Latest / In the Loop AI / Revealing Elon's Tesla AI Secrets: $TSLA Is Winning AI and Automation

Transcript

Revealing Elon's Tesla AI Secrets: $TSLA Is Winning AI and Automation · In the Loop AI

0:03
Josh: One of the most important technologies in the world that is happening as we
0:06
Josh: speak every day is the rise of autonomy, and particularly around autonomous robots.
0:11
Josh: Robots can be many things. Robots can be humanoids, they can be cars.
0:14
Josh: And today we're going to talk about both, because there's one company that is
0:17
Josh: at the frontier of both of those areas, and that's Tesla.
0:21
Josh: Tesla has the most unbelievable set of autopilot software that I think exists in the world.
0:27
Josh: I've been using it personally for eight years now. And it's been amazing to
0:30
Josh: see how good it's gotten.
0:31
Josh: And EJS, now there's, for the first time ever, we have the secrets.
0:35
Josh: The secret sauce that shares exactly how they've been able to get autonomy this
0:39
Josh: powerful, this impressive.
0:40
Josh: And there's now very clearly a world in which I can imagine waking up in the
0:44
Josh: morning, getting ready to go to work, stepping outside, and there's a cyber
0:47
Josh: cab waiting for me outside that will just take me wherever I want for a fraction
0:50
Josh: of the cost that it takes for a normal driver.
0:53
Josh: And I think this is an incredibly powerful unlock and to see a behind the scenes
0:57
Josh: of this is awesome so the entire episode today is behind the scenes of the most
1:02
Josh: impressive new front-end tier technology that exists.
1:05
Ejaaz: I think what i'm most excited about today josh is the fact that i've always
1:10
Ejaaz: thought tesla ai and robotics is so cool but i i just don't know how any of
1:15
Ejaaz: this works and they've refused to tell us and finally they've they've spilt their secrets today
1:20
Ejaaz: to quickly paint some context for the listeners here, up until yesterday,
1:24
Ejaaz: we only thought of Tesla AI as something called a neural network.
1:29
Ejaaz: That's their secret source. And a neural network can be thought of as a software
1:33
Ejaaz: program that is designed to function like the human brain.
1:37
Ejaaz: So it takes in information and it discovers patterns, trends,
1:41
Ejaaz: and it can also sometimes make predictions.
1:43
Ejaaz: Now, this contrasts directly to some of Tesla's competitors,
1:47
Ejaaz: which do self-driving and robotics in a very different way.
1:50
Ejaaz: They take more modular and sensor-driven approaches, right?
1:54
Ejaaz: The reason why Tesla's neural network is so special is they have an end-to-end
1:59
Ejaaz: neural network, which means that they feed a bunch of raw data from one side
2:04
Ejaaz: and out comes the output, which is an action.
2:07
Ejaaz: In this case of Tesla cars, it would be driving, steering, and acceleration.
2:12
Ejaaz: And they took this approach for a few different ways.
2:15
Ejaaz: The most important being, it's really hard Josh to codify what human values
2:21
Ejaaz: are and what I mean by that is let's say in this example that you're seeing
2:25
Ejaaz: on your screen right now you are driving your car and there's a massive puddle on your lane but
2:30
Josh: You see that you.
2:32
Ejaaz: Could potentially drive into the oncoming lane to skirt around it now for humans
2:36
Ejaaz: it's really easy to do that right it's like okay maybe I should just go through
2:39
Ejaaz: it because there's no cars coming but for a machine to do that it requires a
2:45
Ejaaz: lot of effort. It's hard to hard code.
2:47
Ejaaz: So that's one special thing around the neural network. But Josh,
2:50
Ejaaz: I want to jump into the secrets.
2:52
Ejaaz: Can you lead us with the first one?
2:54
Josh: Well, what you mentioned is really important, the end-to-end stuff.
2:57
Josh: And I want to walk through a little experiment.
2:59
Josh: So when you kick a soccer ball, I think this is an experience everyone's kind
3:02
Josh: of went through, right? What do you do when you kick a soccer ball?
3:05
Ejaaz: Yeah, I see the soccer ball coming towards me. I kind of prepare my legs ready to kind of kick.
3:12
Ejaaz: I'm right-footed, so I'm kicking with my right foot.
3:14
Ejaaz: And then I guess the rest is kind of intuitive, Josh. I just kind of run up to it and kick it.
