Latest / Coopers' Code / Tesla on Trial: $329,000,000 Autopilot Judgment with Brett Schreiber, Part 1

Transcript

Tesla on Trial: $329,000,000 Autopilot Judgment with Brett Schreiber, Part 1 · Coopers' Code

0:00
Howdy and welcome to the show. Cooper's code focuses on legal
0:03
issues and notable practitioners, distilling wisdom
0:06
so we all achieve the best results for our clients.
0:09
I'm Miles Cooper and I'm very excited for this after action
0:11
review where we sit down with Brett Schreiber from Singleton
0:14
Schreiber. We discussed what it took to
0:17
achieve the first jury verdict against Tesla and what a verdict
0:20
it was if Florida federal jury hit Tesla with a $243 million
0:25
judgement after learning about the harm caused.
0:27
Deny Bell Benavides and Dylan Angulo and Tesla's actions that
0:31
led to the incident. BRO will help us understand the
0:34
groundwork and courtroom decisions that helped achieve
0:36
with this record verdict. Welcome to the show, Brad.
0:39
Thanks for having me. I'm really excited to be able to
0:43
sit down and talk with you about your verdict and that I think
0:48
the best way to kind of launch into things, if you don't mind,
0:50
is can you just give us a. A paragraph or two snapshot of
0:56
the case and then we'll start digging into how you built the
0:59
case. Trial moments, other pieces that
1:02
work. Absolutely.
1:03
OK, happy to. So what this case was was the
1:06
first third party autopilot defect case to go to trial in
1:11
America. And what that means is this was
1:13
not something where the plaintiff was the driver of the
1:18
Tesla on autopilot. This was a young couple, 22
1:22
years old and 26, who were minding their business at the
1:28
end of AT intersection in Key Largo, FL, having lawfully
1:32
parked beyond the edge of the road behind some signs and had
1:36
done nothing other than go fishing that evening.
1:39
And we're packing up, looking at the stars and talking when a
1:42
Tesla on autopilot drove through that intersection, through that
1:47
stop sign, through the flashing lights and the signs at over 70
1:50
miles an hour colliding with them, causing the 22 year old
1:54
young woman to lose her life and her 26 year old boyfriend to be
1:59
forever injured. So that in a nutshell, is is
2:03
what the case was about. So when you describe it that
2:06
way, it sounds very straightforward and a case where
2:09
Tesla would say to itself, gosh, it was on autopilot, that's on
2:16
us. Tell us how to compensate the
2:19
family. It sounds like they didn't do
2:21
that. Indeed they did not.
2:23
There's a lot of wrinkles, as I'm sure we'll probably unpack.
2:26
But I think one of the things that is is critically important
2:29
in this, and really one of Tesla's dirty little secrets, of
2:34
which there are many, is that they took the position in
2:38
hearings long before I was asked to get involved in the case, a
2:42
little less than about a year before trial, that this was not
2:45
an Autopilot case because Autopilot was off at the time of
2:48
impact. And having cases like this
2:51
across the country and having been working on them for the
2:54
better part of several years, that is a constant refrain.
2:58
Because the autopilot system is designed that when it recognizes
3:03
a crash is imminent, it's beyond the point of no return.
3:07
This is usually about 1.3 seconds or so before impact.
3:12
Autopilot shuts off. So technically, when vehicles
3:17
are smashing into each other, when metal is crunching into
3:20
itself, when bodies are being foiled and vaulted about,
3:25
autopilot is off. And so Tesla routinely, when
3:31
inquiries are come from law enforcement after a collision,
3:34
they go autopilot. There's nothing to see here.
3:38
Autopilot was off and so that was one of the claims that were
3:43
made. And the other major issue for
3:45
them was they took the position it was never designed to stop in
3:49
this type of environment. This was, you know, a
3:52
perpendicular vehicle off the road.
3:56
This is not something that Autopilot of this era and really
3:59
even Autopilot today was designed to to respond to.
4:05
And so they very much took the position that, you know,
4:09
Autopilot had nothing to do with this crash for them.
4:12
This case was all about a bad driver.
4:15
And in nine out of 10 cases, it's always about the driver.
4:20
But as I know, we'll talk about more, right?
4:23
There's no doubt that this driver was a bad actor, but
4:26
every actor needs a stage. And Tesla has routinely set the
4:29
stage for these preventable collisions to continue to occur.
4:34
Does Tesla make people aware of this cut out of the self driving
4:41
feature? Absolutely not.
4:42
And and effect and think I think it's one of the reasons why
4:47
autopilot crashes and autopilot failures in this country, I'm
4:51
going to estimate are under reported probably by an order of
4:53
magnitude of 10X. It's almost like public health.
4:57
Right routinely if there's an outbreak of some disease, the
5:01
reporting to public health authorities is 10% of the actual
5:04
population of people. The same is true for autopilot
5:07
because so often times and I see this in other cases that I have
5:12
and this was not unique to the to the Miami trial law
5:16
enforcement, you know, bless their hearts are are not trained
5:20
in this. You got a lot of these old kind
5:22
of Smokey the bear looking guys, right?
5:25
The old highway patrolman where the Smokey the Bear hats and
5:28
they've been reconstructing accidents for 40 years, back
5:31
when they used to do it on, you know, little graph.
5:33
Paper and drawn little square cars you know on the on the
5:37
diagram and so when they get involved in a case that
5:42
ultimately is looking at really what has become a computer that
5:46
just happens to be on wheels. They're out of their element.
5:50
And so when Tesla, the world's largest corporation, comes
5:54
forward and says, oh, Autopilot wasn't on, many of these
5:59
individuals go, well, I guess they must be.
6:01
Right. Nothing to see here.
6:03
So I think there are probably, like I said, well, we know
6:08
about, you know. 1000 + 2000 plus crashes based upon NITSA
6:12
investigations where autopilot was involved.
6:15
We know the real numbers are, are orders of magnitude higher
6:18
and, and, and to put a finer point on that, the government
6:21
realized this. And so about five years ago,
6:24
NITSA put in place what's called a Standing General Order 95.
6:28
And that requires autonomous vehicle manufacturers to report
6:32
a crash if the autonomous features were engaged 30 seconds
6:37
prior to impact. But it's an honor system, OK?
6:41
And they have to report it. And not surprisingly, in a post
6:45
Doge world that we lived in, all of the people who worked within
6:49
NITSA who were investigating this, there were eleven of them
6:53
that were dedicated to investigating autonomous
6:56
vehicles and and primarily. Tesla, because they have a
6:59
disproportionate number of these types of crashes, about 80% of
7:03
them, I think 8, seven or eight of them have been fired.
