Latest / In the Loop AI / Tesla Rolls Out Full Self-Driving: How Your Car Can Make You Money

Transcript

Tesla Rolls Out Full Self-Driving: How Your Car Can Make You Money · In the Loop AI

0:00
Josh: For the first time ever, cars are driving around a major American city with
0:03
Josh: literally nobody inside.
0:05
Josh: No driver, no passenger, no safety monitor, just software sensors and a steering
0:10
Josh: wheel turning itself through the streets.
0:12
Josh: But soon that steering wheel even is going to disappear.
0:15
Josh: It's clear now that I think like
0:16
Josh: autonomy isn't this abstract science fair project. It's an actual product.
0:20
Josh: On the episode today, we're going to talk about how Tesla has been able to deploy
0:23
Josh: these driverless tests starting in Austin, Texas, hopefully throughout the country
0:26
Josh: very soon, and the two philosophies between the top competing companies.
0:29
Josh: We have Waymo and Tesla, of course.
0:32
Josh: And I think we're going to make the case that if Waymo was Uber,
0:34
Josh: Tesla is more closer to iOS.
0:36
Ejaaz: Josh, you've owned and have driven your Tesla for about eight years, right?
0:41
Ejaaz: And I think for the last two years, you haven't even touched the steering wheel. Is that right?
0:46
Josh: The progress of full self-driving has been remarkable, particularly over the
0:51
Josh: last two years. I've had...
0:53
Josh: These cars now with full-slope driving for eight years now. So I've seen the
0:56
Josh: entire ramp up from very scary white knuckle driving to I feel like I could
1:01
Josh: take a nap and it'll be totally fine.
1:02
Josh: And the difference happened in the last two years because they switched over
1:05
Josh: from hard-coded numbers.
1:07
Josh: So they use like hard-coded C++ where if it saw a stop stein,
1:10
Josh: it knew that it needed to stop.
1:12
Josh: They switched over fully to neural nets. And that was the thing that made a
1:15
Josh: difference. Once they switched to neural nets, the progress got unbelievable.
1:18
Josh: And yeah, now there's, I don't even need to touch the steering wheel.
1:21
Josh: It actually just drives me around anywhere I want to go.
1:23
Josh: It parks in the parking lot. It backs out of the driveway. It's unbelievable.
1:27
Ejaaz: What is like the longest journey you've done? Because you drive home like most
1:31
Ejaaz: weekends, right? How long is that?
1:32
Josh: Yeah, most of the time, like 50 miles, no problem. No oversights, door to door.
1:36
Ejaaz: So it's navigating New York traffic and everything until it parks and your parents like-
1:40
Josh: That's been the most impressive thing is the New York City navigation.
1:44
Josh: Because it seems like everyone in New York City wants to get hit by a car.
1:47
Josh: They're trying to get hit.
1:48
Josh: And the fact that these cars are able to recognize that and navigate through
1:52
Josh: all of that. And there's a lot of weird stuff that happens.
1:54
Josh: There are cars parked in the middle of the road. There's people who are working
1:57
Josh: on construction, giving hand signals.
1:58
Josh: It's now capable of reading and understanding all these things and actually
2:01
Josh: navigating through the world. And it's so impressive.
2:03
Josh: EJS, we're going to have to, we'll have to record a video soon of us actually in a car doing a demo.
2:08
Josh: Maybe we'll record a podcast in the car as it's driving because it really,
2:11
Josh: it's truly remarkable. And if you haven't experienced it, it's tough to describe,
2:15
Josh: But it is genuinely a full self-driving cyber cab out in production right now.
2:20
Ejaaz: That's great. It's going to be like carpool karaoke, but just two nerdy guys talking about tech.
2:25
Ejaaz: That's actually a good idea. Okay, let's get right into the news.
2:30
Ejaaz: Something was spotted over the weekend in Austin, Texas, Josh.
2:35
Ejaaz: And it was a normal regular Tesla Model Y, except there was no one in the front driving it at all.
2:42
Ejaaz: And this was later confirmed by
2:45
Ejaaz: Elon himself as the first test cases for unsupervised full self-driving.
2:50
Ejaaz: Now, if you're new to this podcast or if you're new to Tesla in general,
2:53
Ejaaz: Tesla's FSD or full self-driving is basically their autonomous driving program
2:57
Ejaaz: or software stack that allows their cars to drive itself, similar to what Josh
3:02
Ejaaz: has exactly explained, actually.
3:04
Ejaaz: And why this is such a major leap forward is typically up until before the weekend,
3:11
Ejaaz: these cars required at least one human person to be in the front seat.
3:17
Ejaaz: They didn't need to be in the driver's seat, but they needed to be somewhere
3:19
Ejaaz: in the passenger seat just in case the car got into any kind of an accident.
3:22
Ejaaz: And it marks the transition from level two autonomous driving,
3:26
Ejaaz: which is what Teslas have been fully capable of doing for a while now,
3:29
Ejaaz: to level four, which is this driverless addition or rather subtraction.