3:19
Josh: Yeah, yeah. And I think that's exactly the point is when you kick a soccer ball,
3:23
Josh: this is something a lot of people have experienced.
3:25
Josh: You're not actually thinking about all the parts of kicking a soccer ball.
3:28
Josh: You're not thinking of where it is on the ground, where your ankle is,
3:31
Josh: where your knee is, where your leg is, the positioning, how hard you're going
3:34
Josh: to kick it. It just feels very intuitive.
3:35
Josh: And with a lot of other car companies, they're hard coding these intuitions as code.
3:40
Josh: So it does have to think about each section. It does have to calculate each
3:43
Josh: section. And what's different about Tesla and what we learned from this article,
3:46
Josh: this is from Ashok, who is the person who's in charge of Tesla AI,
3:50
Josh: is that they use this thing called end-to-end neural networks.
3:52
Josh: And what does that mean? In like a fun, simple way, it's basically the intuition
3:56
Josh: that you just described with kicking a soccer ball, the AI model,
3:59
Josh: the chip on a car is able to emulate that.
4:01
Josh: So instead of making these minute decisions all the way through a fixed decision
4:05
Josh: tree, they're able to take a ton of data and use these things that we've learned
4:08
Josh: over time, which are gradients and weights, and basically move the gradients
4:13
Josh: and weights throughout the decision process to reach an end goal.
4:16
Josh: So if the end goal is to kick a soccer ball, there's a very clear stated end goal.
4:20
Josh: And the neural network's job is to figure out the full sweep of gradients as
4:24
Josh: it goes across to get to that end goal.
4:26
Josh: And it uses a bunch of this training data that they collect in order to get there.
4:29
Josh: So this is this remarkable technology that breakthrough that they have.
4:32
Josh: And they have some really interesting examples here.
4:34
Josh: So in the case of the ducks, like we're looking at an example on the screen
4:37
Josh: right now, there's ducks standing in the middle of the road.
4:40
Josh: When you're coding an AI system, when you're coding a car, you're not hard coding
4:44
Josh: in, if you see ducks, do this.
4:46
Josh: What the car is understanding intuitively is like, okay, there's an obstacle
4:49
Josh: here and they are ducks. They're not moving.
4:52
Josh: The interesting thing is the example above is the car recognizes that the ducks
4:56
Josh: are actually moving across the road.
4:57
Josh: So it knows to wait and then it could pass once they've moved.
5:00
Josh: But the second one, it notices they're just kind of chilling.
5:03
Josh: The ducks aren't going anywhere. And what does it do? It understands that intuitively
5:06
Josh: and it is able to back up and then move around them. And that's the difference
5:10
Josh: in how Tesla does it versus some other companies is they're not hard coding
5:13
Josh: a series of fixed parameters.
5:15
Josh: They are doing it all entirely through these neural networks.
5:17
Ejaaz: If we move on to secret number one, Josh, it kind of explains how they're able
5:22
Ejaaz: to achieve this at a pretty high level, right?
5:25
Ejaaz: So it's titled The Curse of Dimensionality. And what it basically describes
5:30
Ejaaz: is you can imagine for a car to self-drive, it requires a ton of data.
5:35
Ejaaz: I think Tesla, the average car, has about seven cameras.
5:39
Ejaaz: It ingests a ton of audio data, a ton of navigation GPS data,
5:44
Ejaaz: and kinematics. So speed is tracking your speed.
5:46
Ejaaz: And so all this data is roughly equivalent to 2 billion tokens.
5:51
Ejaaz: And if you think about it, it needs to run through this end-to-end neural network
5:54
Ejaaz: that you just described, Josh, and it needs to output pretty much two tokens.
5:58
Ejaaz: One token, which determines which way the car should steer, and the other token
6:03
Ejaaz: determining how fast should that car be at that point? Should it decelerate
6:07
Ejaaz: or should it accelerate?
6:09
Ejaaz: And you can imagine this is an incredibly nuanced and complex process.
6:13
Ejaaz: And the way that the Tesla neural engine or the neural network is designed is
6:18
Ejaaz: it has really special data lanes that process this data in a very nuanced way
6:23
Ejaaz: to understand what exactly it needs to map onto when it comes to steering and acceleration.
6:28
Ejaaz: Now, you might think that's pretty cool, but Tesla's secret source when it comes
6:33
Ejaaz: to this particular component is the driving data, right, Josh?
6:37
Ejaaz: So they get access to all the camera data, audio data, GPS data that I just
6:41
Ejaaz: mentioned from their entire fleet of Tesla cars.