7:06
So rest well at night knowing that the entire autonomous
7:10
vehicle industry in America is currently being policed by three
7:14
or four people. Given that some people may be
7:17
visiting this down the road, that DOGE is the Department of
7:20
Governmental Efficiency where Elon Musk, who runs Tesla, was
7:27
brought in to in essence be the fox guarding the hen house.
7:30
And he went through and he gutted all the people who were
7:33
watching his companies. Indeed.
7:35
OK, good summary. Is there any position that Tesla
7:41
took with you or with the jury or with others about this cut
7:45
out feature that you know, it's not it's not a bug, it's a
7:47
feature. It's it's for safety.
7:49
Oh yeah, no, they say. Look, once the vehicle knows
7:52
it's going to crash, we have to hand the control back to the
7:55
human. But the problem is they hand the
7:58
control back to the human at a point where the human is no
8:02
longer capable of doing anything to prevent or mitigate the
8:06
crash. And, and that was, you know,
8:09
part of our beef. Not only did this vehicle
8:12
perceive and, and predict that it was going off the road, that
8:17
there was a vehicle there, that there was a pedestrian there,
8:20
that it was beyond what's called the end of drivable space, which
8:23
is the way the computer vision system understands what is
8:26
allowed to drive on and what it's not.
8:28
It knew all of those things and it ultimately did nothing.
8:34
And what we now know from years of research of this.
8:38
Humans are terrible at monitoring autonomous features,
8:43
right? I mean, this goes back to, you
8:45
know, the old I Love Lucy episode, you know, with the
8:48
chocolates coming down the conveyor belt, right?
8:50
We're just, we're not good at it.
8:52
We weren't good at it in, you know, the era of Lucille Ball.
8:55
And we haven't gotten much better in the last 80 years.
8:58
And so they know that the. Time for a human to take over is
9:04
is multiples more than if you were actively driving.
9:08
When you're on autopilot, you essentially become a passenger
9:11
in your own vehicle, right? And your passenger in the car
9:14
and the driver's doing something.
9:15
When was the last time you reached over and grabbed the
9:17
steering wheel or stuck your leg across and jumped on the brakes?
9:21
We're just not wired that way. We don't do that.
9:23
And Autopilot creates that same sense of, of overtrust.
9:28
And you know, people rely on it, sadly, to their detriment.
9:32
So it sounds like we're talking a bit about, as you described
9:36
the the driver was a bad actor, but every actor needs a stage.
9:40
The stage being not just the cut out, but the way that Tesla
9:44
markets its autonomous features period.
9:49
Are are we at a point where we should talk about that a little
9:51
bit? No, for sure.
9:52
And, and, and that's very much a part of it because you know, one
9:55
thing I've been really crystal clear about and talking about
9:58
this verdict in the same way when I talked to to the jury
10:00
about it and when I've talked to the media about it, right?
10:03
Like this was not an indictment of autonomous vehicle
10:06
technology. To the contrary, this was an
10:09
indictment of Tesla. Tesla is an industry outlier
10:15
both with respect to how they've designed their system and how
10:19
they market their system, right. And so, you know, there has not
10:23
been ACEO of an automotive manufacturer for, you know,
10:28
since the Model T rolled off the line who is like perpetually
10:32
live tweeting over hyping, overstating and misrepresenting
10:37
the capabilities of its vehicles.
10:40
And, you know, say what you will about Elon today, as we're
10:43
talking in November of 2025, but going back for many years, he
10:48
was and in some circles still is considered a credible source,
10:53
right? I got these cases all over the
10:54
country. I focus group him.
10:56
He plays differently in different environments.
10:59
And I'll tell you, in Miami, you know, there were a lot of folks
11:02
in focus groups. Who were like.
11:03
He's a genius, he's a visionary, he's the world's richest man and
11:08
that. Carries with it credibility and
11:11
so people listen to. Him, for better or for worse,
11:16
right? And so when he's telling them,
11:18
as he had been for a decade leading up to this crash,
11:23
autonomous driving is a solved problem.
11:26
The car will stop for anything, including an alien spaceship in
11:30
the road. You're going to be able to drive
11:32
from New York to LA without a touch, right?
11:35
Those statements were as untrue the day he said them as they
11:39
remain untrue today. And he continues to double down
11:44
and continues to misrepresent and mistake and overhype.
11:48
And it's it's as simple as the the name of the thing, right?
11:51
There is nobody else in the autonomous vehicle space, any
11:55
other manufacturer in the world that calls there what they
11:58
consider Level 2 autonomy, partial autonomy, where the
12:01
human has to be monitoring. No one calls it auto anything
12:06
because it's not, you know, they use terms like copilot, driver
12:10
assist, supercruise. Right.
12:12
Words matter. And, and that was a big issue,
12:16
right? That That that I think is, is so
12:18
much about what this case and ultimately what this verdict
12:21
ended up representing was going back to that theme that that
12:25
words matter. And I'm not a Tesla driver.
12:31
Peripherally pay attention to some of the terms that are used.
12:34
Is full self driving mode one of the terms that is used to
12:37
describe their semi autonomous mode?
12:40
Yes. And and that was one of the
12:42
things they love to come back in court and call this essentially
12:48
an, a driver assistance system, right?
12:51
And you know, they're they're in there talking about how this is
12:54
basically like, you know, it's just glorified cruise control.
12:58
This is like what you had on your grandmother's Oldsmobile,
13:01
right? Which then you juxtapose with
13:04
their own statements, their own marketing, Musk's, you know,
13:07
press releases and press conferences where they're
13:10
calling it enhanced autopilot or full self driving, right?
13:14
How do you call something that by definition, you know, in the
13:19
engineering realm is a partial something that you then marketed
13:23
as full self driving, right? And, and so again, thank you
13:29
Tesla. They opened a perfect, very
13:33
understandable and digestible metaphor to talk about Tesla in
13:37
the showroom versus Tesla in the courtroom, right?
13:41
Because Tesla in the showroom will tell you that this is the
13:44
greatest enhanced full self driving vehicle the world has
13:48
ever seen. And Tesla in the courtroom makes
13:51
it sound like it's a jalopy, right?
13:54
And So what I was able to tell that jury.
13:57
Is you cannot show up in court and raise as a defense the fact
14:04
that the full self driving enhanced autopilot, greatest
14:08
autonomous vehicle you claim to have made never actually
14:13
existed. Words matter.
