone through the future of self

182
00:11:43.759 --> 00:11:48.679
driving. Whereas it's hardware, it's
vision, and then there's a lot of

183
00:11:48.759 --> 00:11:52.120
software, and there's the software problem
here should not be Minimi. It's a

184
00:11:52.120 --> 00:11:56.720
massive software problem that managing vast amounts
of data, training against the data,

185
00:11:56.919 --> 00:11:58.679
how do you control the car based
on the vision. It's a very difficult

186
00:11:58.720 --> 00:12:03.399
software problem. So going after going
over just like teslim massive plan. Obviously,

187
00:12:03.440 --> 00:12:05.200
we've made a bunch of forward looking
statements, as they call it.

188
00:12:05.759 --> 00:12:09.879
And but let's go through some of
our other forward looking statements that we've made.

189
00:12:09.039 --> 00:12:11.759
No way back when we created the
company, we said we build the

190
00:12:11.799 --> 00:12:15.559
Tesla Road, Sir. They said
it was impossible and that even if we

191
00:12:15.559 --> 00:12:18.240
did build it, nobody were buy
it. It was like universal opinion was

192
00:12:18.279 --> 00:12:22.080
that building an electric car was extremely
dumb and would fail. I agree with

193
00:12:22.120 --> 00:12:26.480
them that probably failure was high,
but that this was important. So we

194
00:12:26.519 --> 00:12:30.159
built the Tesla Roads, going into
production in two thousand and eight and shipping

195
00:12:30.200 --> 00:12:33.360
that car it's not collector's item.
Then we would build a more affordable car

196
00:12:33.399 --> 00:12:35.320
with the Model S. We did
that. Again, we were told that

197
00:12:35.279 --> 00:12:39.320
it's impossible. I was called a
fraud and a liar and it's not going

198
00:12:39.360 --> 00:12:43.360
to happen. This is all untrue, Okay. Famous last words now is

199
00:12:43.600 --> 00:12:46.879
we were into production with the Model
S and twenty twelve exceeded all expectations.

200
00:12:46.960 --> 00:12:52.639
There is still in twenty nineteen no
car that can compete with the Model s

201
00:12:52.720 --> 00:12:56.000
of twenty twelve. It's seven years
later. Were waiting, so we'd build

202
00:12:56.039 --> 00:13:00.600
on an affordable car, maybe highly
affords affordable, more affordable. With the

203
00:13:00.600 --> 00:13:03.120
Model three we built the Model three, we're in production, said we'd get

204
00:13:03.159 --> 00:13:05.919
over five thousand cars a week for
Model three. At this point, five

205
00:13:05.960 --> 00:13:09.639
thousand cars week is a walk in
the park for us. It's not even

206
00:13:09.679 --> 00:13:11.600
hard. So we do large scale
solar, which we did through the Sol

207
00:13:11.679 --> 00:13:15.919
City acquisition, and that we develop
into play the solar roof, which is

208
00:13:15.960 --> 00:13:18.080
going really well. We're now on
version of three of the Solo tail roof,

209
00:13:18.200 --> 00:13:22.519
and we expect to will out production
of the Solo Tawer roof significantly later

210
00:13:22.559 --> 00:13:26.360
this year. I have it on
my house and it's great. And I

211
00:13:26.480 --> 00:13:30.919
make the power roll and the power
pack, and we made the power role

212
00:13:30.960 --> 00:13:33.480
and power pack. In fact that
the power pack is now deployed in massive

213
00:13:33.519 --> 00:13:39.360
grid scale utility systems around the world, including the largest operating battery projects in

214
00:13:39.399 --> 00:13:43.720
the world that at above one hundred
megawatts, and in the next or probably

215
00:13:43.279 --> 00:13:46.799
the next year two years at the
most. We expect to have a gigawatt

216
00:13:46.840 --> 00:13:52.159
scale battery project completed. So all
these things I said we would do them,

217
00:13:52.240 --> 00:13:52.960
we did it, so we'ld do
it. We did it. We're

218
00:13:52.960 --> 00:13:56.159
going to do the rover taxi thing
too. Only criticism and it's a fair

219
00:13:56.200 --> 00:14:00.039
one. And sometimes I'm not on
time, but I get it done and

