Latest / Elon Musk Thinking / Latest Interview, Elon Musk and Alex Karp!!!

Transcript

Latest Interview, Elon Musk and Alex Karp!!! · Elon Musk Thinking

• 0:00
  Speaker 1: Which is a great segue. Optimists is I think going
• 0:07
  Speaker 1: to be the greatest product in the history of humanity.
• 0:14
  Speaker 1: What's the progress like and how much of your how
• 0:17
  Speaker 1: many of your cycles are going specifically to Optimists. What's
• 0:20
  Speaker 1: the timeline? I think you're on version three maybe four?
• 0:24
  Speaker 1: Tell us everything?
• 0:27
  Speaker 2: Well, yeah, everything will gonna take a long time. We've
• 0:32
  Speaker 2: got time.
• 0:38
  Speaker 3: We're finalizing the design of Optimist version three, and.
• 0:46
  Speaker 2: That really is going to be a very re multiple robot.
• 0:51
  Speaker 3: It will have the essentially the manual dexterity of a
• 0:55
  Speaker 3: human It's meaning a very complex hand the an AI
• 1:01
  Speaker 3: mind that can navigate and comprehend reality, and we made
• 1:08
  Speaker 3: in very high volume. Those are the three things that
• 1:11
  Speaker 3: I'm missing, Like if you see any other robotics company,
• 1:15
  Speaker 3: they're missing those three things. Those are the three really
• 1:19
  Speaker 3: hard things.
• 1:23
  Speaker 2: And I spend Actually, at this point.
• 1:32
  Speaker 3: It might be more of my mental cycles than anything
• 1:35
  Speaker 3: anything else, any other single thing on Optimis. That's that
• 1:40
  Speaker 3: that's solving pool real world AI all of the electro
• 1:46
  Speaker 3: mechanical issues of Optimists, the supply chain and production challenges
• 1:51
  Speaker 3: of it, because there is no supply chain that exists
• 1:54
  Speaker 3: for humanoid robots, so it has to be we have
• 1:56
  Speaker 3: to recreate it from scratch, and which requires doing a
• 2:00
  Speaker 3: lot of vertical integration. None of the actuators in Optimism
• 2:06
  Speaker 3: are available from an existing supply chain. But I think
• 2:13
  Speaker 3: it is accurate to say that if successful, Optimus will
• 2:16
  Speaker 3: be the biggest product ever and the.
• 2:20
  Speaker 1: Cost of it at scale twenty thirty forty thousand dollars
• 2:24
  Speaker 1: a robot. What do you think the first wave of
• 2:27
  Speaker 1: them will cost and when will we be able to
• 2:32
  Speaker 1: buy one to work on the ranch?
• 2:37
  Speaker 3: I think that the Martinmall cost of production once you
• 2:41
  Speaker 3: hit a million units per year is probably around the
• 2:46
  Speaker 3: twenty thousand dollars range. It so of depends on how
• 2:51
  Speaker 3: much you spend on the AI chip in the robot
• 3:00
  Speaker 3: to achieve a lot of efficiencies in the actuators. There
• 3:03
  Speaker 3: are twenty six actuators per arm, like twenty six like motors,
• 3:09
  Speaker 3: gearboxes and par electronics.
• 3:16
  Speaker 2: So, but the AI chip will be pretty expensive.
• 3:21
  Speaker 3: But like that, that might be like five, five or
• 3:23
  Speaker 3: six thousand dollars of the of the bill of materials,
• 3:26
  Speaker 3: maybe more.
• 3:29
  Speaker 2: And so but I think at volume.
• 3:33
  Speaker 3: At a million units a year, the production cost is
• 3:37
  Speaker 3: probably in the order of twenty thousand dollars maybe twenty
• 3:40
  Speaker 3: five something like that, and it price will be as
• 3:44
  Speaker 3: a function of demand.