6:45
Ejaaz: So the equivalent of data that they get every day is something crazy like 500
6:50
Ejaaz: years worth of driving data.
6:52
Ejaaz: Now, you can imagine if it processes this amount of rich data,
6:56
Ejaaz: and not all of that data is important, right? It's kind of like the same kind of standard things.
7:00
Josh: Over those years of data.
7:02
Ejaaz: You get access to the one or two random nuanced incidents which feed in and
7:07
Ejaaz: improve the collective intelligence of the entire Tesla fleet.
7:11
Ejaaz: So whether you're on the other side of the world driving a Tesla or you're in
7:14
Ejaaz: the local neighborhood, you still benefit from the same types of improvements.
7:18
Josh: I want to talk a little bit about the scale because you mentioned 2 billion
7:20
Josh: inputs and it's kind of difficult to comprehend what 2 billion actually means.
7:25
Josh: And as a good example, I want you to imagine your phone processing every TikTok
7:31
Josh: that exists on the platform every single second in order to determine the next
7:35
Josh: turn. That is two billion inputs.
7:37
Josh: It is an astronomical amount of data. You're basically, you take the whole TikTok
7:41
Josh: catalog every second in order to make every decision and you distill that entire
7:45
Josh: data set into two single points.
7:48
Josh: And it's just, it's a remarkable amount of compression and then a remarkable
7:52
Josh: amount of precision to make the right decision over and over and over again,
7:55
Josh: and then adjust and calculate as things change.
7:58
Josh: The way that they do this, they're not doing this raw. They're not actually
8:01
Josh: ingesting all this data.
8:02
Josh: They have this data curation process that they use in order to help them kind
8:06
Josh: of figure out what is important and what is just noise.
8:09
Josh: And what they do, and we have a great example on screen here,
8:11
Josh: is they pick the juiciest clips.
8:13
Josh: It's like kind of curating like a viral playlist and they use it to train the
8:17
Josh: AI on these weird scenarios.
8:18
Josh: So we're seeing on the screen, there's someone rolling through an intersection of wheelchair.
8:22
Josh: It's actually very funny to see and scary to see what types of things happen.
8:26
Josh: I mean, this is crazy. Two cars crashing right in front of you,
8:30
Josh: driving on a snow blind street.
8:31
Josh: There's kids that are running out in the middle of the road.
8:33
Josh: There's these tremendous amount of edge cases that are really difficult to understand.
8:38
Josh: And because of the 500 years of driving data every single day that they ingest,
8:43
Josh: they're able to analyze and to kind of sift through.
8:45
Josh: And they've come up with systems to curate the most viral clips,
8:48
Josh: not viral, but the clips with the most implications of safety that are kind of the weird edge cases.
8:54
Josh: And then we have this example here. Do you want to walk through the chart that
8:57
Josh: we're seeing, because it's really fascinating how the car can kind of see it before the human does.
9:01
Ejaaz: Yeah. So what's interesting is when I first watched this clip and for those who are listening,
9:06
Ejaaz: it is a car driving on a very rainy evening on the highway and a car in front
9:12
Ejaaz: of it kind of crashes out and goes and starts to spin and kind of enter its own lane.
9:17
Ejaaz: When I first watched this video, Josh, I didn't even notice the car spinning
9:21
Ejaaz: out because it happens so far away.
9:24
Ejaaz: And so what's effective about this particular video is, given everything that
9:29
Ejaaz: you just described, the Tesla self-driving software and machinery is able to
9:33
Ejaaz: detect things that you necessarily as a human aren't able to do this.
9:39
Ejaaz: This graph specifically, Josh, can you explain what I'm looking at here?
9:43
Josh: Yeah, so this is the gradient. This is the weighted decision tree in real time.
9:46
Josh: So you could kind of see every single frame that it receives,
9:49
Josh: the chart moves, and then you could actually see the point in which it realizes
9:52
Josh: there's a threat and it adjusts very quickly.
9:54
Josh: So what you're seeing here is the real time visual representation of what the
9:57
Josh: brain sees. And we're going to get into this a little bit later where you can
10:00
Josh: actually communicate with this system.
10:02
Josh: You could talk to it just like it's a large language model. It's pretty insane.
10:05
Josh: But I want to move on to the next section because this is my favorite.