220
00:14:00.120 --> 00:14:03.000
the Tesla team gets it done.
So what we're going to do this year

221
00:14:03.279 --> 00:14:07.200
is we're going to reach a combined
production of ten thousand a week between SX

222
00:14:07.200 --> 00:14:09.559
and three. Feel very confident about
that, and we feel very confident about

223
00:14:09.559 --> 00:14:13.480
being feature complete with self driving.
Next year, we'll expound the product line

224
00:14:13.519 --> 00:14:18.559
with Model Y and Semi, and
we expect to have the first operating robotaxis

225
00:14:18.639 --> 00:14:22.559
next year with no one in them
next year. It's always difficult to like

226
00:14:22.600 --> 00:14:26.559
when things are on an exponential at
an exponential rate of improvements. It's very

227
00:14:26.559 --> 00:14:30.480
difficult to wrap one's mind around it
because we're used to extrapolating on a linear

228
00:14:30.519 --> 00:14:33.840
basis. But when you've got massive
amounts of as the hardware, massive amounts

229
00:14:33.840 --> 00:14:39.320
of hardware on the road, the
cumulative data is increasing exponentially. Software is

230
00:14:39.360 --> 00:14:43.519
getting better at an exponential rate.
I feel very confident in predicting autonomous rover

231
00:14:43.600 --> 00:14:48.080
taxis for Tesla next year. Not
not an old jurisdictions because we won't have

232
00:14:48.120 --> 00:14:52.360
regulatory approval everywhere, but I'm confident
we'll have at least regulatory approval somewhere literally

233
00:14:52.480 --> 00:14:56.320
next year. So any customer will
be able to add or remove their car

234
00:14:56.440 --> 00:14:58.799
to the Tesla network, so expect
this to operate. Is like a combination

235
00:14:58.840 --> 00:15:03.559
of maybe the Uber and Airbnb model. So if you own the car,

236
00:15:03.679 --> 00:15:05.799
you can add or subtract it to
the Tesla network, and Tesla would take

237
00:15:07.279 --> 00:15:11.679
twenty five to thirty percent of the
revenue. And then in places where there

238
00:15:11.679 --> 00:15:15.559
aren't enough people sharing their cars,
we would just have dedicated to Tesla vehicles.

239
00:15:15.840 --> 00:15:16.879
Well, when you use the car, we'll show you our right sharing

240
00:15:16.879 --> 00:15:20.480
app, so you're able to be
able to summon the car from the parking

241
00:15:20.519 --> 00:15:22.279
lot, get in and go for
a drive. It's really simple to just

242
00:15:22.320 --> 00:15:26.120
take the same Tesla app that currently
have. We'll just do it. We'll

243
00:15:26.279 --> 00:15:28.879
update the app and add some in
Tesla or commit your car to the fleet,

244
00:15:30.000 --> 00:15:33.399
so see that some in your car, or add some many Tesla,

245
00:15:33.600 --> 00:15:35.279
or add your add or subtract your
car to the fleet. You'll be able

246
00:15:35.279 --> 00:15:39.879
to do that from your phone.
So we see potential for smoothing out the

247
00:15:39.919 --> 00:15:43.759
demand distpution curve and having the car
operates at a much higher utility than an

248
00:15:43.759 --> 00:15:48.519
amal car it operates so like typically
the use of a car is about ten

249
00:15:48.600 --> 00:15:50.600
to twelve hours a week, so
most people will drive one and a half

250
00:15:50.600 --> 00:15:54.480
to two hours a day, typically
ten to twelve hours a week of total

251
00:15:54.519 --> 00:15:58.240
driving. But if you have a
car liken operate autonomously, then most likely

252
00:15:58.279 --> 00:16:02.559
you could probably most likely you'd have
that car operate for a third of the

253
00:16:02.559 --> 00:16:04.200
week or longer. So they're a
hundred hours in a week, so probably

254
00:16:04.200 --> 00:16:07.759
you've got something on the order of
fifty five to sixty hours a week of

255
00:16:07.799 --> 00:16:12.200
operation, maybe a bit longer.
So the fundamental utility of a vehicle increases

256
00:16:12.240 --> 00:16:15.919
by a factor of five. So
you look at this from a macroeconomic standpoint