• 3:45
  Speaker 4: Lelon, can you maybe explain to everybody why the hand
• 3:50
  Speaker 4: is so important to get right and why you know,
• 3:53
  Speaker 4: the actuator design is so unique, and you know why
• 3:56
  Speaker 4: it's so difficult, why nobody makes it, and why you
• 3:59
  Speaker 4: have to start there almost to build the rest of
• 4:01
  Speaker 4: the robot properly.
• 4:05
  Speaker 3: It turns out human hands are incredibly that they've evolved
• 4:09
  Speaker 3: to to be this incredibly sophisticated machine like that. Your
• 4:13
  Speaker 3: hand is no actually a remarkable thing. It's look look
• 4:19
  Speaker 3: closely at your hands and and think of all the
• 4:23
  Speaker 3: things you can do with your hands, which is a lot.
• 4:28
  Speaker 2: Just thinking about something. Your hands a very versatile instrument.
• 4:33
  Speaker 1: Yeah, you can give them a high five.
• 4:36
  Speaker 2: Very versatile.
• 4:39
  Speaker 3: You know, you can swing a baseball bat, you can
• 4:43
  Speaker 3: thread needles, you put thread in a needle, you can
• 4:48
  Speaker 3: play the cannon with violin. You know, you could disassemble
• 4:52
  Speaker 3: or assemble a car. The hands are incredibly versatile instruments.
• 5:00
  Speaker 3: And most of the muscles of the hand are actually
• 5:04
  Speaker 3: in the forearm, so your hand is kind of like
• 5:07
  Speaker 3: a like a like it's like a puppet like it's
• 5:10
  Speaker 3: mostly a puppet.
• 5:11
  Speaker 2: The muscle.
• 5:11
  Speaker 3: The muscles are coming from the forearm and they're pulling
• 5:13
  Speaker 3: the tendons, which are you know, also human tendon desizing
• 5:18
  Speaker 3: or human human tendent evolution is incredibly good.
• 5:23
  Speaker 2: So you've got this web of tendons. You've got I
• 5:29
  Speaker 2: think I think that.
• 5:30
  Speaker 3: Human hand is something like depending on how you counted,
• 5:32
  Speaker 3: twenty seven or twenty eight degrees of freedom per you know,
• 5:36
  Speaker 3: in the hand, it's uh, it's amazing. So in order
• 5:42
  Speaker 3: to create a robot that can be a generalized, uh, humanoid,
• 5:47
  Speaker 3: you must solve the hand hands problem.
• 5:51
  Speaker 2: And we had just got hands, knees, hands, and so is.
• 5:55
  Speaker 4: It like when you were first building Tesla where the
• 5:58
  Speaker 4: supply chain doesn't exist, So now you have to go
• 6:00
  Speaker 4: out and find folks to work with and you know,
• 6:03
  Speaker 4: build all this vertical integration, get support.
• 6:05
  Speaker 2: Is it?
• 6:06
  Speaker 4: Is it literally like it's just nowhere to be found
• 6:08
  Speaker 4: and you're going to have to build all of this
• 6:10
  Speaker 4: stuff up.
• 6:12
  Speaker 3: Yes, we could not actually buy the actuators for any
• 6:15
  Speaker 3: amount of money. They simply didn't exist, even though there
• 6:17
  Speaker 3: are ten twenty thousand electric.
• 6:20
  Speaker 2: Motors out there.
• 6:21
  Speaker 3: Versus sizes and shapes, we've had to design every electric motor,
• 6:27
  Speaker 3: gearbox and the controlling electronics from scratch, basically from physics
• 6:31
  Speaker 3: post principles.
• 6:33
  Speaker 1: The good news is you've got a lot of experience
• 6:36
  Speaker 1: with factories over the last couple of decades. So how
• 6:39
  Speaker 1: challenging is this versus cyber truck, Model YDEL, gigafactory, you
• 6:46
  Speaker 1: know the Yeah, the Faberge egg known as the Model X.
• 6:50
  Speaker 2: Yeah, right, it's harder than any any of those things.
• 6:57
  Speaker 1: Okay, yeah much hart significantly am Yes.