10:07
Josh: When I saw this, it just really blew my mind on how they
10:10
Josh: were able to basically emulate real world
10:13
Josh: driving scenarios and each as I want to start this section with an
10:16
Josh: example that they showed if you don't mind scrolling down and sharing the one
10:18
Josh: of the the fake screen so after these splats there's one a little bit later
10:23
Josh: and basically it shows a driving further down even sorry the like next section
10:28
Josh: then we'll go right back up oh sure sure yeah this one yeah yeah so this example
10:34
Josh: that we're looking at on the screen.
10:36
Josh: This looks like a standard traditional driving setup.
10:40
Josh: So the car has, what is that, seven cameras and each one of them ingest data.
10:44
Josh: The thing with this EJAS is what you're seeing on screen is not real.
10:47
Josh: That is a 100% virtual representation of this real world.
10:52
Josh: And it's unbelievable because it looks so good. And as I'm watching this,
10:55
Josh: I'm like, man, I hope GTA 6 looks like this because the quality,
11:00
Josh: the fidelity of this artificially generated world is indistinguishable from
11:04
Josh: real life, the entire thing.
11:06
Josh: And the reason they're able to do this is by ingesting all this data.
11:09
Josh: So now that you've seen how impressive it gets, this is kind of how they build
11:12
Josh: it. So we can go back up to the Gaussian splatting examples.
11:15
Josh: And Gaussian splats are kind of a fancy way of saying, as the car drives through,
11:19
Josh: you could imagine the cameras as scanners.
11:22
Josh: So if you flipped a camera into a scanner, it maps this 3D world and creates a world.
11:27
Josh: And then they're actually able to move around and navigate the 3D world they
11:31
Josh: create using just the cameras on your car.
11:33
Josh: And I want to reiterate that every Tesla you see on the road,
11:36
Josh: regardless of when it was made, is capable of collecting this data and creating
11:40
Josh: these 3D models that you see on the screen. So...
11:43
Josh: The interesting thing here is that top bar is what the car sees.
11:45
Josh: The bottom bar is what the car is generated to see.
11:48
Josh: And what it's able to do as a result is it's able to kind of get a better understanding
11:51
Josh: of the world around it and make much better decisions that in turn make it much
11:55
Josh: safer than a human driver does.
11:56
Ejaaz: This just looks like a computer game, Josh. Like one of those massive MMORPGs
12:01
Ejaaz: that kind of generates the world as I navigate and move through it as I interact
12:05
Ejaaz: with different characters.
12:06
Ejaaz: This is kind of that, but for self-driving specifically. And why I think this is so cool,
12:12
Ejaaz: and these are kind of like widely known as world simulators,
12:15
Ejaaz: it's like an AI model that generates simulated realities, is that this data
12:20
Ejaaz: can be modified in so many different ways and so many different scenarios to
12:25
Ejaaz: train the car for experiences or accidents that it hasn't even,
12:29
Ejaaz: that hasn't even encountered just yet.
12:32
Ejaaz: And this is really cool because I think one major constraint that a lot of AI
12:36
Ejaaz: models and self-driving models come up against is sometimes there's not enough
12:40
Ejaaz: data to account for every single different type of scenario.
12:44
Ejaaz: So a way to kind of address that is to create something known as synthetic data.
12:49
Ejaaz: World simulators is one step towards being able to do that super effectively
12:53
Ejaaz: whilst bending this simulated reality to how the actual world works,
12:59
Ejaaz: right? Physics is super important, but hard to translate into an AI model.
13:03
Ejaaz: And so seeing something like this at scale for a product, a car,
13:07
Ejaaz: that is used by almost every human on the world is just so amazing to see.
13:13
Josh: And the answer to the question, well, why hasn't everybody done this?
13:16
Josh: Is because to generate these world models generally takes tens of seconds to do.
13:19
Josh: Tesla's figured out a way to do it in 0.2 seconds. So it's a remarkable efficiency
13:24
Josh: improvement that allows them to actually do this.
13:26
Josh: It's not like the rest of the world doesn't want to do this.
13:28
Josh: Is that technically speaking, it's just very, very difficult to do.
13:31
Josh: And the next example they shared was one of my favorite ones because it really just created.
13:35
Josh: It made it feel very familiar where you can actually talk to these models like
13:39
Josh: they're a language model.
13:40
Josh: Yeah. And the example above where you could just say, well, why are you not turning left?
13:44
Josh: And it will explain to you, well, there's a detour sign. And why shouldn't you
13:48
Josh: turn right? Well, because the detour sign is pointing to the left.