3:34
Ejaaz: And this worked out really well for the stock.
3:37
Ejaaz: On monday's market open tesla closed at the highest weekly and monthly close
3:42
Ejaaz: in tesla's history 475 dollars and 31 cents which is just insanely bullish it
3:48
Ejaaz: was up five percent uh on the day end
3:50
Josh: And it's funny because if you look at the stock you would think the whole world
3:53
Josh: is aware of what's going on but the reality is this is not true this is a great
3:57
Josh: visual that we have on the screen and for those listening it's basically it's
4:00
Josh: a visual showing that the population of the u.s is 343 million people.
4:04
Josh: Of that, 243 million are actual drivers.
4:07
Josh: Of those 243 million, four of them are Tesla owners. And of the four million Tesla owners,
4:14
Josh: only half a million people have actually used full self-driving.
4:17
Josh: So the reality of this is that only 0.145% of the population have experienced
4:23
Josh: the technological revolution that is being rolled out as we speak in real time
4:28
Josh: to cities like Austin, Texas.
4:30
Josh: And this, I believe, is going to lead to one of those moments,
4:33
Josh: like the, oh my God, it's real moment, where things tend to advance very slowly.
4:39
Josh: But if you're not in the know, it appears as if it's happening very quick.
4:42
Josh: And there's a reality in which very soon, that someone's going to walk outside
4:45
Josh: and get into a full self-driving car and freak out because they had no idea
4:49
Josh: that this technology was even close to being ready, let alone here today.
4:53
Ejaaz: What's really funny is of this 500,000 people that do know about it,
4:57
Ejaaz: Josh, one of them is my auntie.
4:58
Ejaaz: And I just texted her saying, hey, you need to give this thing a go.
5:01
Ejaaz: And she replied, I'm terrified. I don't want to give it a go.
5:04
Ejaaz: So even the people that do know about it are still kind of like they haven't
5:07
Ejaaz: jumped the chasm across the chasm to yet to get there.
5:11
Ejaaz: But Josh, I want to take a look at like what these things actually look like,
5:14
Ejaaz: because they're not just rolling out this software upgrade, right?
5:18
Ejaaz: It's going to become an entirely autonomous taxi service called the RoboTaxi.
5:23
Josh: The cyber cab. And there's two parts of this story. So one of it is the software,
5:27
Josh: one of it's the hardware.
5:28
Josh: The hardware story is the cyber cab. The software story is the one that we just
5:31
Josh: mentioned though, which is the existing fleet of 8 million Tesla vehicles that
5:36
Josh: will be eligible to have this ability via over the air software update.
5:40
Josh: So there is a person on this planet who is going to type a command into a computer,
5:44
Josh: send this update to millions of cars, and overnight they will wake up and become
5:48
Josh: fully autonomous cyber cabs.
5:50
Josh: Now that is one thing, That is the existing fleet.
5:53
Josh: The new fleet is what we're looking at on screen right now. That's the cyber cab, the robo taxi.
5:58
Josh: This is the crown jewel of autonomy as we move into this next era.
6:02
Ejaaz: Josh, sorry, I have to say this looks like a Hot Wheels car that I used to buy
6:07
Ejaaz: as a kid. What is this? It's like it's like die cast.
6:10
Josh: It's funny you say that because the Hot Wheels car that you had as a kid was
6:13
Josh: made via die cast machine, which is injection molding.
6:16
Josh: So basically you have this big metal mold you punch a bunch of this like hot
6:20
Josh: molten stuff into it and then out pops this very cooled,
6:24
Josh: um piece of a car and it looks like a car and what
6:27
Josh: tesla's done is they've created this small die cast model that you
6:29
Josh: use for your hot wheels back in the day but they made it
6:32
Josh: into the biggest casting model in the planet and it's
6:35
Josh: so large that it prints out actual full-size cars like the one you're seeing
6:39
Josh: now which comes with a bunch of benefits because a lot of times if you have
6:42
Josh: a car one of the pain points is your paint it scratches it's expensive to do
6:46
Josh: there's no paint because it's injection molded the actual material that they
6:50
Josh: mold it with is the color of the paint So there's no paint on this car.
6:53
Josh: It's just the raw material that's on the outside.
6:55
Josh: And because of that, there's other benefits like wires. There's miles and miles
6:59
Josh: of wires in traditional cars and electronics.
7:02
Josh: All of that's just printed right in. So it's this really efficient way of doing
7:05
Josh: it, but it also has the second order effect of kind of looking like a spaceship.
7:08
Josh: I don't know. I think it's pretty cool.
7:09
Ejaaz: Okay, so do I. But of course, we can't really talk about cybercab or autonomous
7:15
Ejaaz: driving without bringing up the obvious elephant in the room,
7:18
Ejaaz: which is Waymo, which is already out there.
7:20
Ejaaz: They have a paid taxi service, which does hundreds of thousands of paid trips
7:24
Ejaaz: every week. It's just rolled out in London, my hometown.