14:17
We're, I think jumping ahead of part partly because I'm, I'm
14:21
curious about how they describe this and, and what tools you
14:24
used in the courtroom. Can I'm going to take us back a
14:28
little bit chronologically. Can you tell us how you got
14:31
involved in this case and what steps you were involved in, in
14:34
terms of building it for the jury trial itself?
14:37
Sure. It really started.
14:41
With a mojito in Old San Juan, PR.
14:45
OK, I was in Puerto Rico, in this bar in Old San Juan.
14:51
Looked like Ernest Hemingway. Couldn't, you know, come walking
14:54
out of the the backroom at any moment.
14:57
And I'm there with my wife. We're there for a legal
14:59
conference. And I had the Zoom call on my
15:02
calendar at, like, 2 in the afternoon.
15:04
We're at lunch. And she goes, I'm going to have
15:06
a mojito. She's going back to, you know,
15:08
the pool to, you know, hang out, read a book and take a nap.
15:10
I was like, you know what? I'm going to join you.
15:13
So I have a mojito. And I learned this place, which,
15:16
you know, invented the mojito. Every bar in Old San Juan claims
15:19
that they invented the mojito, but they pour them like
15:21
gasoline. And I had one.
15:24
And then she's like, I'm going to have another.
15:25
Yeah, me too. And so then I get on the Zoom
15:27
call with these guys who I knew had a Tesla case and I knew it
15:30
was in Miami and that's all I knew.
15:32
Well, by the time I got on the Zoom, I was, I was a little
15:36
motivated and, and one might say a bit lubricated.
15:41
And so when they tell me the story of, of what had happened,
15:45
I was very much very quick to be like, I'm in, let's go, let's
15:51
kill these guys. I know this story as I had a few
15:54
of these cases going at the time.
15:55
I said, send me the Dropbox. I'll take a look at the file
16:00
next day, send me the Dropbox. I'm on my flight home through
16:03
Miami. And I open it up.
16:05
And I realized that there's been one liability deposition taken
16:09
in the last 2 1/2 years. And we're in federal court and
16:13
expert reports are due in six weeks.
16:17
And I suddenly started to question my decision making,
16:21
which was partly informed by my mojitos.
16:24
And I looked at it and go, oh, oh wow.
16:28
That's that's a lot of work. That needs to be done in a
16:30
really short amount of time. Thankfully, Weisel try lawyer
16:34
who I came up with one of my, you know, through my first
16:38
really adult job taught me that, you know, defense lawyer is your
16:41
adversary, not your enemy. And I knew these guys.
16:43
I knew the the defense firm that does a lot of Tesla's work
16:46
nationally and I knew they needed to do some work too.
16:49
So I was able to associate into the case and get an agreement to
16:54
push things out by 6 weeks. So now I had the opportunity to
16:59
do about three years of work in three months.
17:01
And that's what we started doing.
17:04
And it was just, it was just a, a full on full court press.
17:09
So the first thing I did was, of course, it's an auto product
17:12
case. Where's the car, right.
17:14
So we get out there, we do an inspection of the vehicle.
17:17
There was something missing the autopilot computer.
17:21
And I'm going to interject a question here.
17:23
At this point, when you're coming in, with it being shortly
17:27
before expert reports are due, are you relying on the team, the
17:32
existing team, or do you have your own team of people who you
17:34
bring in at that moment? It was a combination, the team.
17:40
The team had been loosely put together, OK.
17:43
But they didn't have documents. They hadn't reviewed deposition.
17:47
Yeah. They hadn't done anything in a
17:50
meaningful way. And they hadn't been.
17:53
There wasn't sufficient ammunition for them to do what
17:57
they needed to do. And so that's where I came in.
18:00
And so I knew enough to know that these cases, I said before
18:04
are computers on wheels and it's all about data.
18:07
So we needed data to understand what happened in the crash.
18:11
And that's where we started. So you got to the inspection and
18:15
sounds like the computer that the important part was gone.
18:18
It was gone. And I, I've had looked at the
18:21
file and I had seen that there had been a subpoena that had
18:24
been sent to the Florida Highway Patrol, who had initially taken
18:27
the vehicle because they got involved and, you know, they
18:29
looked at this initially. As a homicide, right?
18:31
There was a homicide investigator, they were looking
18:33
at whether it's vehicular manslaughter or some sort of,
18:36
you know, heightened, you know, felony or something like that
18:39
for, for the driver's malfeasance.
18:41
And so they had the cops had had taken control of the vehicle
18:47
from the jump and then it was eventually released.
18:51
Tesla ended up buying it long before I showed up.
18:55
So it was preserved, but the computer was gone.
18:59
And I went to my local council and I said, guys, where's,
19:02
where's the computer? Right?
19:03
We, we served the subpoena. And you know, they said they
19:06
didn't have it. I reached out to Tesla.
19:09
You all know where the autopilot computer is and, and what's it
19:12
called? The MCU?
19:13
There's the media control units, the big screen, and then the,
19:16
the, the ECU as it's called, the electronic control unit, which
19:19
is the computer that lives kind of inside the dash behind it.
19:22
And they're both Mia, you know. Hey, Tesla, any you know where
19:25
any of this stuff is? Nope.
19:27
Don't know what to tell you. I said, all right, well, that's
19:30
a problem. So there was an evidence storage
19:35
log and it was signed by some guy and I was in the midst of,
19:40
of discovering some other case at the time.
19:42
So I couldn't do it. But I contacted my, my lawyers
19:45
who brought me in. I said, guys, we need to depose
19:47
this person. And they're like, well, why?
19:51
You know, I said, you know, they've already served the
19:53
subpoena. You know, they they've said they
19:55
don't have it. I go just just take the
19:57
deposition. You never know.
19:59
So they do. Officer, whoever it was shows up
20:03
at his deposition say, hey, officer, we slide the oven and
20:06
storage log over and it has an address.
20:08
And Hialeah said, hey, we we served a subpoena.
20:12
Didn't find it. There he goes.
20:14
Yeah, but did you check the other facility?
20:16
And Tamiami, I'm like, no, you know, yeah, not everybody knows
20:21
this, but we store a lot of evidence there.
20:25
OK? So the next day, we go down,
20:28
knock on the door, evidence storage, you know, clerk comes
20:33
up. It's like, basically like
20:34
sweeping up in the back, you know, And we handle this thing.
20:37
She goes into the back and she comes out.
20:39
She goes, oh, you mean this whole thing?
20:41
It's been sitting on a shelf for 4 1/2 years.
20:46
We're like Eureka, you know? We found it.
20:51
Now this had been an ongoing issue for about a year and a
20:54
half before when they had done had discovery motions on this.