257
00:16:17.039 --> 00:16:19.080
and say, just if this was
like some if we were operating some big

258
00:16:19.120 --> 00:16:23.759
simulation, if you could upgrade your
simulation to increase the utility of cars by

259
00:16:23.759 --> 00:16:29.399
a factor of five, that would
be a massive increase in the economic efficiency

260
00:16:29.480 --> 00:16:33.000
of the simulation. Just gigantic,
we'll do Model three S three and excess

261
00:16:33.000 --> 00:16:36.440
taxis. But we made an important
change to our leases, so if you

262
00:16:36.559 --> 00:16:38.399
lease a Model three, you don't
have the option of buying it. At

263
00:16:38.440 --> 00:16:41.759
the end of the lease, we
want them back. If you buy the

264
00:16:41.799 --> 00:16:44.279
car, you can keep it,
but if you lease it, you have

265
00:16:44.320 --> 00:16:47.600
to give it back. And as
as I said, where in any locations

266
00:16:47.600 --> 00:16:51.320
where there's not enough supply for sharing, TESL will just make its own cars

267
00:16:51.360 --> 00:16:52.759
and add them to the network in
that place. So the current cost of

268
00:16:53.480 --> 00:16:56.679
a Model three Robotaxi is less than
thirty eight thousand dollars. We expect that

269
00:16:56.759 --> 00:17:02.480
number to improve over time resigning the
cars. The cars currently being built are

270
00:17:02.519 --> 00:17:06.200
all designed for a million miles of
operations. The drive units designed, find

271
00:17:06.240 --> 00:17:08.799
and test and validated for a million
million miles of operation. The current battery

272
00:17:08.839 --> 00:17:14.359
pack is about maybe three hundred to
five hundred thousand miles. The new battery

273
00:17:14.400 --> 00:17:18.759
pack that probably going into production next
year is designed explicitly for a million miles

274
00:17:18.759 --> 00:17:22.200
of operation. The entire vehicle well, it's designed to operate for a million

275
00:17:22.200 --> 00:17:26.400
miles with minimum maintenance. So we'll
actually be adjusting tire design and really optimizing

276
00:17:26.400 --> 00:17:30.759
the car for a hyper efficient robotaxi, and at some point you won't need

277
00:17:30.960 --> 00:17:34.480
steering wheels or pedals and we'll just
delete those. So as these things become

278
00:17:34.519 --> 00:17:37.279
less and less important, we'll just
delete parts just they won't be there.

279
00:17:37.319 --> 00:17:41.400
If you say, like probably two
years from now, we make a car

280
00:17:41.440 --> 00:17:44.480
that has no steering wheels of pedals, and if we need to accelerate that

281
00:17:44.480 --> 00:17:47.519
time, we can always just delete
parts easy. You know, probably say

282
00:17:47.599 --> 00:17:52.359
long term three years robotaxis with eliminated
parts, maybe it ends up being twenty

283
00:17:52.400 --> 00:17:55.559
five thousand dollars or less. And
you want a super efficient car. So

284
00:17:55.720 --> 00:17:59.440
the electricity consumption is very low.
So we're currently at four and a half

285
00:17:59.440 --> 00:18:02.359
miles per kill whatever, but we
will improve that to five and beyond.

286
00:18:02.480 --> 00:18:06.160
And there's just really no company that
has the full stack integration. We've got

287
00:18:06.359 --> 00:18:10.559
the vehicle design and manufacturing, we've
got the computer hardware in house, we've

288
00:18:10.559 --> 00:18:14.480
got the in house soft development and
AI, and we've got by far the

289
00:18:14.519 --> 00:18:18.680
biggest suite. It's extremely difficult,
not impossible paths, but extremely difficult to

290
00:18:18.680 --> 00:18:22.039
catch up when Tela has one hundred
times more miles per day than everyone else

291
00:18:22.039 --> 00:18:26.720
combined this is these todays. Is
the cost of running a gasoline car or

292
00:18:26.960 --> 00:18:30.279
the average cost of running a car
in the US. This is taken from

293
00:18:30.319 --> 00:18:33.880
Triple A. So it's currently about
sixty two cents a mile. Half thousand

294
00:18:33.880 --> 00:18:36.680
miles from fifty million vehicles adds up
to two trillion a year. These are