7:27
Ejaaz: And so the question that comes into mind is, can it compete?
7:31
Ejaaz: And at least from a cost perspective, if that's the first framework we apply here,
7:35
Ejaaz: robo taxis are just or like cybercams are
7:38
Ejaaz: just way way cheaper we've pulled up a
7:41
Ejaaz: graphic here which shows what you gotta keep scrolling
7:44
Ejaaz: down yeah there's so many
7:47
Ejaaz: which shows us what one million dollars worth
7:50
Ejaaz: of waymo gen 6s which is their latest versus one
7:53
Ejaaz: million dollars worth of tesla cybercams and uh you have about 10 waymos for
7:58
Ejaaz: the price of one million dollars and for robo taxis or cybercabs you get five
8:02
Ejaaz: and a half times more than that we've got 55 this is insane josh like the cost
8:08
Ejaaz: must be like super cheap for these things and
8:10
Josh: This makes all the difference so as we start to get into differences between
8:13
Josh: waymo and tesla one of the most important things is the form factor what does
8:17
Josh: it actually look like and why does it look that way
8:19
Josh: Um, for those who are not familiar with the Waymo system, basically what it
8:23
Josh: is, is the current way that it exists.
8:24
Josh: And these are out on the road today is it's a Jaguar car that costs about $75,000,
8:28
Josh: but then it costs an additional $100,000 to equip it with 29 cameras,
8:33
Josh: five LiDAR sensors, and it's very custom. It requires a lot of human interfacing.
8:37
Josh: So the net cost of those cars are about $175,000.
8:40
Josh: What we're seeing on screen now is a much cheaper version of this car,
8:43
Josh: the next generation Waymo,
8:45
Josh: which they're hoping to get down to $85,000, but still
8:48
Josh: does not compare to the cyber cab tesla cyber
8:51
Josh: cab is looking to price itself at thirty thousand dollars per
8:55
Josh: car and the reason this is important is because of
8:57
Josh: the the cost per mile and we're going to get into cost per mile but the
9:00
Josh: innovation that we're showing on screen now how they're able to do it is this
9:03
Josh: thing called the unboxing process so easy as we briefly talked on the
9:06
Josh: die cast process the casting how they're able to quickly mold them
9:09
Josh: together the amazing thing the zinger here the big
9:12
Josh: punch line is tesla is able to produce one of these cyber cabs
9:15
Josh: every six seconds it'll roll off production line
9:18
Josh: so if you're watching the end of a production line every six seconds a new car
9:22
Josh: will be coming off of this line why because of the die casting and because of
9:25
Josh: this unboxing process so typically when you assemble a car it goes down a singular
9:29
Josh: production line it has to be assembled and then disassembled to paint it and
9:32
Josh: then reassembled for final production this does it all in one shot it's assembled in six areas.
9:37
Josh: Pressed together in one location and every six seconds a new one rolls off the
9:40
Josh: line even if waymo gets this next version it's still going to be three times
9:43
Josh: the cost of tesla cyber cab which has big downstream effects to you and i the
9:47
Josh: user in how much it's going to cost to actually use these services i just.
9:50
Ejaaz: Did some uh back of the napkin math and i figured if each waymo costs around
9:55
Ejaaz: a hundred grand um if they wanted to deploy like let's say 500 000 more of these Waymo cars,
10:02
Ejaaz: which is, you know, not crazy to think about, that would cost them $50 billion
10:06
Ejaaz: worth of upfront CapEx, right?
10:09
Ejaaz: Now, if they wanted to do this for Tesla, they could just push in a software
10:14
Ejaaz: upgrade for just 10% of their existing fleet.
10:19
Ejaaz: So I'm talking about cars that already exist, Josh. On the road right now,
10:23
Ejaaz: there's like 5.5 million eligible Teslas that can get the software upgrade and
10:29
Ejaaz: have the entire FSD benefits that we've just described to you,
10:33
Ejaaz: just 10% of those can get that benefit right now from a single software push.
10:37
Ejaaz: So it just goes to show that like Tesla versus Waymo is a completely different
10:41
Ejaaz: beast because of the way that they've constructed FSD in this particular case.
10:46
Ejaaz: And so it costs Tesla basically $0 because the customers that are purchasing
10:50
Ejaaz: this, all fleet operators that want to buy these cars in the future can And
10:53
Ejaaz: just do so for a cheaper cost, $30,000 versus 100 grand really kind of amplifies over time.
10:59
Ejaaz: But Josh, I've noticed something, which is these robo taxis,
11:03
Ejaaz: or rather the cybercabs, have only two seats in them.
11:07
Ejaaz: When Waymo has five, surely the Waymo is better in that way, right?