13:50
Josh: And it really, you start to get a sense the same way yesterday in our episode
13:55
Josh: yesterday, where you can see the behind the scenes of how the model thinks when it trades.
13:59
Josh: You can now see the behind the scenes of the brain and you could start to understand
14:03
Josh: how it works, why it works, how it's reasoning.
14:05
Josh: And the results from this is pretty fascinating. It's not only is it interpreting
14:09
Josh: inputs like where the lines on the road are, but it's also able to read signs.
14:12
Josh: They have an example where you're able to see a human who's like kind of giving
14:16
Josh: you a high five, like saying, wait one second, I'm about to pull out.
14:18
Josh: And then the car recognizes that and stops.
14:20
Josh: So there's these like unbelievable improvements that they have.
14:22
Josh: And this section I want to get into next is because they can reevaluate these
14:27
Josh: new decision trees on existing historical models.
14:30
Josh: So my car, I've had a few near collision experiences that have been a little
14:34
Josh: scary, but they've been narrowly avoided.
14:36
Josh: What they can do is they can actually take the exact camera inputs from the
14:39
Josh: car and emulate if the collision had actually happened.
14:42
Josh: And then they could run these new tests on it and see how the new models would
14:46
Josh: compare to the old models.
14:47
Josh: So in the case that you narrowly miss an accident, well, you could test it on
14:51
Josh: a new model and see if it does better. And in the first example, it does.
14:54
Josh: And it actually moves away faster than the others.
14:57
Josh: The second example that they have here is that you can create artificial examples.
15:01
Josh: So you can take a car, remove it, place it into this virtual world,
15:05
Josh: but it looks like the real world. It emulates a real world scenario. And it just.
15:09
Josh: As I'm looking at this, Ejas, to your point, it all feels like a video game.
15:12
Josh: And it's a really high fidelity video game where they can take things from reality.
15:16
Josh: They can distort them. They could create fake realities. And as I was scrolling
15:20
Josh: through this post, I started to lose track of what was real and what wasn't
15:24
Josh: because it all looks so real to me.
15:26
Josh: And to the video game point, which you might be able to share,
15:29
Josh: is that they actually allow you to play it as if it was a video game.
15:32
Josh: You can drive through these virtual worlds without actually needing a Tesla vehicle.
15:36
Ejaaz: Yeah, so what I have here is the Tesla's Neural World Simulator,
15:41
Ejaaz: where you have someone that is in basically a driver's seat,
15:45
Ejaaz: but it's one of those video gaming driving setups.
15:47
Ejaaz: And they are driving through what looks
15:50
Ejaaz: like a pretty pleasant suburban neighborhood on a sunny blue sky day.
15:54
Ejaaz: And it looks really real, Josh. It looks like something that would be recorded
15:58
Ejaaz: from Tesla's seven cameras, except that none of it is real.
16:02
Ejaaz: He is navigating through roads. He's skirting around cars. He's narrowly avoiding collisions.
16:08
Ejaaz: And every single perspective and animal that you see from the three different
16:11
Ejaaz: cameras on the screen here is completely and utterly simulated.
16:16
Josh: The most remarkable part is that all of this amazing stuff that we've just talked
16:19
Josh: about for the last 20 minutes, it's actually cross compatible with the next
16:23
Josh: most important form of autonomy, which is robots.
16:26
Josh: Now, everyone knows Tesla's making Optimus. They signal plans to make hundreds
16:29
Josh: of thousands of these by next year.
16:31
Josh: And the problem with training robots for a lot of other companies is that they
16:34
Josh: don't have the data, they don't have the neural models.
16:36
Josh: Well, all of the progress and all of the data that's been made previously through
16:39
Josh: Tesla is cross-compatible directly with the robot team and Optimus as a humanoid robot.
16:46
Josh: And that is one of the most impressive things because as the program gets better
16:51
Josh: through AI's autopilot stack, it improves dramatically through Optimus.
16:56
Josh: And what you're able to see is,
16:58
Josh: A lot of, like you mentioned, Ejaz, the goldmine is the digital data because
17:01
Josh: you just want more data to train.
17:03
Josh: Optimus gets better. And that
17:04
Josh: moves us on to the price of Tesla and the second order effects of Tesla.
17:09
Josh: Because now that we have humanoid robots that are learning quickly,
17:13
Josh: now that we have cars that are able to drive themselves, well, there's two things.
17:17
Josh: One of them is being the chip that unifies the two.