20:58
So I had read from hearing testimony that that Tesla's
21:01
lawyers had represented if we ever found the Autopilot
21:04
computer, they would help us to to essentially pull the data and
21:08
decode it, right? Because that's.
21:10
Part of the problem, right? It all goes into Tesla's magic
21:13
box, or so they make you believe.
21:15
And so we think to ourselves, Oh well, I hate Tesla.
21:19
Remember when you guys said six months ago that if we ever found
21:22
the computer, you'd help us get the information off the
21:24
computer? He won't believe it.
21:27
But we found him. I'm sure they honoured that and
21:30
we're pleased to help set that up for you.
21:32
A. 100% no. And, and to which their response
21:36
was, well, that's great, but discovery's closed and there's
21:42
nothing we can do here. So, you know, you learn anything
21:47
in this industry. You know, sometimes you got to
21:50
go over, around above, you know, or through the obstacle and, and
21:57
that's what we did. When you say you went through
22:00
it, there are a variety of ways that one can do that.
22:04
The most logical to me, it sounds like we'd be going to the
22:08
judge and saying, hey judge, you know, fair is fair.
22:13
We got this thing. It's time for us to take a look.
22:16
Which which is one of the things that we did.
22:19
OK. So we brought a motion to compel
22:21
that Tesla had to help to pull the data off and and ultimately
22:26
because we did have some data, OK, they then that's part of it,
22:30
right? Tesla will produce in cases,
22:34
begrudgingly. But they often.
22:36
Times will produce some of the onboard videos of, of the of the
22:41
vehicle and they'll produce what's called log data.
22:44
And it's basically a giant Excel spreadsheet.
22:46
It's got 200 columns, it's got 10,000 rows.
22:49
It's just a bunch of data because again, remember, Tesla
22:53
is a tech. Company that got into cars,
22:55
right, they are not a car company that got into tech,
22:58
right and so. You know, the blessing and the
23:00
curse for them is they got data on everything but the stuff.
23:03
That they give you only tells you time.
23:06
This spreadsheet will tell you all these different things in
23:10
time and and really the I think the best way to describe it is
23:13
what log data is is it's kind of like the table of contents.
23:17
Of a book, and it's not even a complete table of contents, it's
23:20
a, but you know 80% of the table of contents and you can open up
23:24
a book and you can thumb through the table of contents and you
23:28
have a general idea about the content of the story that
23:32
follows. But you don't know the
23:34
characters, you don't know the plot twist, you don't know, you
23:37
know the the tension that that built up or the resolution.
23:41
And that's the stuff that you have to get from the autopilot
23:45
computer. That's the stuff they don't want
23:47
you to have. And the way my experts described
23:50
it is when you when you get that data from the autopilot
23:53
computer, you can superimpose it onto the videos.
23:57
And then that gives you what they call ground truth.
24:00
That is space and time OK. And you have to understand space
24:05
and time to understand what the autopilot system perceived,
24:10
predicted and reacted to or failed to perceive, failed to
24:14
predict and failed to react to. And that's where the fight was
24:17
led. So we said, Judge, you got to
24:20
help us hold Tesla to what they said they would do, give us the
24:25
data. And then we have to have them
24:28
produce this called augmented vision, augmented vision or
24:32
augmented video where they superimpose the data onto the
24:36
video so we understand what the system, what the computer was
24:40
really thinking and doing in the moments.
24:42
And so that's process started. But like anything, even in
24:48
federal court in the Southern District of Miami to get a
24:50
hearing and to brief it and all of that takes months.
24:53
And we were bumping up on a trial day.
24:56
And So what I've also learned doing cases, especially cases
25:01
like this, you know, there it was been a fair amount of media
25:06
attention, right? There are a lot of intrepid
25:08
reporters who have been spending a lot of time digging into Tesla
25:12
and Elon and how they manage data, how they handle things,
25:15
you know, just a whole host of things.
25:16
And I found it's very helpful to have good relationships with our
25:21
friends in the media and, you know, to share what you know.
25:24
And they can often timeshare what they know.
25:26
And there was one such reporter from The Wall Street Journal who
25:30
had just published a piece about Tesla.
25:33
And it was involving their data and what they did with it.
25:37
And she had located a Russian hacker who was capable of
25:42
pulling data from Autopilot computers.
25:47
And so we reached out. I talked to the reporter before
25:51
and. All I knew was his Twitter
25:54
handle. Green the only.
25:56
Green the only. Green the only.
25:59
OK. And so I reach out to Green
26:03
through Twitter DMS. No response.
26:07
Days go by. Bumping up on this motion.
26:10
We're bumping up on trial. I've got expert reports in, but
26:13
they have to supplement. I got a lot going on.
26:15
I finally reached back out to the Wall Street Journal.
26:17
Like, listen, can you let this guy know like, like, we're the
26:20
good guys here until I reach out and she does.
26:24
And she knew that because we had shared information and we had
26:26
talked and, you know, she understood what we were doing
26:29
and, and how, like I said, we all had kind of similar goals
26:32
to, to shine a light on, on a lot of Tesla's misdeeds.
26:36
And eventually I get an e-mail back from Greenlee only, and he
26:43
agrees to set up a Zoom call. And I get on Zoom, and this guy
26:48
is everything you would want in your Russian hacker.
26:52
He has a ponytail. He has a lazy eye.
26:54
He has a thick Russian accent that would make a Bond villain
26:58
jealous. And I get off the call with him
27:01
and we explain our circumstance and he looks at us and he's
27:05
like, of course, we can do this is no problem.
27:10
I was like, really? I mean, Tesla's made this sound
27:13
like this is impossible. No one other than them can do
27:15
it. He goes, no, no, no, I, I take
27:17
that off autopilot computer. I superimpose on video, I post
27:21
on Twitter. This is what they do.
27:23
It's fun. I was like amazing, right?
27:27
Like my jaw is on the floor, right?
27:29
Because I've been told for years, in other cases as well as
27:32
this one, Tesla's the only entity in the universe that is
27:35
smart enough to do this. And I got some dude.
27:39
I don't know where he is in the world, but he has figured this
27:41
thing out just as a side hustle. But does he get past Deber
27:47
eventually? Yes.
27:48
OK, sorry, go on. And so I'm on the zoom.
27:52
He's like, I show you there's no problem.
27:55
He goes, do you want to go to my basement?
27:56
I'm thinking to myself, not really, but but but OK, you
28:00
know, and he's on he's on zoom, right?
28:02
And he's on his iPad and he's walking down the stairs, right?
28:05
And we're, we're going down the stairs together and, and there's
28:08
like wood panelling and has a certain kind of silence of the
28:11
Lambs vibe. Like I'm not sure what I'm going
28:12
to find when I get down to the basement.