11:11
Josh: Well, this chart begs to differ. This chart says that 90% of robo taxi trips
11:15
Josh: have been with two people or less. so what the differences you're seeing is
11:20
Josh: a hyper customized approach versus a traditional approach i guess we're like
11:26
Josh: waymo is basically saying well we have the
11:28
Josh: software but we don't really have the hardware so we're just going to take an existing car
11:31
Josh: strap some sensors on it and apply our super fancy software tesla's
11:35
Josh: taking a total first principles approach so looking at this and saying well
11:38
Josh: if nine out of ten trips are with two people or less let's just strip out the
11:41
Josh: back seat and make the car even more efficient and more cost effective so the
11:45
Josh: cyber cab only has two seats in it for two passengers whereas the waymo has
11:49
Josh: five people so the assumption is that you will mostly need it for two people.
11:53
Josh: If you need it for five, well, maybe get a Waymo or just get two cyber caps,
11:57
Josh: because I mean, the idea is that they will be cheap enough that you can get
12:00
Josh: a couple for the price of one Waymo ride.
12:03
Ejaaz: I saw this amazing breakdown of Tesla versus Waymo by Marcelo and I've pulled
12:09
Ejaaz: up his tweet and he goes, let's talk about scale first.
12:14
Ejaaz: Waymo just reported over 100 million fully autonomous miles driven,
12:18
Ejaaz: while Tesla has 6.8 billion
12:22
Ejaaz: full self-driving miles driven, over 10,000 miles per minute.
12:27
Ejaaz: And if you're wondering how on earth they were able to get this scale,
12:30
Ejaaz: it's because the existing 5.5 million Teslas that are eligible for this FSD
12:35
Ejaaz: upgrade gets input data every time a human takes it for a drive.
12:40
Ejaaz: So its data can get fed back to its single generalized neural network that trains
12:45
Ejaaz: FSD and gets fed to every single other Tesla that is operational on the planet
12:50
Ejaaz: that has this FSD software, which is just the coolest thing ever.
12:55
Ejaaz: It's ridiculous. Yeah. So it says here that Tesla has 5.1 million cars with
13:00
Ejaaz: the AI4 computer chip, which is just like kind of like the base chip.
13:04
Ejaaz: Actually, Josh, you were telling me before we started recording this that it's
13:06
Ejaaz: actually a chip prior to this that is also capable for this,
13:09
Josh: Right? They're going to be building software for that one as well.
13:11
Josh: So they'll all be eligible.
13:12
Ejaaz: Okay. Okay. So that would mean that Tesla's fleet is 2,000 times larger than
13:19
Ejaaz: Waymo's when it comes to like a mile advantage. But wait. Today.
13:23
Ejaaz: Yeah, today. Today. And then it goes, Tesla's FSD miles driven is growing at
13:28
Ejaaz: over 10,000 miles every minute. That is crazy.
13:31
Ejaaz: Waymo is doing 450,000 paid drives per week.
13:34
Ejaaz: So the average Waymo ride is about six miles in length. This means Waymo is
13:38
Ejaaz: collecting 2.7 million miles of data every week from its paid rides.
13:43
Ejaaz: That's only 268 miles per minute versus Tesla's 10,000 miles per minute.
13:48
Ejaaz: Josh, if I'm like an analyst looking at this from an investment perspective,
13:52
Ejaaz: also from a customer experience and safety, like which car do I trust more when
13:56
Ejaaz: it comes to autonomous driving?
13:57
Ejaaz: It's going to be the one that has the most experience and mileage.
14:00
Ejaaz: How does Tesla lose this?
14:02
Josh: I don't believe they do in any reality, and I'm not sure it's particularly close.
14:07
Josh: If you look at Waymo's strategy, it very much feels like they're shipping a
14:11
Josh: service where they have this full self-driving software, but they don't own the full stack.
14:17
Josh: And therefore, they're not able to manufacture cars quickly.
14:20
Josh: And they also haven't really approached it correctly because theirs is very
14:23
Josh: expensive. They have LiDAR sensors, which we're going to get into.
14:25
Josh: And they also have it geofenced. So the way that Waymo expands is they have to first scan a region,
14:30
Josh: get a high detailed map, and then they have to go city by
14:33
Josh: city and it takes a very long time for this to roll
14:36
Josh: out whereas the reality is tesla they're more shipping instead
14:39
Josh: of a service they're shipping a model and their bet is that this vision first
14:42
Josh: system plus the fleet data plus this continuous trading data gives you autonomy
14:47
Josh: that is not specific to a city but generalized so then you take that technology
14:52
Josh: and then you wrap it into a service but the reality is that when tesla rolled
14:56
Josh: out their service in Austin.
14:58
Josh: I believe Waymo had been there for already about 136 days.
15:02
Josh: And in 42 days, Tesla was able to four times the original Waymo size because
15:07
Josh: they have this generalized way of doing it where the neural network is smart
15:11
Josh: enough to go anywhere in the world and drive fully autonomously. Whereas Waymo is not.
15:15
Josh: Waymo not only has the hardware stack, but they don't have the software that
15:18
Josh: can scale because they haven't gone deep in these neural nets.
15:20
Josh: And the funny thing with neural nets, just a quick aside, is that they're trained
15:25
Josh: learning based on what they see instead of programming them.
15:27
Josh: So previously, they would program a car to stop at a stop sign.