• 7:02
  Speaker 2: Well, there not hot starships harder.
• 7:08
  Speaker 1: Okay, So somewhere between a Model X and a starship.
• 7:13
  Speaker 5: Is it is the what's harder the hardware or the software?
• 7:19
  Speaker 3: Right now we're struggling with the the final design of
• 7:23
  Speaker 3: the hardware. Like I said, it's really primarily the hand,
• 7:27
  Speaker 3: not to just dismiss the rest of the robot. Rest
• 7:30
  Speaker 3: of it's also important, but that the hands are hands
• 7:34
  Speaker 3: and inclusive of the forum, are a majority of the
• 7:37
  Speaker 3: engineering difficulty of the entire robot.
• 7:39
  Speaker 6: And then let's let's assume you get past the hardware challenges.
• 7:42
  Speaker 6: How much do you sort of get for free based
• 7:44
  Speaker 6: on all the progress that's happening with l ll ms.
• 7:47
  Speaker 6: Will you know we'll consumers just be able to interact
• 7:51
  Speaker 6: with this. Talk to the robot, ask you to do
• 7:53
  Speaker 6: things and we'll understand.
• 7:54
  Speaker 2: And so yeah, yeah, no problem.
• 7:58
  Speaker 1: You're spending a lot of time with Annie. I know
• 8:00
  Speaker 1: noticed online not that long.
• 8:03
  Speaker 3: Maybe I went a little over the top promoting rock Imagine.
• 8:07
  Speaker 1: Well, but in all seriousness, those characters and these robots
• 8:13
  Speaker 1: that seems to be you know, like maybe.
• 8:15
  Speaker 2: They you could get the embodiments of ani as.
• 8:18
  Speaker 7: Yeah, why the human form factor? Elon, you could make
• 8:22
  Speaker 7: something that's maybe better than a human, or maybe simpler
• 8:25
  Speaker 7: than a human to do specific tasks, and maybe better
• 8:27
  Speaker 7: than a human to do more things than a human
• 8:29
  Speaker 7: can do. How do you decide to make it just
• 8:32
  Speaker 7: like a human?
• 8:35
  Speaker 3: Well, if you wanted to do all the things that
• 8:36
  Speaker 3: a human can do, it turns out you need a
• 8:38
  Speaker 3: humanoid robot. So if you wanted to do a subset,
• 8:44
  Speaker 3: that's much easier. But it turns out humans evolved to
• 8:52
  Speaker 3: the shape and capabilities that we have for good reasons.
• 9:00
  Speaker 3: That actually is that there is like this value to
• 9:04
  Speaker 3: having quiet you four fingers in the thumb, and even
• 9:09
  Speaker 3: the pinky actually is quite useful.
• 9:13
  Speaker 2: Tows are much more question walk but.
• 9:19
  Speaker 5: Well also humans.
• 9:20
  Speaker 6: Humans have designed the world as well, so we designed
• 9:23
  Speaker 6: it for us. Exactly if we could make a humanoid robot,
• 9:26
  Speaker 6: it'll be immediately backwards compatible with what we've.
• 9:28
  Speaker 5: Built the world for.
• 9:30
  Speaker 4: There's another there's another part of the robot. So there's
• 9:35
  Speaker 4: the l l MS, there's the actuation in the hands,
• 9:38
  Speaker 4: but also there's the the silicon that runs it, and
• 9:43
  Speaker 4: there was you know dojo. I think you you posted
• 9:48
  Speaker 4: on x A I five and AI six and it
• 9:50
  Speaker 4: just seemed like you were incredibly excited about the direction
• 9:54
  Speaker 4: in which the silicon layer was also going. Can you
• 9:57
  Speaker 4: tell us about that and what that is? And what what?
• 10:00
  Speaker 5: What are we?
• 10:01
  Speaker 4: What are we building here? What is being built? Is
• 10:03
  Speaker 4: it a complement to everything that exists in the world,
• 10:06
  Speaker 4: Is it a potential long term competitor?