295
00:18:36.680 --> 00:18:40.200
literally just taken from the Triple A
website. Cost of ride sharing is the

296
00:18:40.319 --> 00:18:42.319
pointing to uber left is two to
three dollars a mile. The cost run

297
00:18:42.319 --> 00:18:48.279
a robotaxi, we think less than
eighteen cents a mile, and dropping like

298
00:18:48.359 --> 00:18:51.799
this is current, This would be
current, this current cost, future costs

299
00:18:51.839 --> 00:18:53.240
will be lower. You say,
what would be the probable gross profit from

300
00:18:53.240 --> 00:18:57.480
a single robotaxi. We think probably
something on the order of thirty thousand bells

301
00:18:57.519 --> 00:19:02.160
per year, and we expect that
we're literally designing. We're designing the cars

302
00:19:02.400 --> 00:19:06.680
the same way that commercial semi trailer
semi trucks are designed. Commercial semi trucks

303
00:19:06.680 --> 00:19:08.640
are all designed for a million mile
life, and we're designing the cars for

304
00:19:08.680 --> 00:19:12.079
a million my life as well.
So phenomenal dollars that would be a little

305
00:19:12.119 --> 00:19:15.519
over three hundred thousand dollars over the
course of eleven years might be higher.

306
00:19:15.640 --> 00:19:19.240
I think these consumptions are actually relatively
conservative, and this assumes that fifty percent

307
00:19:19.240 --> 00:19:22.319
of the miles driven art does nothing
are not useful, so this is only

308
00:19:22.559 --> 00:19:26.680
at fifty percent utility. By the
middle of next year, we will have

309
00:19:26.759 --> 00:19:30.759
over a million Tesla cars on the
road with full self driving hardware feature complete

310
00:19:30.839 --> 00:19:33.880
at a reliably level that we would
consider that no one needs to pay attention,

311
00:19:34.119 --> 00:19:37.240
meaning you could go to sleep and
from ashtandpoint if you fast forward a

312
00:19:37.319 --> 00:19:41.680
year, maybe a year and three
months at it. But next year,

313
00:19:41.799 --> 00:19:45.519
for sure we will have over a
million robo taxis on the road. The

314
00:19:45.559 --> 00:19:48.920
fleet wakes up with an over the
air update. That's all it takes.

315
00:19:49.000 --> 00:19:52.319
You say, what is in their
personal value of Roverotaxi. Probably on the

316
00:19:52.400 --> 00:19:56.119
order of a couple hundred thousand dollars, So buying a Model three is a

317
00:19:56.119 --> 00:20:00.440
good deal. Questions our production rates
Generally, if you look at a compound

318
00:20:00.440 --> 00:20:03.759
annual production rate since twenty twelve,
which is like the that's our first full

319
00:20:03.839 --> 00:20:08.720
year of model Model as production,
we went from twenty three thousand vehicles produced

320
00:20:08.720 --> 00:20:14.440
in twenty thirteen to around two hundred
and fifty thousand vehicles produced last year,

321
00:20:14.559 --> 00:20:18.559
So in the course of five years
we increased output by a factor of ten.

322
00:20:18.759 --> 00:20:22.519
I would expect that something similar occurs
over the next five or six years.

323
00:20:22.640 --> 00:20:25.799
As for sharing sharing versus, I
don't know. The nice thing is

324
00:20:25.799 --> 00:20:29.680
that essentially customers are fronting us the
money for the car. That's great.

325
00:20:29.799 --> 00:20:33.359
Sure, we expect the solving for
the snake charger. It's pretty straightforward.

326
00:20:33.359 --> 00:20:37.839
It's from a vision prop standpoint,
it's like a known situation. Any kind

327
00:20:37.839 --> 00:20:40.640
of known situation with vision is like
a charge port. It's a trivial.

328
00:20:41.119 --> 00:20:45.279
Yeah, the cars were just automatically
park and automatically plug in, there would

329
00:20:45.279 --> 00:20:49.279
be no humans of supervision required.
Yeah. During what was a pricing that

330
00:20:49.279 --> 00:20:52.559
we just threw some numbers on there, I think definitely plug in order pricing

331
00:20:52.599 --> 00:20:56.279
you think makes sense. We just
randomly said okay, maybe a dollar.