15:31
Josh: But the car was never told that in this new version. It was inferred that.
15:34
Josh: But when it was reading drivers, it was watching drivers drive.
15:38
Josh: The cars just kept rolling through the stop signs because no one really stops at a stop sign.
15:41
Josh: So FSD started rolling through stop signs. And then the Highway Safety Committee
15:44
Josh: got really upset because they were like, wait a second, your full self driving
15:48
Josh: software isn't following the rules. It's rolling through stop signs.
15:50
Josh: So they had to actually create artificial data of humans stopping.
15:54
Josh: That way you could teach the car to stop and actually follow the rules because
15:56
Josh: it was too much like a human.
15:58
Ejaaz: This was really the big bet that
16:01
Ejaaz: elon made almost a decade ago which is uh visual
16:05
Ejaaz: data or vision learning uh in a single
16:08
Ejaaz: neural net is way better than slapping a
16:11
Ejaaz: bunch of detectors and sensors on a car he
16:14
Ejaaz: thought maybe like if we just get visual input you could train a car much more
16:18
Ejaaz: faster and when we look at the stats today 10 000 miles per minute uh across
16:23
Ejaaz: an entire fleet of 5.1 million cars versus 268 miles or whatever that stat was
16:29
Ejaaz: for Waymo. It's just obvious at this point.
16:32
Ejaaz: But I did read that for a while, Elon was using LiDAR for Tesla cars.
16:37
Ejaaz: In fact, I think he uses it to pre-train, and I say pre-train intentionally
16:41
Ejaaz: here because it sounds like he's training a model for his Tesla cars, right?
16:44
Ejaaz: He doesn't exclusively not use LiDAR. He does use it. Is that right?
16:47
Josh: This is the most commonly wrong misconception about Tesla, where everyone believes
16:51
Josh: that Tesla hates LiDAR. And the reality of it is that it is just not true.
16:56
Josh: And it hasn't been spoken much, but the idea is that they do use LiDAR quite
17:00
Josh: extensively, actually, and they like LiDAR as a sensor in the training time.
17:05
Josh: So when cars are out collecting training data, when they're testing these things,
17:09
Josh: the cars have LiDAR strapped to them.
17:10
Josh: In fact, if you see any of these vehicles driving around that they've been running
17:13
Josh: demos on, they have these big LiDAR things on top and they're really collecting a lot of data.
17:17
Josh: The arbitrage opportunity occurs during test time when you actually package
17:21
Josh: up these learnings and you deploy to the fleet
17:24
Josh: They've taken what they've learned from the LiDAR and they have removed it,
17:28
Josh: but they've compressed it into a way that the car can do just with cameras.
17:31
Josh: So the car benefits from the safety of LiDAR in training, but does not require
17:35
Josh: it in production when it's out in the real world driving.
17:37
Josh: And that arbitrage that they're collecting is huge because the cost of these
17:40
Josh: sensors is tens of thousands of dollars per car that they're removing.
17:44
Josh: And the idea is that, well, if a human eyeball can see, and if we as humans
17:48
Josh: can drive anywhere, then cameras with, what is it? I think there's eight cameras
17:52
Josh: on these cars now with eight eyeballs can not only see much better,
17:55
Josh: but they can then therefore drive much better than we can.
17:57
Josh: And it doesn't require all the extra sensors that cost so much money.
18:00
Ejaaz: Yeah, the more I learn about this, Josh, I keep seeing Waymo or LiDAR specifically
18:06
Ejaaz: as this really kind of clunky mechanism.
18:08
Ejaaz: And this isn't to kind of like crap on it. But if you kind of like,
18:12
Ejaaz: think about this over the long term, it just seems that Vision scales way easier
18:18
Ejaaz: and way better, like higher fidelity.
18:21
Ejaaz: Like if I am Waymo and let's say we enter a new city or a city gets like an
18:25
Ejaaz: extension or a new district, you now need to go map out that entire area, update your sensors.
18:31
Ejaaz: Your sensors won't work as well in fog or during the dark as Vision does.
18:35
Ejaaz: It just seems to have so many of the odds stacked against it.
18:39
Ejaaz: I kind of think of it as, not to put my AI hat on all the time,
18:44
Ejaaz: but it sounds like LiDAR is kind of like used for pre-training.
18:48
Ejaaz: So it's like to train the corpus or the base level of the model.
18:52
Ejaaz: Elon's kind of approach with visual learning is the post-training,
18:56
Ejaaz: the reasoning, which typically so far for the last at least six months has made
18:59
Ejaaz: AI models way, way more smart.
19:02
Ejaaz: Like that's the kind of analogy that at least applies for me here.
19:04
Josh: Yeah, no, that's absolutely right. And the point that you made earlier I think
19:08
Josh: is an important one because if you asked any analyst on the street or if you
19:12
Josh: asked any person on the street who's winning the cyber cab robo taxi wars currently
19:16
Josh: the clear and obvious answer is Waymo they have,
19:19
Josh: orders of magnitude more rides. They actually have a lot more cars than cyber
19:23
Josh: cabs that are in the world actually giving people autonomous rides.