• 10:08
  Speaker 3: What is it?
• 10:13
  Speaker 2: Yeah, So.
• 10:15
  Speaker 3: At Tesla, we basically had two different chip programs, one
• 10:20
  Speaker 3: Dojo and one UH dojo on the training side, and
• 10:23
  Speaker 3: then over call you know AI or it's just our
• 10:28
  Speaker 3: inference jib that the AI Force is currently shipping in
• 10:33
  Speaker 3: all vehicles and we're finalizing finalizing the design of AI five,
• 10:37
  Speaker 3: which will be an immense jump from AI four by
• 10:42
  Speaker 3: some metrics, the improvement in AI five will be forty
• 10:46
  Speaker 3: times better.
• 10:47
  Speaker 2: Than AI four times. And and this is because we
• 10:55
  Speaker 2: work so closely at a.
• 10:57
  Speaker 3: Very fine grade level on the AI software and the
• 11:00
  Speaker 3: AI hardware, so we know exactly where the limiting factors are,
• 11:05
  Speaker 3: and so effectively the AI hardware and software teams are
• 11:09
  Speaker 3: co designing the chip.
• 11:12
  Speaker 4: So forty x improvement in the silicon I think then,
• 11:15
  Speaker 4: as it as everybody here in the audience experience is
• 11:18
  Speaker 4: it is that just an almost like an order of
• 11:21
  Speaker 4: magnitude increase in the quality of FSD and the safety
• 11:25
  Speaker 4: that you experience as a Tesla driver, and then the
• 11:27
  Speaker 4: quality of the robot. Like word, is it all manifest
• 11:30
  Speaker 4: when you, you know, bring it up and actually get
• 11:34
  Speaker 4: it into production.
• 11:36
  Speaker 2: Yeah, it's be precise.
• 11:37
  Speaker 3: The forty x is on if you said, like compared
• 11:40
  Speaker 3: to the worst limitation on AI four, which is running
• 11:45
  Speaker 3: the softmax operation. Yeah, we currently have to run softmax
• 11:49
  Speaker 3: in around forty steps in emulation mode, whereas that will
• 11:53
  Speaker 3: just be done in a.
• 11:54
  Speaker 2: Few steps natively in AI five.
• 11:58
  Speaker 3: A five trip will also be easily handled mixed precision models,
• 12:04
  Speaker 3: so you don't have it'll dynamically handle mixed there's a
• 12:07
  Speaker 3: bunch of sort of technical stuff that AIF I will
• 12:10
  Speaker 3: do a lot better in terms of nominal sort of
• 12:15
  Speaker 3: raw compute. It's it's eight times more compute, about nine
• 12:21
  Speaker 3: times more memory, roughly five times more memory boundwidth. So
• 12:28
  Speaker 3: but because we're addressing some core limitations in AI four,
• 12:31
  Speaker 3: you multiply that by that that eight x compute improvement
• 12:35
  Speaker 3: via another five x improvement because of optimization at at
• 12:40
  Speaker 3: a very fine grain silicon level of things that are
• 12:42
  Speaker 3: currently suboptimal in.
• 12:43
  Speaker 2: AI four, that's where you get the forty x improvement
• 12:46
  Speaker 2: you had. Keep going.
• 12:49
  Speaker 3: So now that said, I am confident that the current
• 12:54
  Speaker 3: chips AI A four chips that are in the cars
• 12:57
  Speaker 3: will achieve self dry having safety that is at least
• 13:02
  Speaker 3: two to three times that of human and and maybe
• 13:06
  Speaker 3: even ten x. And the software that will be released
• 13:10
  Speaker 3: for that is coming out over the next few months,
• 13:14
  Speaker 3: so version fourteen will be the biggest upgrade until the
• 13:18
  Speaker 3: software stance version twelve. We are increasing the parameter account
• 13:24
  Speaker 3: by an order magnitude. The there's there's there's a lot
• 13:30
  Speaker 3: of reinforcement learning that's been used. As we were there,
• 13:38
  Speaker 3: There's like you can think of AI sort of as
• 13:41
  Speaker 3: a way of compressing reality, and some of those compression
• 13:45
  Speaker 3: steps we were too lossy, and we address the lostiness
• 13:51
  Speaker 3: in the compression steps. So these are all software updates
• 13:55
  Speaker 3: that will that will go out. So just over there updates.