332
00:20:56.440 --> 00:21:00.240
And the thing is there's like on
the order of two billion cars and trucks

333
00:21:00.279 --> 00:21:03.559
in the world, So robotaxis will
be an extremely high demand for a very

334
00:21:03.599 --> 00:21:07.279
long time. And from my observation
thus far is that the auto industry is

335
00:21:07.359 --> 00:21:10.839
very slow to adapt. Like I
said, there's still not a car in

336
00:21:10.880 --> 00:21:12.720
the road that you can buy today
that is as good as the model S

337
00:21:12.960 --> 00:21:18.480
was in twenty twelve. So that
suggests a pretty slow rate of adaptation for

338
00:21:18.640 --> 00:21:22.240
the car industry, and so probably
a dollar is conservative for the next ten

339
00:21:22.319 --> 00:21:29.240
years because people to think there's like, actually not enough appreciation for the difficulty

340
00:21:29.240 --> 00:21:32.599
of manufacturing is insanely difficult. But
a lot of people I talked to think,

341
00:21:32.720 --> 00:21:36.039
if you just have the right design, you can instantly make as much

342
00:21:36.160 --> 00:21:40.240
of that thing as the world wants. This is not true. It's extremely

343
00:21:40.279 --> 00:21:45.240
hard to design a new manufacturing system
for new technology. Audi's having major problems

344
00:21:45.279 --> 00:21:48.480
manufacturing Eutron, and they are extremely
good at manufacturing. And if they're having

345
00:21:48.480 --> 00:21:52.599
problems, what about others. So
on the order of two billion Coston trucks

346
00:21:52.640 --> 00:21:56.519
in the world, on the order
of about one hundred million units per ure

347
00:21:56.599 --> 00:22:00.519
of production capacity of vehicles, but
only of the old design, it will

348
00:22:00.559 --> 00:22:04.119
take a very long time to convert
all of that to full self driving cars.

349
00:22:04.240 --> 00:22:07.200
And they really need to be electric
because the cost of operation of a

350
00:22:07.279 --> 00:22:11.880
gasling diesel car is much higher than
an electric car. It's any robotax that

351
00:22:11.960 --> 00:22:15.240
isn't electric will absolutely not be competitive. There's a close that we put into

352
00:22:15.240 --> 00:22:17.720
our cars. I think it was
about three or four years ago. They

353
00:22:17.720 --> 00:22:19.200
can only be using the Tesli the
network, yes, but it's like the

354
00:22:19.200 --> 00:22:22.880
app store that you can add,
only add or remove them through the tails.

355
00:22:22.920 --> 00:22:26.799
The network and then tells it gets
revenue share. I guess you could

356
00:22:26.839 --> 00:22:30.119
operate a rental car fleet, but
I think this is very unwieldy. I

357
00:22:30.160 --> 00:22:33.039
think there will be a phone home
thing where if the car gets stuck,

358
00:22:33.160 --> 00:22:36.799
it'll just phone home to Tesla and
ask for a solution. Things like being

359
00:22:36.799 --> 00:22:40.839
pulled over for by athlete softshare.
That's easy, fresh program and that's not

360
00:22:40.880 --> 00:22:44.359
a problem. It will be possible
for somebody to take over using the steering

361
00:22:44.359 --> 00:22:47.119
wheel, or at least for some
period of time, and then probably down

362
00:22:47.160 --> 00:22:51.319
the road we'll just cap the steering
wheel so there's no steering control. We'll

363
00:22:51.319 --> 00:22:53.039
just take steering wheel off, put
a cap on that were literally just unbolt

364
00:22:53.079 --> 00:22:56.799
the steering wheel and put a cap
on where the steering wheel handle currently is.