19:26
Josh: But the truth is that it does not scale quickly.
19:31
Josh: In fact, the earlier Waymo rides, they were giving rides over a decade ago.
19:35
Josh: The difference in approach between the two allows Tesla to scale so much faster.
19:41
Josh: So while Waymo has a very clear and obvious lead now, it is very clearly and
19:45
Josh: obviously deteriorating rapidly.
19:47
Josh: And by this time next year I suspect that lead will be zero if not they will
19:51
Josh: be behind by a significant margin and it's all due to this pre-training post-training
19:55
Josh: phase and just the way that they're really considering all these things from
19:58
Josh: first principles the vertical integration they're building the car they have
20:00
Josh: the software stack and it's ready to go in any place in the world that has roads.
20:05
Ejaaz: Yeah, I think one really cool unlock of all of this, Josh, is that transport
20:10
Ejaaz: is just going to be reduced to a cheap commodity that anyone can access.
20:14
Ejaaz: And, you know, you don't need to blow tens of thousands of dollars on a car.
20:17
Ejaaz: You can just hop into a cyber cab and get from A to B.
20:21
Ejaaz: I was looking into the economics cost behind this and roughly with the new Model
20:25
Ejaaz: Y RoboTaxi, it'll run at about at a cost of 60 cents per mile.
20:31
Ejaaz: So let's say Tesla decides to charge $1 per mile.
20:36
Ejaaz: Tesla will pocket a 40% margin per mile, which is way, way higher than any kind
20:41
Ejaaz: of ride-hailing service that exists today.
20:43
Ejaaz: And it just goes to show that the investments that Elon made a decade ago really
20:47
Ejaaz: translates into both economical and scalable costs for his models today. It's just super cool.
20:52
Josh: Yeah, and there's this important distinction to be made with this post,
20:55
Josh: where that 60 cents per mile is referencing the cost per mile of a Model Y.
20:59
Josh: That's the car that I drive. That's the car that is publicly available.
21:01
Josh: The Cybercab, when it starts production in Q2 of next year, brings that cost
21:05
Josh: per mile from 60 cents down to 20 cents because the cost to produce the car is much lower.
21:11
Josh: The efficiency, the cost per kilowatt, the amount of range that the car will
21:15
Josh: get per kilowatt of energy is much lower.
21:17
Josh: It's the most aerodynamically efficient car in the world.
21:20
Josh: It's the most efficient car in the world, and it's built that way from the ground
21:23
Josh: up. So that cost drops even lower, which brings Tesla's margins much higher.
21:28
Josh: And what's funny is, I mean, comparing against Waymo, some of
21:31
Josh: these rides the shorter ones cost up to five to six dollars per
21:34
Josh: mile with typical trips costing up to three dollars
21:36
Josh: per mile and the tesla team has a
21:39
Josh: very clear trajectory to reaching 60 cents per mile today but 20 cents per mile
21:44
Josh: by the end of next year and 20 cents relative to six dollars is a gigantic delta
21:50
Josh: that starts to really change the unit economics of how we move things around
21:53
Josh: the earth as a whole when things cost 20 cents to transport per mile privately and
21:59
Josh: Um, that begins to change things. Like it starts to make it economically viable
22:03
Josh: or even economically incentivized to never even need to own a car because the
22:07
Josh: ownership cost of owning a car will be higher than it would be to just call
22:11
Josh: up one of these cyber caps to take you anywhere you ever want to be in the world.
22:14
Josh: And that doesn't include delivery.
22:15
Ejaaz: I just don't see a world where people are forced to buy cars in the future.
22:18
Ejaaz: I just think it'll be, I don't want to say everything becomes a subscription,
22:22
Ejaaz: but it seems like people will just kind of like go on a paper ride basis. Yeah. Yeah. It could be.
22:27
Josh: Like the ballers in the future will own their own cyber cabs and they'll have
22:30
Josh: their own custom matted out fancy like tits, but they're not driving it.
22:35
Josh: There's no steering wheel or pedals in this car.
22:37
Josh: And that's probably the difference is it will become economically unviable for
22:41
Josh: most to have a car, nor will they want one.
22:44
Josh: Who wants to spend their time driving when you could just be driven around by a private chauffeur?
22:48
Ejaaz: Yeah, it's funny. We went from having to buy your own car and drive it to,
22:52
Ejaaz: oh, get someone to drive for you to get no one to drive you.
22:56
Ejaaz: And it's your own personal space.
22:58
Ejaaz: And it's going to be cheaper per ride, which is awesome.
23:01
Ejaaz: But despite all of this, if I went to my friends today, and I have,
23:06
Ejaaz: and I've said like, look how cheap this is, look how affordable it's going to
23:08
Ejaaz: be. their number one bit of criticism for Tesla specifically,
23:12
Ejaaz: Josh, has been around safety.