• 13:58
  Speaker 2: Your car has got to feel like it is sentient
• 14:01
  Speaker 2: by the end of the year.
• 14:03
  Speaker 1: It feels that way already, to be honest, I saw
• 14:07
  Speaker 1: in the trades that you spent about seventeen billion dollars
• 14:10
  Speaker 1: on some spectrum and that. Yeah, so some couch change
• 14:17
  Speaker 1: to enable your satellites and the Starlink network to connect
• 14:22
  Speaker 1: directly with phones. What will that look like in a
• 14:25
  Speaker 1: year or two? Are we going to drop our Verizon
• 14:28
  Speaker 1: account and just expand our Starlink account. We're kind of
• 14:35
  Speaker 1: hoping because Verizon kind of sucks.
• 14:38
  Speaker 5: How many of you on a Starlink phone?
• 14:41
  Speaker 1: Who wants a Starlink phone? Is it technically possible?
• 14:46
  Speaker 7: I know you can't see it, but it's everyone.
• 14:48
  Speaker 2: Yeah, definitely A right cool. So this is a kind
• 14:52
  Speaker 2: of a long term thing.
• 14:55
  Speaker 3: It will allow SpaceX to deliver high band with connectivity
• 15:04
  Speaker 3: directly from the satellites to the phones, but there are
• 15:10
  Speaker 3: hardware changes that need to happen in the phone so
• 15:13
  Speaker 3: since these.
• 15:14
  Speaker 2: Frequencies aren't supported in current.
• 15:15
  Speaker 3: Phones, the chipset has to be modified to add these frequencies,
• 15:22
  Speaker 3: and that probably.
• 15:23
  Speaker 2: Is a two year time frame.
• 15:24
  Speaker 3: So the phones that are able to use the spectrum
• 15:27
  Speaker 3: that was acquired probably start shipping.
• 15:30
  Speaker 2: In around two years.
• 15:32
  Speaker 3: And then we also need to build the satellites that
• 15:36
  Speaker 3: are going to communicate on those frequencies. So in parallel
• 15:39
  Speaker 3: building the satellites and working with the handset makers to
• 15:43
  Speaker 3: add these frequencies to the phones, and then the satellites
• 15:47
  Speaker 3: and the phones will then handshake very well to achieve
• 15:51
  Speaker 3: high bound with connectivity. But the net effect is that
• 15:53
  Speaker 3: you should be able to watch videos anywhere on your phone.
• 15:58
  Speaker 2: Wow, and it's.
• 16:00
  Speaker 1: Going to be crazy.
• 16:01
  Speaker 5: What do these do? These frequencies?
• 16:04
  Speaker 6: Would they work indoors inside buildings, you know, like like
• 16:06
  Speaker 6: that phone currently does?
• 16:07
  Speaker 5: Okay, and so.
• 16:10
  Speaker 3: Will you be able to have basically like if you
• 16:12
  Speaker 3: if you if you're in a building with it with
• 16:13
  Speaker 3: it like a thick metal roof, then no, but.
• 16:16
  Speaker 5: The same the same types of normal normal homes.
• 16:20
  Speaker 4: Yes, yes, Elon, is your vision for this that instead
• 16:23
  Speaker 4: of you know, having an AT and T account or
• 16:25
  Speaker 4: and then roaming when you're in the UK or you're
• 16:28
  Speaker 4: in India. It's just we could have one direct deal
• 16:30
  Speaker 4: with starlink. It works all over the world eventually, not today,
• 16:34
  Speaker 4: but at some point. Is that the end goal that
• 16:37
  Speaker 4: basically we don't need a regional carrier, we have a
• 16:40
  Speaker 4: global carrier and that would be you.