365
00:22:56.839 --> 00:23:00.799
I think there'll be a transition period
where people will be able to take

366
00:23:00.839 --> 00:23:03.960
over and should be able to take
over from the rover taxi, and then

367
00:23:03.240 --> 00:23:07.319
once regulators are comfortable with us not
having the steering wheel, we'll just delete

368
00:23:07.319 --> 00:23:11.440
that. And for cars that are
on the are in the fleets, obviously

369
00:23:11.480 --> 00:23:14.160
with the permission of the owner.
If it's owned by somebody else, we

370
00:23:14.160 --> 00:23:17.240
would just take the steering worl off
and put a cap where the steering world

371
00:23:17.279 --> 00:23:19.640
currently attaches. In future, will
the probabity of the steering world being taken

372
00:23:19.680 --> 00:23:22.279
away in the futures one hundred percent. Consumers will demand. It does not

373
00:23:22.359 --> 00:23:26.640
mean prescribing a point of view about
the world. This is me predicting what

374
00:23:26.720 --> 00:23:30.720
consumers will demand. Consumers will demand
in the future that people are not allowed

375
00:23:30.759 --> 00:23:33.279
to drive these two ton death machines. There were amphibians, but then pretty

376
00:23:33.319 --> 00:23:37.359
much that things just become like land
creatures. It'll be a little bit of

377
00:23:37.359 --> 00:23:41.279
an amphibian phase. I think are
a mistake. We actually had ht maps

378
00:23:41.279 --> 00:23:44.680
for a while. Actually can that
because you either need hd mapps, in

379
00:23:44.720 --> 00:23:48.079
which case, if anything changes about
the environment, the car will break down,

380
00:23:48.240 --> 00:23:49.519
or you don't need HTAM apps,
in which case why you're wasting your

381
00:23:49.519 --> 00:23:56.519
time during hd mapps. The hd
maps saying like the two main crutches that

382
00:23:56.119 --> 00:24:00.400
are that should not be used and
will be obviously fall and foolish are light

383
00:24:00.480 --> 00:24:03.000
aer and amaps. If you need
a geofest area, you don't have real

384
00:24:03.039 --> 00:24:06.880
self driving. I think we're actually
going to want to push our sort of

385
00:24:06.880 --> 00:24:11.839
standar range plus battery more than our
long range battery because the energy content in

386
00:24:11.880 --> 00:24:15.680
the long range pack is to send
higher kilo what hours. So essentially you

387
00:24:15.720 --> 00:24:19.400
can make third more cars if you
just if they all sort of standard range

388
00:24:19.400 --> 00:24:22.839
plus instead of the long range pack. It's ones like around fifty kilo what

389
00:24:22.880 --> 00:24:26.759
hours the other ones around seventy five
kilo what hours. So we're actually probably

390
00:24:26.759 --> 00:24:30.759
gonna binance our sales intentionally towards the
smaller battery in order to have a higher

391
00:24:30.839 --> 00:24:34.559
volume of what basically you want to
the obvious thing to maximize the number of

392
00:24:34.559 --> 00:24:40.000
autonomous units or the number of maximize
the output that will subsequently result in the

393
00:24:40.000 --> 00:24:42.720
biggest autonomous leap down the road.
So we're making doing a number of things

394
00:24:42.720 --> 00:24:45.759
in that regard, but it's just
on our for today is meeting the million

395
00:24:45.799 --> 00:24:52.079
mile of life is basically just about
getting the cycle life of the pack to

396
00:24:52.720 --> 00:24:55.759
you need. Basically on the order
let's say you've got a basic math,

397
00:24:55.799 --> 00:24:57.559
if you've got a two undred and
fifty mile range pack. You're gonna need

398
00:24:57.640 --> 00:25:03.039
four thousand cycles. Very achievable.
We already do that with our stationary storage

399
00:25:03.200 --> 00:25:07.880
some stay stationary storage solutions like Powerpack. We're ready to play power Pack with

400
00:25:07.920 --> 00:25:11.319
four thousand cycle life capability. Really
fundamental message that consumers should be taking today

401
00:25:11.519 --> 00:25:18.720
is that it's financially insane to buy
anything other than a Tesla. They were

402
00:25:18.880 --> 00:25:22.559
like owning a horse in three years
find if you don't own a horse,

403
00:25:22.599 --> 00:25:25.200
but you should go into it with
that expectation. If you buy a car

404
00:25:25.359 --> 00:25:27.640
that it does not have the hardware
necessary for full self driving, it is

405
00:25:27.680 --> 00:25:30.160
like buying a horse, and the
only car that has the hardware nessary for

406
00:25:30.400 --> 00:25:33.640
full self driving as a Tesla for
people should really think about their approaches any

407
00:25:33.680 --> 00:25:37.279
other vehicle. It's basically crazy to
buy any other car than a Tesla.