28:15
But we get down there and there's computers and wires and
28:17
shit all over the place. Like it's like, I guess what you
28:21
would expect again, I didn't even know what to expect.
28:23
So I can't say I had an expectation.
28:25
But we get down there and he goes, look, I, I take Dada and,
28:29
you know, I then I put on this machine and I, I take the chip
28:32
off and then I take here and I put in my, my computer and I
28:35
extract and, and he's just walking us through like it's
28:38
just, you know, it's just, that's something he does in an
28:40
afternoon. He goes, let me show you this
28:42
easy. He picks up a little plastic
28:45
container. It's probably, you know, 8
28:47
inches by 8 inches kind of looks like the thing you would put.
28:51
It's from like a tackle box or like a sewing kit.
28:53
It's got little partitions in it.
28:55
And he, he holds it up to the camera and he pours the corner
28:58
out and 203 hundred of these little bits fall into his end.
29:03
He goes, you see, these are autopilot chips.
29:05
He goes. People send me, I take that up,
29:08
I pull it off. I post on Twitter.
29:10
No problem. Like you've done this hundreds
29:14
of times. Oh yeah.
29:16
I go, he goes. I'm like, I can't believe it.
29:19
I go, all right, well, can I hire you to to help us with
29:23
this? He goes, sure.
29:25
I go. What's your?
29:26
Hourly rate, he goes, I don't know, 50.
29:29
I go US like $50.00. I said, bro, I go, I'm going to
29:34
have to possibly have you swear out a declaration in federal
29:38
court and say that you were like a forensic data extraction
29:41
specialist. I can't have you charging me
29:44
$50.00 an hour. I'm like.
29:46
How about 500? He goes 500, I go, yeah, $500
29:49
and I go, OK, we could. So next thing you know, 3 weeks
29:54
later, Oleg, as it were, who prefers to remain anonymous so
29:58
I'll just use his first name is on a plane to Miami and he
30:04
brings his machinery. His computer had to bring
30:08
another expert that I had found out of Florida who does
30:12
essentially he's a cloning expert.
30:14
So once we. Took the data, we had to be able
30:16
to create identical clones of it, which includes A
30:19
cryptographic signature, which is just fancy digital speak for
30:23
making sure they are one of one or two of two.
30:26
And so we did that. And then we head over to a
30:32
Starbucks in Coral Gables and Oleg orders a Venti hot
30:38
chocolate because he's a manchild who drinks Venti hot
30:43
chocolates in the middle of the day in in October in Miami.
30:46
And within 15 minutes using Starbucks Wi-Fi, Oleg has pulled
30:53
the data, has pulled it up on his computer, has taken the
30:57
video and starts to superimpose what happened in the crash.
31:04
And not only that, but he also tells us for the first time that
31:10
this snapshot of data, as it's called, it's called a collision
31:14
snapshot. It's basically a zip file,
31:17
right? When, when Teslas crash, as I
31:20
said before, they know in the moment, like we talked about,
31:23
that they're going to crash. And so autopilot shuts off.
31:27
But what it first does as a part of that process is it takes all
31:31
of this information. It takes the cameras.
31:33
Takes what's called. Canned data, which is basically
31:35
all the computers talking to each other.
31:37
It takes that log data that I talked about and a few other
31:39
things, little packets of information or buckets of
31:41
information, and it downloads them.
31:44
Now Tesla, for years, their 30B6, their corporate reps had
31:47
told me for years. It's called an urgent upload.
31:50
And they said, look, you know, sometimes we get it because this
31:54
was the position they had taken in this case.
31:56
We didn't get all of this data. Yeah, plaintiff wants us to do
31:59
this, but there's no way we we didn't get it.
32:01
You know, nothing to see here. That was the position they took
32:06
with law enforcement. That was the position they took
32:08
with federal investigators. That was the position they took
32:11
in court for 3 1/2 years. And in this moment, we finally
32:16
were able to show that they lied.
32:20
My gut had told me all along that they were lying, that they
32:23
knew more than they were letting on.
32:26
I know, shocker, right? Giant trillion dollar
32:28
corporation is not being forthright and and transparent
32:31
and information that's potentially damning to it.
32:33
I know that comes as a big surprise to everybody, but
32:37
fundamentally it was it was that moment where we were able to
32:40
show again, tech company got into cars.
32:43
What happens is the data is downloaded to the autopilot
32:47
computer because the vehicle doesn't know in that instant
32:51
before the impact if it's going to have cell coverage.
32:53
It doesn't know if it's going to end up on fire or it doesn't
32:55
know if it's going to end up in a lake.
32:57
So it downloads it first. Then after the crash, it checks
33:01
if it has cell coverage, almost like your e-mail refreshing.
33:04
It's like, wait, do I have cell coverage?
33:05
Do I have cell coverage? Oh, I do.
33:07
Boom. And it uploads it to the
33:11
mothership. And it includes a line of data
33:14
almost like your fax machine from the 1990s that says
33:17
transmission successful. And in that Starbucks, 15
33:22
minutes after Oleg points to the line of code that says they had
33:28
this 4 1/2 years ago, well before the 1st ambulance
33:34
responded to the scene, before the dust from the crash had
33:39
fully settled, Tesla had all of the information about how and
33:45
why this crash occurred. And for 4 1/2 years, they lied.
33:51
They lied to the court, they lied to law enforcement, they
33:54
lied to the victims. And now we could prove it.
33:58
And so at that moment, I knew we had them.
34:02
So again, when we first started talking about the case car goes
34:10
out of control during full self self driving, seems like at that
34:13
point Tesla says, gosh on us, tell us how to properly
34:17
compensate when you now have this piece.
34:20
I presume through the taking of expert depositions etcetera that
34:25
they learn that you have this piece at some point before you
34:28
go to trial and they say to themselves, this is probably not
34:32
good for us. Let's talk about resolution,
34:34
Brett. You'd think.
34:37
You would think. They didn't.
34:40
So for the next 10 days they were under an order.
34:44
This was Halloween, so it's kind of the day before Halloween.
34:48
It's kind of spooky that we figured it out that day.
34:52
They were under obligation about 10 days later to produce this
34:56
augmented video. They now have the data.
34:57
They had it for a couple of weeks.
34:58
They said that's how long they need it.
35:01
And we knew we had it. We were already, we had built
35:04
out the video and all the things.
35:06
And so we waited, we waited till the day that they were supposed
35:09
to produce to see what they would come up with.