• 16:43
  Speaker 2: Uh, that would be one of the options.
• 16:45
  Speaker 3: To be clear, We're not going to put the other
• 16:46
  Speaker 3: carriers out of business. They're still going to be around
• 16:48
  Speaker 3: because they own a lot of spectrum. So, uh, there's
• 16:52
  Speaker 3: a but. But yes, you'll you should be able to
• 16:55
  Speaker 3: have a Starlink like you have like you have an
• 16:58
  Speaker 3: A T and T or the mobile Bride and whatever
• 17:00
  Speaker 3: you should you could have a you know account with
• 17:03
  Speaker 3: stallink that works with your you know, starlink h antenna
• 17:09
  Speaker 3: at home, uh, for your Wi Fi as well as
• 17:12
  Speaker 3: on your phone. And yeah, it would be a comprehensive
• 17:16
  Speaker 3: solution for high bandwidth at home and for high end
• 17:20
  Speaker 3: with Director sell.
• 17:21
  Speaker 7: Could you buy some carriers to have more spectrum, maybe
• 17:26
  Speaker 7: five Hoverrizon.
• 17:27
  Speaker 2: Not out of the question. I suppose that may happen.
• 17:30
  Speaker 1: Let's talk about let's talk about Starship. You just had
• 17:34
  Speaker 1: a really what appeared to be a phenomenal launch. How
• 17:41
  Speaker 1: close is it to you know, being predictable and ready
• 17:45
  Speaker 1: to go in a commercial setting.
• 17:49
  Speaker 2: I think we'll recover the ship next year.
• 17:54
  Speaker 3: We've got one more launch of the Stallink version two to.
• 18:02
  Speaker 2: Stack, but there's only one one booster and ship left
• 18:06
  Speaker 2: that's in the version two design.
• 18:10
  Speaker 3: And then thereafter it's it's version three, which is a
• 18:13
  Speaker 3: gigantic upgrade because that's got Raptor three and pretty much.
• 18:17
  Speaker 2: Everything changes on the rocket with version three.
• 18:20
  Speaker 3: So version three you might have some initial tieving pains
• 18:25
  Speaker 3: because it's such a radical redesign, but it's it's capable
• 18:31
  Speaker 3: of over one hundred tons to of it fully reusable,
• 18:36
  Speaker 3: and I think it's I think I think.
• 18:39
  Speaker 2: Unless we have unless we have some very.
• 18:41
  Speaker 3: Major setbacks, SpaceX will demonstrate full reusability next year, catching
• 18:48
  Speaker 3: both the booster and the ship and being able to
• 18:51
  Speaker 3: deliver over one hundred tons to a useful orbit.
• 18:55
  Speaker 1: What is the best rocket in the world do now
• 18:58
  Speaker 1: in terms of page to space.
• 19:03
  Speaker 3: Well, in terms of uh sort of commercial rockets, this
• 19:07
  Speaker 3: this Falcon Heavy, which we'll do in with with with
• 19:15
  Speaker 3: side booster reuse, we'll do about forty tons.
• 19:20
  Speaker 1: So this is five times bigger.
• 19:21
  Speaker 3: Yeah, well two and a half times bigger. But but
• 19:26
  Speaker 3: starsha would be full reuse, full reusability, got it? Okay, Hey,
• 19:30
  Speaker 3: everything comes back.
• 19:32
  Speaker 7: Elan after after the explosion that happened with the failed launch,
• 19:39
  Speaker 7: there was a lot.
• 19:39
  Speaker 2: Of sorry which which failed, the more.