23:13
Ejaaz: They're like, oh, I've seen a
23:14
Ejaaz: bunch of these viral videos go around where Tesla's crashed or something.
23:17
Ejaaz: And the truth is, that's just not true.
23:20
Ejaaz: In fact, recently, Tesla Cybertruck specifically was rated as one of the safest vehicles out there.
23:29
Ejaaz: And if you just dig into the data beyond just that award,
23:32
Ejaaz: you'll see that for every 6.36 million miles driven, as is demonstrated in this
23:38
Ejaaz: tweet of where people were using autopilot technology, only one crash came from
23:44
Ejaaz: all of those rides, which is just insane.
23:46
Josh: This is an important point. Being in a Tesla that is full self-driving is the
23:50
Josh: safest experience on the road, period.
23:53
Josh: There is no safer car and there is no safer software than the current Tesla stack.
24:00
Josh: And this is important because if we are making this transition,
24:03
Josh: people need to be highly incentivized to do so.
24:05
Josh: And one of the largest causes of death, I don't remember the numbers off the
24:09
Josh: top of my head, but it is an astronomical amount of people that die every year to car crashes.
24:13
Josh: And with this new software stack combined with these Tesla cars,
24:17
Josh: like you just mentioned, they're the number one crash safety tested cars in
24:21
Josh: the world every year since inception.
24:23
Josh: Across the line even that big metal stainless
24:26
Josh: steel cyber truck is safer than your like
24:29
Josh: honda pilot or whatever you're driving around today by like a
24:32
Josh: lot in fact they invented a brand new rating
24:35
Josh: just for these cars because they're so much safer and that's
24:39
Josh: one part of it but the second is the software when you're driving you're a
24:42
Josh: human who makes errors who only has two data points that can process
24:45
Josh: them so quickly and that variance happens a lot
24:47
Josh: by distractions or whether you're tired or not and with
24:50
Josh: a full self-driving robo taxi like this there's eight
24:53
Josh: sensors that are always on that have a full 360 picture of everything
24:56
Josh: all at once and this has happened to me on multiple occasions where a
24:59
Josh: car will come from behind you and kind of swerve into your lane and and your
25:02
Josh: first urge is to turn out of the way but you don't know if there's a car in
25:05
Josh: the lane next to you the cyber cap has all of this in real time and they're
25:08
Josh: able to keep you so much safer so that's why you see these numbers where miles
25:12
Josh: driven per one accident is getting higher and higher and higher because more
25:15
Josh: people are using it and it's getting better and it is the clear and obvious
25:18
Josh: safe winner for yourself or your loved ones to put them in this.
25:22
Ejaaz: Josh, I want to move on to what the future is going to look like,
25:27
Ejaaz: because I don't think this story is just about autonomous driving or even like a robo taxi service.
25:34
Ejaaz: I think this is going to completely change the way we kind of interact with
25:38
Ejaaz: one another and the way we deliver items and do a whole bunch of other stuff
25:41
Ejaaz: that isn't just about sticking someone in a driver's seat and getting from point A to B.
25:45
Ejaaz: A few things that are at the top of my head is
25:49
Ejaaz: Cars are typically seen as a depreciating asset, right? As soon as you buy the
25:53
Ejaaz: car, it goes down like what, 20% in value or something ridiculous like that.
25:56
Ejaaz: With this new version of a car, it starts becoming an investment asset that earns money for you.
26:03
Ejaaz: Okay, what the hell do I mean by that? Well, think about all the hours you don't drive your car, Josh.
26:08
Ejaaz: It could be out there doing a number of different activities.
26:13
Ejaaz: Number one, it could be an autonomous Uber driver, which takes someone in your
26:17
Ejaaz: local area from A to B and you collect a revenue or a fee off of that, right?
26:21
Ejaaz: The other thing is last mile delivery is the most painful point of any kind
26:25
Ejaaz: of delivery experience, right?
26:27
Ejaaz: And so it could just simply be a thing that picks up a package and drops it off.
26:31
Ejaaz: Someone kind of like opens your door and picks it up. Now, of course,
26:34
Ejaaz: people might be initially uncomfortable with these kinds of ideas because it's
26:37
Ejaaz: like, I don't want to let my personal kind of vehicle be out there.
26:41
Ejaaz: But if it's earning you tens of thousands of dollars a month,
26:44
Ejaaz: which is like what one estimate I saw earlier on before, as we were prepping
26:48
Ejaaz: for this episode, seems to suggest, I feel like people would end up kind of
26:52
Ejaaz: utilizing something like this.
26:53
Ejaaz: And before I pass it on to you, the second thing that really excites me, Josh, is
26:57
Ejaaz: I think traffic is going to reduce, dude, because of this.
27:01
Ejaaz: Like imagine where everyone kind of subscribes to this and kind of gets in the
27:04
Ejaaz: car when they need to, to go from point A to point B.
27:08
Ejaaz: You technically don't need to park a car at any given point.
27:12
Ejaaz: You can have a fluid motion.
27:14
Ejaaz: Traffic always flows at the right speed. No one's held up in traffic trams.