17:20
Josh: The other is the second order effects of what happens when this gets rolled out across the world.
17:25
Josh: And he says, maybe you want to tee that up for us, because this is a very bullish
17:28
Josh: scenario that we're guiding towards.
17:30
Ejaaz: Okay, so this is the most exciting part for me for this entire episode,
17:35
Ejaaz: because as you mentioned, this data and these neural networks aren't just super
17:39
Ejaaz: valuable for the Tesla cars.
17:41
Ejaaz: It's for the robots and pretty much any other kind of robotic machine that they create in the future.
17:47
Ejaaz: And the beautiful thing about this is that it's self-recursive.
17:50
Ejaaz: So whatever is learned from all the camera information and audio information
17:53
Ejaaz: that's pulled from the cars can feed into the robots,
17:56
Ejaaz: which is like kind of what we're seeing in the demo on our screen here with
17:59
Ejaaz: this Optimus robot navigating through what seems to be a manufacturing site, right?
18:04
Ejaaz: This is incredibly bullish for Tesla, the stock, in my opinion,
18:09
Ejaaz: because it takes it from, well, it's currently breaching or sitting under its
18:13
Ejaaz: all-time high, right, Josh? What is that market cap right now?
18:16
Josh: We're just under an all-time high, which puts it right around $1.5 trillion.
18:19
Ejaaz: Okay, so $1.5 trillion in today's age seems pretty small.
18:25
Ejaaz: You just had Microsoft and Apple today cross $4 trillion market cap.
18:29
Ejaaz: If you compare that to Tesla, and if you factor in the fact that these humanoid
18:33
Ejaaz: robots are largely going to replace or work in conjunction with a large swathe
18:38
Ejaaz: of the human manual labor force,
18:41
Ejaaz: that prices this up at least up until a $10 trillion company as this scales out.
18:47
Ejaaz: Josh, I have a feeling you're probably similarly bullish when it
18:49
Josh: Comes to this. Obviously, I share your sentiment. I have been maximally bullish
18:53
Josh: on Tesla for over a decade now. It's about,
18:56
Josh: 12 years. Did your dad.
18:58
Ejaaz: Buy your Tesla stock for you at the start? You asked him to?
19:02
Josh: Yeah, I was too young to have my own brokerage account. So we were very early
19:05
Josh: shares of Tesla and continue to be maximally bullish on it.
19:09
Josh: And we're actually, I'm going to be recording a bull thesis episode about Tesla
19:12
Josh: because I'm so bullish on it. So if you're interested in that, let me know.
19:15
Josh: But I'm going to pull some notes from that to use here, just to kind of outline
19:19
Josh: the humanoid robotic opportunity.
19:20
Josh: Because EJ, as you said, $10 trillion, which is an outrageous market cap,
19:24
Josh: considering NVIDIA is the largest company in the world sitting at four trillion.
19:27
Josh: So that's a long way to go. And NVIDIA is on top of the world.
19:30
Josh: But if you think of humanoids as labor, right, you have kind of four billion
19:36
Josh: people in the labor market.
19:37
Josh: And this becomes a global trend. This is not just for the United States.
19:40
Josh: And if the average wage, which is what it is right now, is about $10,000 per
19:44
Josh: year, that's a $40 trillion market size.
19:47
Josh: So the labor opportunity is $40 trillion, assuming we don't have any productivity
19:52
Josh: unlocks that generate brand new opportunities, that generate more use cases for labor.
19:56
Josh: So that's just given the current state of the world today.
19:59
Josh: So if one humanoid at $5 an hour can replace two humans working at $25 an hour,
20:04
Josh: the value per humanoid becomes $200,000 per robot, which is pretty high given
20:10
Josh: that the costs are projected to be around $20,000 to $30,000 once it's all said and done.
20:15
Josh: The US labor market, there's 160 million people.
20:18
Josh: So if just 1% is substituted by humanoid robots, that is greater than $300 billion in value.
20:24
Josh: That's a lot of revenue. That is a tremendous amount of revenue.
20:27
Josh: And then you get to a point where you're starting to offset significant percentages of GDP.
20:31
Josh: So in the 1950s, the US manufacturing share of GDP, it was 30%.
20:35
Josh: Today it sits at 10%. And if this goes further, we'll have a total reliance on foreign entities.
20:40
Josh: So there's all the incentives in the world to bring robots into the United States.
20:44
Josh: So we don't continue this trend of decreasing our manufacturing capabilities.