408
00:25:37.480 --> 00:25:41.240
We need to make that convey that
argument clearly, and we will have to

409
00:25:41.240 --> 00:25:45.000
today. Yeah, we talked to
regulators around the world all the time.

410
00:25:45.200 --> 00:25:49.839
As we introduce additional features like navigator
and autopilots, this requires like regulatory approval

411
00:25:49.839 --> 00:25:55.519
on a perju jurisdiction basis. I
think fundamentally regulators in my experience are convinced

412
00:25:55.559 --> 00:26:00.279
by data. So if you have
a massive amount of data that shows that'omy

413
00:26:00.319 --> 00:26:03.839
is safe, they listen to it. They they may take time to digest

414
00:26:03.839 --> 00:26:07.880
the information. Their process may take
a bit of time, but they have

415
00:26:07.960 --> 00:26:11.079
always come to the right conclusion on
what I've seen. I think like probably

416
00:26:11.200 --> 00:26:15.359
yeah, like Tesla owned robot taxis
would be in dense open areas along with

417
00:26:15.599 --> 00:26:19.799
customer vehicles, and then as you
get to medium and low density areas,

418
00:26:19.880 --> 00:26:25.599
it would chand to be more that
people own the car and occasionally lend it

419
00:26:25.640 --> 00:26:27.680
out. Yeah, there are a
lot of edgecases in Manhattan and say downtown

420
00:26:27.680 --> 00:26:32.519
San Francisco, but those are and
there are variousties around the world that have

421
00:26:32.680 --> 00:26:36.519
a challenging open environments. But we
do not expect this to be a significant

422
00:26:36.720 --> 00:26:41.240
issue when I say future complete IDOL
work in downtown San Francisco and downtown Manhattan

423
00:26:41.279 --> 00:26:45.359
this year. Right now, AI
neural nets are used really for object recognition,

424
00:26:45.640 --> 00:26:49.160
and we're still basically just using it
as still frames, so identifying objects

425
00:26:49.160 --> 00:26:55.079
and still frames and tying it together
in a perception path planning layer thereafter the

426
00:26:55.160 --> 00:26:59.640
But what's happening is steadily is that
the neural net is eating into the software

427
00:26:59.680 --> 00:27:03.759
base more and more, and so
over time or expect the neural net to

428
00:27:03.759 --> 00:27:07.079
do more and more. Now,
for a computational cost standpoint, there are

429
00:27:07.119 --> 00:27:11.880
some things that are very simple for
a HIC and very difficult for a neural

430
00:27:11.920 --> 00:27:15.680
net, and so it probably makes
sense to maintain some level of heuristics in

431
00:27:15.680 --> 00:27:21.160
the system because they're just computationally a
thousand times easier than a neural net.

432
00:27:21.279 --> 00:27:22.960
Neural net is like a cruise missile, and if you're trying to swat a

433
00:27:23.039 --> 00:27:26.519
fly, just use a fly swatter, not a cruise missile. So bit

434
00:27:26.599 --> 00:27:32.200
over time, I would expect that
it moves really to just training on against

435
00:27:32.279 --> 00:27:37.359
video and then video in car steering
and pedals out or basically video in that

436
00:27:37.640 --> 00:27:41.559
lateral longitudinal acceleration out almost entirely.
That's what we're going to use the Dojo

437
00:27:41.640 --> 00:27:44.759
system for. There's no system that
can currently do that. Essentially, the

438
00:27:44.759 --> 00:27:47.319
car is going to do what a
human would do. It can think with

439
00:27:47.440 --> 00:27:51.960
humans like basically a camera on a
slow gimble and it's a remarkable that people

440
00:27:51.960 --> 00:27:53.440
are able to drive the car in
the way that they are, because if

441
00:27:53.519 --> 00:27:56.640
you know, you can't look at
all directions at once. The car can

442
00:27:56.680 --> 00:28:00.160
literally look in all directions at once
with multiple cameras, so cumbs are able

443
00:28:00.200 --> 00:28:03.920
to drive just by looking this way, you're looking that way. They're actually

444
00:28:03.960 --> 00:28:06.640
stuck in their driver's seat. They
can't really get out of the driver's seat.