35:12
And instead they responded on that day, sorry guys, there's
35:18
nothing to see here. These are placeholders.
35:21
There's, there's not even data here.
35:23
We don't know what you're talking about.
35:26
We said, well that's interesting because look what we have from
35:30
that same exact cloned version of data.
35:34
Look what we figured out. So we go to the court on that
35:37
day, have the hearing where Tesla goes, we don't have
35:41
anything. And we go, yeah, you do, It's
35:45
right here. We built.
35:46
This and it was at that point where the magistrate said on the
35:52
record, So you're telling me the plaintiffs have out Teslaed
35:57
Tesla? So it looks that way, Your
36:00
Honor, It looks that way. That was also about the time
36:04
where a nearly 7 figure sanction order was entered for the fact
36:12
that we had to go through all of these efforts to to do this,
36:17
retain all of these folks, spend all of this money, attorney
36:20
time, expert time, etcetera, resulting in one of probably the
36:24
largest, you know, sanction awards in the civil justice
36:26
system in the last year, if not in the last decade for this
36:30
misconduct, which was, you know, gratifying if I'll be honest.
36:36
But it got us to the then point of thinking, well, they're going
36:39
to do exactly as you suggested. Now they know their goose is
36:42
cooked. A Whether I call it a diversion
36:45
or a tangent or what have you, I know that there are a lot of
36:49
lawyers who end up getting into whether it's technology or a
36:53
morass of other complicated material and then run into
36:58
challenges with judges where the defense in essence obfuscates
37:03
like here's a lot of Excel spreadsheets or here's a lot of,
37:06
you know, ones and zeros. And it's it's judge.
37:09
It's really complicated. What they say is and so.
37:12
How? Did you, did you get lucky with
37:13
a smart magistrate? Did you take this information
37:16
and package it in a very clean way?
37:19
How were you able to get it so that someone has to call balls
37:22
and strikes on lots of things? Could do exactly what the
37:26
magistrate had the courage to do here.
37:29
Both we were, we were fortunate with a judge who got it also who
37:35
understood from the years of litigation before that, I mean,
37:39
nearly everything they produced, including the stuff that they
37:42
should routinely produce, had to be brought by way of motion to
37:44
compel, right. And and so the judge understood
37:48
the state of play. They understood what we were
37:51
dealing with and and then as as time went on.
37:57
I think through a combination of, of our efforts, right and
38:00
packaging inappropriately, the judge started to get it right as
38:05
as they tend to right. Because, you know, in fairness,
38:08
I don't think this court had ever, and I know my trial judge
38:11
had, you know, had never dealt with an autopilot case before.
38:14
They were new to all of this. And, and really for us on this
38:19
data stuff. I mean, I, I had these other
38:21
cases, but I had never gone deep into the, you know, into the
38:26
data sphere that I found myself in in this case.
38:29
And so invariably through because we're also continuing to
38:33
focus group the case while all of this is going on.
38:36
I focus group the case before I had the augmented video and the
38:39
data. I focus grouped it after.
38:41
I focus grouped it before Elon was a part of the
38:44
administration. I focus grouped it after, right
38:46
you. And so that also helps with the
38:48
messaging, right? How am I going to convert these
38:50
very complex in shoes? And and and complex evidence
38:55
into Gerber baby food so that I can spoon feed it, you know, to
38:59
eight 1012 people, whoever they are sitting in the box.
39:02
And so you know to that end I had said at the beginning Tesla
39:08
initially takes the position autopilot wasn't even on the
39:12
judge six months later realized what a bunch of bullshit that
39:16
was. It was technically true, but it
39:21
was entirely disingenuous and misrepresentative of the actual
39:27
truth. And he got that same thing of we
39:35
never got the data. What do you want from us?
39:37
We, we, we never got it. And now I can say, you see it
39:41
right here. This is the code.
39:43
This is the confirmation. And they couldn't say it wasn't.
39:49
There was nothing that they could say.
39:51
In fact, I ultimately ended up taking the deposition of their
39:56
data infrastructure lead. And this man was probably in
40:01
remains today. I haven't checked his LinkedIn.
40:03
He's probably been fired because he offered truthful testimony.
40:06
But he was the most truthful of Tesla employees I've ever
40:09
deposed. It's because he didn't really
40:11
have a dog in the fight. He's the data guy.
40:14
I think you know, before Tesla he worked it.
40:17
I don't remember Ray Ban or something.
40:18
Like it just it's data infrastructure.
40:20
OK. He doesn't care what the data
40:22
is. It's how do you organize it?
40:24
How do you process it? How do you find it right.
40:27
And, and there's a teachable moment, I think, for any lawyer
40:29
listening to this or anybody. Listening to this as these cases
40:33
go forward and they're going to continue, right?
40:36
At the end of the day, this case really came down.
40:39
If you just distill it down to its, its scientific essence,
40:43
This was a case about a company being held responsible for
40:45
algorithmic decision making, right?
40:48
And we got to be thinking not about the cases we've been
40:51
working up and trying for the last 20 years.
40:53
We got to be thinking about the cases we'll be working up and
40:55
trying 1020 years from now. And this is this is where all of
40:59
this is going. And so it's critically important
41:03
to understand how this stuff is stored, how it's analyzed, how
41:07
it's processed. And this gentleman we deposed
41:10
confirmed what we knew to be the truth.
41:12
Which is, wait a second, you got the data.
41:15
Yeah. And in fact, he was the first
41:18
one to tell us truthfully that if they receive any of that
41:23
stuff, if they receive the videos or the diagnostic log, if
41:26
they receive 1 bit, he's it's a ZIP file.
41:29
You can't receive just part of a ZIP file through an e-mail.
41:33
You either got the whole thing, you got the whole enchilada, or
41:38
you got nothing. It's it's, it's binary.
41:41
It's all or nothing. And he was like looking at us.
41:45
I'm like, so if Tesla took the position that they only got a
41:49
partial upload, he's like partial upload.
41:52
There's no such thing. It's it's determinative.
41:55
They baked it in to the system. It's all or nothing.
42:00
And so again, Tesla couldn't say anything in response to that
42:04
because their own witness internally confirmed what we
42:08
were able to prove through our hacker.
42:11
So I want to go back to something that you said and, and
42:14
unpack it a little bit because I think it's important and I don't
42:17
want it lost in the story when you said that as we try cases
42:22
going forward the next 20 years that we need to be considerate
42:27
of algorithmic decision making. And I'm, I'm paraphrasing what
42:31
you said, I'd love for you to unpack that a little bit more.
42:34
Sure. I mean, at the end of the day,
42:39
when these machines make decisions, they have to perceive
42:45
the world, they have to make predictions, they have to have.