20:47
Josh: There's a lot of headwinds and a lot of trends that all converge on the humanoid
20:50
Josh: robot opportunity. It's just a matter of making these.
20:53
Josh: And it's possible because of this new software stack and also because of this
20:58
Josh: new chip, which is the AI5 chip.
21:00
Josh: And the AI5 chip is the brand new golden child of Tesla. And it is going to
21:06
Josh: be cross compatible between both robots and,
21:08
Josh: cyber cabs. But you just maybe you want to walk us through exactly why this is interesting.
21:12
Ejaaz: Yeah. So the way I think about this is this is Tesla's bold attempt to replace the GPU.
21:19
Ejaaz: And as we've spoken about many times on this show before, Nvidia kind of rules the kingdom.
21:24
Ejaaz: We mentioned that they are sitting at a $4 trillion or above a well above a
21:27
Ejaaz: $4 trillion market cap. They are the kings of the roost.
21:30
Ejaaz: And the reason why is because they provide the hardware that kind of fuels all
21:34
Ejaaz: these different things.
21:35
Ejaaz: Now, what Tesla identified is whilst all these GPUs that they've been using
21:40
Ejaaz: are really helpful, they're not specifically designed to fit certain niche use
21:45
Ejaaz: cases when it comes to a range of different things that they're involved in, right?
21:49
Ejaaz: Cars, humanoid robots, and an array of different things.
21:52
Ejaaz: And now they've released their AI5 chip, which is basically their brand new
21:57
Ejaaz: chip, which is going to be used across all their different robots.
22:01
Ejaaz: So it's going to be used in cars, on humanoids, and the like.
22:04
Ejaaz: And the coolest part about this, Josh, we were speaking about this before the
22:07
Ejaaz: show, is it improves this whole GPU experience for them by 40 times.
22:13
Ejaaz: But can you help me unpack as to why exactly?
22:15
Ejaaz: Is this like a sizing thing? Can they add more compute? How does this work?
22:19
Josh: Okay, so first thing, AI5 isn't out just yet. It's coming. They have completed the spec.
22:24
Josh: Elon's been working on it. He said on the most recent earnings call that it
22:27
Josh: has been his number one focus for weeks and weeks and weeks on end,
22:30
Josh: which is very high signal that it means a lot.
22:32
Josh: So it is coming soon. They're working on tooling and they're working to roll
22:34
Josh: this out, I assume, in companion with the Optimus robot that is probably coming
22:39
Josh: next year. You mentioned it's 40 times better.
22:41
Josh: Why is it 40 times better? And why do companies make their own chips?
22:44
Josh: I think this is an important question because a lot of people don't know.
22:47
Josh: Well, NVIDIA makes awesome GPUs. Why would I go through all the R&D budgeting
22:50
Josh: costs and pain in the ass because...
22:53
Josh: To make my own chip? And the answer is because vertical integration allows you
22:57
Josh: to be hyper customized in what you're able to do.
22:59
Josh: So what Tesla has done is they, it's funny, they do this with everything,
23:03
Josh: but they kind of, they looked at the chip through first principles.
23:06
Josh: They looked at all the different modules that sit on this chip.
23:08
Josh: You could think one of them processes graphics, one of them processes images, one is processing math.
23:13
Josh: The reason why all of these GPUs from other companies need to have all of these
23:16
Josh: is because they need to satisfy their customers.
23:18
Josh: They need to be able to be diverse in the types of computing they can do.
23:22
Josh: In the narrow band of use cases that Tesla has, they're able to reconsider this and optimize for it.
23:27
Josh: So for example, there's this image signal processor that sits on a chip and
23:31
Josh: it's meant to what it says. It processes image signals that come in.
23:35
Josh: What Tesla has done is they're not actually processing images.
23:38
Josh: They're processing photons and photons can be binary.
23:40
Josh: They could be expressed in code. So there's this.
23:43
Josh: Big chip that sits on a larger chip, they're able to completely remove that
23:46
Josh: image processing chip because they said, actually, we don't need to look at images ever.
23:50
Josh: We're just doing photons in, photons out, baby. And that unlocks X percent of
23:54
Josh: this board to add more compute power to the specific type of compute you need.
23:58
Josh: So for the first time ever, you're getting these chips that don't actually look
24:03
Josh: like traditional chips.
24:04
Josh: They're built very different because of the narrow band use case that's required.
24:07
Josh: And that allows them to not only be much more efficient in terms of compute
24:11
Josh: per watt, but also cost per watt, and also the cross compatibility across all these devices.