445
00:28:06.799 --> 00:28:08.160
So it's like kind of one camera
on a gamble and is able to

446
00:28:08.240 --> 00:28:11.839
drive. A conscientious driver can drive
with very high safety. The cameras in

447
00:28:11.880 --> 00:28:15.480
the cars have a better vantage point
than the person, so they're like up

448
00:28:15.519 --> 00:28:19.200
in the b pillar or at in
front of the review mirror. They've they've

449
00:28:19.240 --> 00:28:22.680
really got a great vantage point.
So if you're turning onto a road that's

450
00:28:22.720 --> 00:28:26.680
got a lot of high speed traffic, you can just do what person is

451
00:28:26.799 --> 00:28:29.519
just like gradual turn a little bit. They've got fully into the road that

452
00:28:29.640 --> 00:28:32.440
the camera see what's going on and
if things look good, and then the

453
00:28:32.599 --> 00:28:34.839
rear cameras don't show on any oncoming
traffic, or if you go and if

454
00:28:34.839 --> 00:28:37.720
it looks sketchy, you can just
pull back a little bit, just like

455
00:28:37.759 --> 00:28:42.079
a person. The behavior is like
remarkably, it starts to become remarkably lifelike.

456
00:28:42.319 --> 00:28:45.240
It's like quite eerie. Actually,
the car just starts behaving like a

457
00:28:45.279 --> 00:28:49.440
person. Well, we publish the
accidents per mile every quarter, and what

458
00:28:49.519 --> 00:28:52.839
we see right now is that water
pilot is about twice as safe as a

459
00:28:52.880 --> 00:28:56.519
normal normal driver on average, and
we expect that to increase quite a bit

460
00:28:56.559 --> 00:29:00.759
over time. Like I said,
in the future, it will be consumers

461
00:29:00.799 --> 00:29:03.359
will want to outlaw saying they will
succeed or am I saying I agree with

462
00:29:03.400 --> 00:29:07.000
this position, But in the future, consumers will want to outlaw people driving

463
00:29:07.000 --> 00:29:11.079
their own cars because they're unsafe.
And if you think of like elevators used

464
00:29:11.079 --> 00:29:14.160
to be operated on a big lever
go up and down the floor. There's

465
00:29:14.200 --> 00:29:18.640
like a big relay and yet elevator
operators, but then periodically they would get

466
00:29:18.680 --> 00:29:22.200
tired or drunk or something and then
turn the lever at the wrong time and

467
00:29:22.319 --> 00:29:26.519
sever somebody in a half. So
now you do not have to elevator operators.

468
00:29:26.599 --> 00:29:30.319
And we're quite alarming if you went
into an elevator that had a big

469
00:29:30.400 --> 00:29:33.680
lever that could just move between floors
arbitrarily. So there's just buttons. And

470
00:29:34.039 --> 00:29:37.880
in the long term, again not
a value judgment saying I want the world

471
00:29:37.880 --> 00:29:41.119
to be this way, I'm saying
consumers will most likely demand the people are

472
00:29:41.119 --> 00:29:45.440
not allowed drive Carson follow up,
Can you share with us how much Tesla's

473
00:29:45.440 --> 00:29:49.359
spending on auto pilot or autonomous technology
by order of magnitude on an annual basis?

474
00:29:49.400 --> 00:29:52.599
Thank you. It's basically our entire
expense structure. Between now and when

475
00:29:52.599 --> 00:29:57.000
the robot taxis are fully deployed throughout
the world, the sensible thing for us

476
00:29:57.039 --> 00:30:00.799
is to maximize rate and drive the
company cashlow neutral. Once the rover taxi

477
00:30:00.839 --> 00:30:04.200
fleet is active, I would expect
to be extremely casually positive. Who is

478
00:30:04.279 --> 00:30:07.519
liable for an accident? Is it
Tesla? Or does it mean if the

479
00:30:08.000 --> 00:30:11.920
vehicle has an accident and harms,
probably Tesla is probably Tesla? Yeah,

480
00:30:12.000 --> 00:30:15.640
very good, right, thing too
is just make sure there are very few accidents