42:49
Confidence in those predictions and then they either have to act
42:54
or not act. That's really all it is.
42:58
We can make it sound way more complicated and they do a
43:03
fantastic job of that. But this is true in all forms of
43:07
artificial intelligence. This is true in what we're
43:10
seeing right now in in some of these, you know, tragic chatbot
43:14
suicide cases. At the end of the day, it's
43:18
about a prompt, some stimulus. It's a perception to the
43:23
stimulus and it's a decision making to do or not to do That's
43:30
it. And when you break it down like
43:33
that and, and go after these corporate witnesses about that,
43:38
they have to concede as much, right?
43:41
They'll, they'll come back at you with word salad and, you
43:43
know, and oh, well, that's, you know, and I routinely was, you
43:46
know, when I would use language like this and, and talk about it
43:49
in these simplistic terms, I would be routinely laughed at by
43:53
Tesla's lawyers, counsel you, you don't understand.
43:58
And they pat me on the head, right?
44:00
You know, figuratively. But that's what it is.
44:05
And so. What we have to look at is, is
44:09
you have to be able to drill into that.
44:11
And the only way you can do that is with access to the actual
44:15
data itself. Not the summary, not the table
44:19
of contents, but the actual words of the story.
44:23
And I'm not saying you need to have a computer science degree.
44:25
I mean, I, I do not, I have absolutely no tech background
44:29
whatsoever. I was a creative writing and
44:31
religion major, right? I, I, I went to, you know, I
44:34
took a liberal arts math 25 years ago.
44:36
That's the extent of my knowledge of these things.
44:38
But I know how to talk in sound bites and distill things down in
44:43
simplistic ways to convey them. As a part of a story.
44:47
Right. And, and so that's what you have
44:50
to do here. And, and despite any effort to,
44:54
to make it sound more complicated than that, you just
44:58
got to keep going back to that truth, right?
45:01
So what did it see? What's the stimulus?
45:04
What was it doing? How confident or not?
45:07
Is it in the thing that it was going to do?
45:09
And they all have different standards.
45:11
And, you know, is it from a zero to 1?
45:12
Is it a zero to five? It's, you know, one to 100 or
45:16
whatever, doesn't matter. But you know, and even this
45:20
whole concept, I mean to, to that same end, neural networks,
45:24
right? I take the director of AI at
45:27
Tesla. You know, neural network sounds
45:29
like a fancy, you know, really complex idea.
45:34
It does, it does and and it's not it's pattern recognition.
45:40
You want to teach something how to stop at a stop sign, you need
45:42
to show it. 200,000 stop signs in 200,000 locations at 200,000
45:48
times a day at 200,000 different weather conditions.
45:52
And then sooner or later, hopefully the system starts to
45:55
figure out, oh, that's a stop sign.
45:57
I need to react to that. That's all it is.
46:00
That's all it is. They're just going out there
46:02
just training. It's it's machine learning.
46:04
It's like when I taught my kid how to swing a baseball bat or
46:07
how to how to shoot a free throw.
46:09
He's got to do it 1000 times. As you're taking the deposition
46:14
of the head of AI for Tesla, it I'm going to do a little bit of
46:19
background on this. I, I happen to be in the midst
46:21
of listening to a different podcast right now of acquired as
46:25
the podcast. It's going through Google's AI
46:28
progression and, and a deep dive into that.
46:30
And so it's talking about neural networks and they talk about
46:33
something called the cat paper where they, the, the machines
46:37
started recognizing through seeing and seeing and seeing and
46:40
seeing and seeing cats. And they, they were able to do
46:44
it successfully and they 75% of the time.
46:48
And then they moved up to 85% of the time.
46:50
So the frame for this question is when you're talking to the
46:54
head of AI, do they have confidence intervals for
46:58
identifying stop signs or cars parked at the end of AT
47:01
intersection or what have you did?
47:03
Was that a piece of what you learned as you went through this
47:05
process? So.
47:06
Yes. Now where they push back is
47:11
what's, what's fascinating about Tesla is that there were very
47:15
few people who could actually describe, and this is true both
47:17
in the Miami case as well as other cases I've had who can
47:20
actually describe what that confidence interval looks like.
47:23
None of their experts are are even entirely aware of what that
47:28
is. And and, you know, and then they
47:30
try to say, well, at a certain point you're getting to our
47:32
source code and we can't, you know, possibly that trade
47:36
secreted. And you know, that's the,
47:37
that's, you know, the the secret, you know, ingredient of,
47:40
of Coca-Cola. We can't possibly give you the
47:42
recipe. But what we were able to show
47:47
and, and tell through the data is clearly the system saw all of
47:53
these things. It knew that it was coming to
47:56
the end of the road. It knew that it was approaching
47:59
a stop sign. It knew that the end of drivable
48:01
space was was there. It knew that there was a vehicle
48:04
and a pedestrian. And we were able to show as it
48:06
cycled through all of these things.
48:09
And it didn't do anything. It didn't warn.
48:12
It didn't. Slow.
48:13
It just crapped out at the end and at that point I don't really
48:19
care. What the confidence interval was
48:22
for any of that decision making, Because clearly it didn't work.
48:27
The proof was literally in the pudding.
48:31
It didn't work. So something happened here,
48:35
right? And I can't tell you, was it
48:37
supposed to be, you know, 90 percent, 95%?
48:39
Now we also knew from the data that it did have certain
48:44
confidence and it continued to cycle, but.
48:47
This goes to a larger. Issue of Tesla that we haven't
48:49
touched on, which is the reason why they remain an industry
48:54
outlier, is because they have now gone all in just on cameras.
48:59
And everybody who knows anything about autonomous vehicles will
49:04
tell you that the Holy Trinity of safety is radar, Lidar and
49:08
cameras. You need lidar for close up.
49:11
You need radar for distance. You need cameras to help to fuse
49:14
the three things together. Tesla, at least in the five
49:19
years ago vintage, was doing radar and cameras.
49:22
Elon never wanted to do Lidar, whether it was aesthetics,
49:27
whether it was cost, whether it was hubris, probably a
49:30
combination of all. They never did that.
49:35
You know, say what you will about Waymo.
49:37
They have followed the the tripod and and they also three
49:43
dimensionally map everywhere they go and they limit where in
49:47
the world the vehicles can be used.
49:50
And that's what other responsible manufacturers at the
49:53
time we're doing right. And so to this end of of kind of
49:57
how the the machine both learns and how it makes decisions.