27:18
Ejaaz: To your earlier point, no one gets into accidents.
27:20
Ejaaz: It just seems like a better world in general. No more parking lots in Manhattan.
27:24
Josh: Yeah. Also think about how it reshapes the actual surface of the cities that we live in today.
27:29
Josh: Like, Ijaz, if you've ever been to MetLife Stadium or any sort of stadium,
27:33
Josh: 90% of the footprint of the stadium is parking.
27:37
Josh: It's almost the entire thing. There's miles of parking lots.
27:40
Josh: And this is true with a lot of cities and even suburban areas where so much
27:43
Josh: space is dedicated to parking.
27:45
Josh: And in the case that you no longer need to park a car because there is these
27:49
Josh: autonomous cybercamps that are rolling around, it actually alters the face of
27:53
Josh: these cities. And those parking lots become either productive or parks,
27:58
Josh: and both of which are great.
27:59
Josh: You can either build more beautiful buildings or more green spaces that aren't
28:03
Josh: required as opposed to these like really dingy parking lots.
28:07
Josh: And the other thing is, to your point, Ejaz, about the appreciating asset.
28:11
Josh: There's a lot of people, I guess more of the ambitious cab drivers,
28:16
Josh: have been buying up cheaper Teslas with full-slub driving licenses in anticipation
28:21
Josh: of being able to turn their fleet into an autonomous vehicle.
28:24
Ejaaz: Operators that already ahead of the game.
28:26
Josh: A lot of these new taxi companies are driving around in Teslas and that's not an accident.
28:32
Josh: One, it's because the cost per mile is so low but two, there's a
28:35
Josh: world in which by next year they don't need to have a driving workforce anymore
28:39
Josh: and that autonomous workforce just becomes net productive without any of the
28:43
Josh: human expenses and it becomes this really interesting business model as well
28:47
Josh: where your car is actually an appreciating asset or if anything a net neutral
28:52
Josh: asset or if you want to buy a car.
28:54
Josh: You could let it go for 85% of the time that it's idle and have it actually make you money.
28:59
Josh: The last final point about this, and this is to tie back to AI,
29:03
Josh: is the distributed compute part of this story also.
29:07
Josh: Because these AI chips, we talked about this on previous episodes,
29:10
Josh: AI5, AI6 are incredibly powerful.
29:13
Josh: And these cards are sitting idle 85% of the time. So that's a lot of idle compute
29:17
Josh: that can talk to each other, that can possibly have second order effects when
29:20
Josh: it comes to training AI models or inference or whatever. So there's a lot of
29:23
Josh: open-ended possibilities. with this.
29:25
Ejaaz: For that Harry Potter fans out there, I think the dark mark has officially been
29:28
Ejaaz: released on LiDAR, Josh.
29:31
Ejaaz: Elon called it back in 2019, right? He said that, you know, LiDAR is great,
29:35
Ejaaz: but it's eventually going to be the thing that kind of kills Waymo and prevents
29:39
Ejaaz: autonomous driving companies from ever winning.
29:41
Ejaaz: Vision is the future. And, you know, after this conversation, I think he's right.
29:46
Ejaaz: And the future is just going to be so cool.
29:49
Ejaaz: Like we haven't even spoken about the other little company that Elon operates,
29:53
Ejaaz: XAI, which is going to integrate Grok and frontier level AI intelligence into the car.
29:58
Ejaaz: So right now with a Christmas update, I believe you told me,
30:01
Josh: Josh. I got it yesterday.
30:03
Ejaaz: Right. You could literally speak to your car and say, hey, can you get me from point A to B?
30:08
Ejaaz: And in the middle of that journey, you could be like, hey, can you stop off
30:10
Ejaaz: at this other gas station? I need to pick up A, B and C.
30:14
Ejaaz: It's just, it's so cool. And like my mind, I think can't fully capture the immersive
30:19
Ejaaz: process that this is going to result in.
30:22
Ejaaz: It's super cool. I'm excited to see what happens. I think it is pegged to start
30:27
Ejaaz: rolling out CyberCap, Robotaxi Service in 2026 in the summer.
30:33
Ejaaz: That's at least the news headlines that I saw, Josh. It might be delayed in a classic Elon style.
30:39
Ejaaz: But this is exciting. This technology is here right now. 500,000 people know
30:43
Ejaaz: about it. Now you guys, you listeners know about it too.
30:47
Ejaaz: Limitless, we try and strive to bring you the latest in news and insights.
30:52
Ejaaz: Josh and I spend a lot of time learning about this stuff and sometimes testing
30:55
Ejaaz: out the products themselves.
30:56
Ejaaz: Josh, eight years of experience, two years not touching his steering wheel.
30:59
Ejaaz: Come on, he hasn't got into a single accident just yet.
31:01
Ejaaz: So if you want to hear more about these types of things, turn on your notifications,
31:06
Ejaaz: subscribe to us, give us a rating.
31:08
Ejaaz: All of this helps us massively. And we will see you in the next one.