24:15
Josh: So a lot of companies, they have, like if you think of Apple,
24:18
Josh: they have the M series chip for the computers and the iPhones,
24:22
Josh: whereas NVIDIA has 12 different GPUs for mobile devices, for power,
24:27
Josh: general computers, for data centers.
24:29
Josh: It's this really remarkable unlock that we're going to start to see roll out
24:32
Josh: next year in both that enables both the CyberCab and the humanoid robot.
24:36
Ejaaz: There's an increasing trend of these new age AI tech companies that once they
24:42
Ejaaz: reach escape velocity for a bunch of consumer and enterprise facing products,
24:47
Ejaaz: they start to vertically integrate with a part of which includes creating their
24:51
Ejaaz: own custom design GPUs and chips.
24:54
Ejaaz: The most recent example I can think of aside from Tesla is OpenAI,
24:58
Ejaaz: who announced that a partnership with Broadgate,
25:01
Ejaaz: that they're going to be developing their own custom GPUs to fuel certain niche
25:05
Ejaaz: use cases that their future GPT-6 models and ahead will utilize.
25:10
Ejaaz: They haven't quite revealed what those chips are going to be facilitating exactly.
25:16
Ejaaz: But what we do know is that they're using the AI model itself to help them design
25:21
Ejaaz: this chip. So this thing around AI5 is the most Elon thing ever,
25:25
Ejaaz: because we've seen what he's done when he's taken a hammer to data centers.
25:29
Ejaaz: And we're seeing now what he's what he's done by creating the probably the most
25:33
Ejaaz: valuable resource going forwards for tech companies at the GPU layer.
25:36
Ejaaz: So I don't know. I'm excited about this, Josh.
25:39
Ejaaz: It makes me unfathomably bullish.
25:41
Ejaaz: My earlier $10 trillion estimate is probably too conservative after what we've just discussed.
25:47
Josh: Well, with Elon's new pay package, there is a direct incentive alignment.
25:50
Josh: One thing on the Broadcom partnership with OpenAI, the difference there is that
25:54
Josh: Broadcom exists and Tesla is a single entity.
25:58
Josh: So OpenAI doesn't really have the resources in order to create their own chips in-house.
26:04
Josh: And I think that's a really big difference because when there is that physical
26:07
Josh: gap between different companies when you're designing these chips,
26:11
Josh: it makes it a little bit more difficult to do that really hardcore,
26:14
Josh: like cost-cutting vertical integration that Tesla has. Tesla's doing this.
26:18
Josh: They're making their own ship in-house. They're designing it in-house.
26:21
Josh: OpenEye is outsourcing that responsibility. And that's where you'll maybe start to see discrepancy.
26:26
Josh: So I am hopeful that they will do great, but I still suspect Tesla will do better.
26:31
Josh: And Tesla also has manufacturing prowess. So yeah, I think if we walk away with
26:35
Josh: anything from this episode is that both of us share the sentiment that we are
26:38
Josh: unfathomably bullish for an assortment of reasons. And this is just one of them.
26:41
Josh: The Tesla bookcase will be coming soon, I promise.
26:44
Josh: And there's a lot more to the company, but this is autonomy. This is autopilot.
26:49
Josh: This is the secrets of Tesla finally unveiled for the world.
26:52
Josh: And I imagine the rest of the world, granted, they've probably been trying to
26:56
Josh: emulate this. It's not really much of a secret, but we'll have a very difficult time in doing so.
27:00
Ejaaz: I think that wraps it up for today's episode.
27:03
Ejaaz: We hope you enjoyed this breakdown. We are unfathomably excited and bullish,
27:07
Ejaaz: as I've said multiple times about Tesla, but are you?
27:11
Ejaaz: Let us know in the comments. Are we crazy? is the vision that we're engaging
27:15
Ejaaz: in around Tesla completely insane?
27:17
Ejaaz: Are robots not really a thing in your opinion? Let us know in the comments.
27:21
Ejaaz: We're also going to be releasing one more episode this week,
27:24
Ejaaz: which is going to be the AI Weekly Roundup, which we're going to cover all the
27:26
Ejaaz: hottest topics. There's some crazy stuff that has happened this week.
27:30
Ejaaz: And if there's anything else that we've missed or that you want to hear about,
27:32
Ejaaz: let us know in the comments. DM us. We're always available. And we will see you in the next one.
27:38
Josh: Thanks for watching. See you guys.