50:01
Part of the problem and why Tesla will never be able to get
50:05
out from underneath itself is because of the steadfast refusal
50:10
to follow the, the, the, the formula that everybody else
50:13
knows has to be followed. In fact, post our crash, and I
50:16
think it was in 22 or 23, they went all in just on cameras.
50:21
They started to walk that back. And I know most recently have
50:24
continued, they've doubled down. Yet again, to go all in on
50:28
cameras, you cannot safely operate autonomous vehicles in
50:34
computer vision solely on cameras.
50:37
It's not enough. Data, it doesn't tell it enough.
50:40
And not for nothing, the megapixel on a 2024 Tesla coming
50:47
off the line of those cameras is less than my iPhone that's
50:52
sitting in my pocket right now. So when you not only are just
50:56
relying on cameras, but you're doing it with shitty cameras,
51:01
it's no surprise, right? This, this is going to continue
51:05
to happen. One of the things that that that
51:08
if one is paying attention to in terms of industries is Tesla's
51:11
big push into trucking. And my guess is you will
51:17
probably expect to see a lot of these types of cases coming from
51:21
much bigger, much heavier, much more dangerous vehicles if Tesla
51:28
has continued to allow to just rely on cameras 100%.
51:32
Now there are other companies, in fact, one of the original
51:36
founders, one of the original autopilot directors, one of the
51:41
ones who raised his hand back in 2014 or 15 and said you can't
51:47
call it full self driving. It's not.
51:53
And then a week later, Elon did a whole press conference calling
51:56
it full self driving. And this gentleman resigned very
51:59
shortly thereafter, is now with a company called Aurora and
52:02
Aurora is trying to do autonomous vehicles with respect
52:06
to fleets of trucks. So you basically have these pods
52:10
of trucks. So rather than 20,000 lbs,
52:12
you're going to have 100,000 lbs of vehicle, you know, heading
52:15
down in a in a fleet, which is a terrifying notion in and of
52:18
itself, but they are trying to do it in a way that
52:23
incorporates. All of those things and are
52:25
trying to do it in a way that is safer.
52:29
Limiting where it can be used, limiting how it can be used,
52:33
trying to rely on the Holy Trinity of safety and things of
52:36
that sort. So again, I, I think both with
52:41
respect to cases involving autonomous vehicles, but I also
52:45
think cases involving algorithmic decision making more
52:48
generally and you know, and artificial intelligence more
52:51
generally. I think in the short horizon
52:54
over the next, let's say, three to five years, the role of the
52:57
civil justice system, if it's to have one, is to go after these
53:02
outliers, right? Because so many of these folks,
53:07
which are tech companies that have got into transportation,
53:11
are running it with the Silicon Valley ethos of not really
53:15
giving a shit. Move fast and break things,
53:18
including people. Exactly.
53:20
And, and do it at the at the expense of safety and do it all
53:24
in an effort to push more metal and to pump the stock, right?
53:28
Understand Tesla's entire trillion dollar valuation is all
53:31
in on autopilot. That's it.
53:33
I mean, they make an OK car. It has some decent crash
53:36
worthiness we're seeing now with the door handle issues.
53:39
People are being entrapped and burning and drowning.
53:42
And I have a few of those cases I'm looking at right now which
53:44
are just horribly tragic again. Which is in large part because
53:48
Elon didn't want a door handle inside because door handles
53:51
aren't cool. The problem is if it gets into a
53:54
crash and the low voltage goes out, the button doesn't work.
53:58
And now you're trapped. But I think that's where the
54:01
civil justice system can can be a deterrent is we have to go
54:08
after the outliers who are doing it in a less Safeway, right?
54:13
Preventable deaths should be prevented, period, full stop.
54:20
And, and there are ways to do this.
54:23
In fact, if you look at at some of the other, like German
54:27
manufacturers, for instance, many of them are just saying,
54:31
forget this partial autonomy thing.
54:33
We're skipping it. We can't do it.
54:35
Well, say what you will about the Germans, they know how to
54:38
build a car. Yeah.
54:39
And, you know, and so I think there's a lot of wisdom in that.
54:43
Some of them are going, no, no, we're going, we're going
54:45
essentially full autonomy. You know, level 4, which is
54:47
where it's kind of like the way MO, right?
54:49
There's still a wheel and a pedals and stuff.
54:51
Level 5 is the stuff where there's actually no wheel and
54:54
pedals. And we're not there at a
54:55
regulatory level and not there, I don't think at a technological
54:59
level yet either. But they're just saying that's
55:02
the only way we're going to do this.
55:04
And so we're going to wait and we're going to study it and
55:07
we're going to do R&D for a decade to be sure that we get it
55:11
right. And that's the exact opposite.
55:15
Of what Tesla has done. That's the exact opposite of
55:18
what we've seen open AI do they they go out, they push an
55:23
update, right? And that's another thing.
55:26
I mean, we're talking about vehicles that weigh thousands of
55:28
pounds hurdling down the freeway at 80 miles an hour.
55:31
A lot of the problems in Tesla's historically would have been
55:34
called a recall. Now they're called a software
55:36
update, right? That's a problem.
55:40
You know, back in the day, we'd all get a little postcard in the
55:41
mail and have to go drive your car to the, you know, to the
55:44
dealership and, and get a loaner and, you know, or take a cab
55:48
home or whatever we used to do in the before times and now.
55:52
They just push out a software update.
55:55
One of the Autopilot leads was talking about and testified at
55:59
trial how when there's an incident, they push out an
56:02
update and there's an incident and then they push out an update
56:06
and again, words matter. I was sure to remind the jury
56:11
that what he called an incident, according to the families and
56:14
the people involved, they called it a funeral, right?
56:17
Generally, someone lost their life every time an incident
56:21
occurred. They're incredibly reactive.
56:25
They're not proactive. They're not taking the steps to
56:29
prevent the harm. They respond to the harm.
56:34
And I'm sorry, but that's just ass backward and and the jury
56:39
understood that. And and I think that is again,
56:43
the bigger role of what the civil justice system in the
56:48
intersection with AI and algorithmic decision making is
56:52
going to look like over the short horizon.
56:56
Is is calling that out and holding these companies
57:00
accountable for making these choices and using that move fast
57:04
and break things ethos instead of elevating safety and in
57:10
people's lives over profits and greed.
57:14
Thank you for listening and please join us next week as we
57:16
continue this conversation. Like what you heard?
57:20
Share us with a colleague and leave us a five star review on
57:23
your podcast platform of choice. Please e-mail us at podcast at
57:26
Coopers dot law with questions, comments, feedback and
57:30
suggestions for all of you doing justice out there.
57:33
Happy hunting.