• 19:43
  Speaker 7: Recent one, the more recent with the starship hat with
• 19:45
  Speaker 7: the big boom, yeah, and the big boom, the and
• 19:48
  Speaker 7: and there's a lot of There was a lot of
• 19:51
  Speaker 7: proclamations that there's going to be environmental and f a
• 19:54
  Speaker 7: A and all these other sorts. The recovery back to
• 19:57
  Speaker 7: the launch pad again was incredible fast. How did you
• 20:02
  Speaker 7: get back so fast? Not just technically and work wise,
• 20:05
  Speaker 7: but just like regulatory clearance wise, because they said there
• 20:08
  Speaker 7: were gonna be all these questions and reviews and so on.
• 20:11
  Speaker 7: How did you guys manage that?
• 20:14
  Speaker 3: Well, there were a lot of questions and reviews. We
• 20:16
  Speaker 3: got through them all. And credit to the SpaceX team.
• 20:19
  Speaker 3: They worked incredibly hard and they got the next sugar
• 20:24
  Speaker 3: and booster tested and on the pad and plan and used.
• 20:29
  Speaker 2: Credit to the SpaceX team.
• 20:31
  Speaker 3: Very proud of them for doing doing such a job,
• 20:34
  Speaker 3: a great job recovering, I mean, creating a fully reusable
• 20:39
  Speaker 3: OpenAL rocket is one of the hottest engineering problems ever.
• 20:44
  Speaker 3: Certainly you know, a candidate for most difficult interineering project ever.
• 20:50
  Speaker 2: You know, it's on the podium at least.
• 20:54
  Speaker 3: So it's a that that's been the goal of SpaceX
• 20:58
  Speaker 3: from the beginning, from two thousand too, and here we.
• 21:01
  Speaker 2: Are twenty three years later.
• 21:03
  Speaker 3: So it's a long journey and with a super talent
• 21:09
  Speaker 3: like by far, I think the most talented group of
• 21:11
  Speaker 3: rocket engineers that's ever been assembled, and we're finally next year,
• 21:19
  Speaker 3: I think.
• 21:20
  Speaker 2: We'll be able to achieve full reusability.
• 21:23
  Speaker 4: On what are the big technical blockers that you're focused
• 21:26
  Speaker 4: on there between now and that full reusability? Are there
• 21:30
  Speaker 4: some showstoppers where you're just kind of literally just obsessing
• 21:35
  Speaker 4: over trying to figure out still or is it more
• 21:38
  Speaker 4: about getting through a side of a laundry list of
• 21:40
  Speaker 4: your learnings and just integrating it into the next launch.
• 21:46
  Speaker 3: Well, well, for full reusability of the ship, there's still
• 21:54
  Speaker 3: a lot of work that.
• 21:55
  Speaker 2: Remains on the heat shield. So no one's ever made.
• 21:58
  Speaker 3: A fully reusable opal heat shield like the Shuttle heat
• 22:01
  Speaker 3: shield had to go through nine months of repair after
• 22:04
  Speaker 3: every flight, right, so no one has ever made a.
• 22:09
  Speaker 2: Fully reusable overal heat shield and is.
• 22:11
  Speaker 4: That a material science problem or is that an engineering
• 22:15
  Speaker 4: problem or both?
• 22:16
  Speaker 3: Yeah, I mean it's a material science engineering problem. So
• 22:20
  Speaker 3: it's but we really are uh looking at the fundamental.
• 22:25
  Speaker 2: Physics here again physics post.
• 22:28
  Speaker 3: Principles and trying to figure out how do we make
• 22:30
  Speaker 3: something that is you know, it can withstand the heat,
• 22:38
  Speaker 3: is very light, doesn't transmit.
• 22:40
  Speaker 2: The heat to the primary primary structure.
• 22:47
  Speaker 1: And.
• 22:50
  Speaker 2: Integrity is holdile, stay on and they don't crack. Yeah. Uh.
• 22:56
  Speaker 3: And then as you ascend, if you hit some rain,
• 23:00
  Speaker 3: the tiles don't dissolve and rain. There's there's a lot
• 23:04
  Speaker 3: of different issues, and then you really need to know
• 23:08
  Speaker 3: that these tiles are working. You can't you know, go
• 23:12
  Speaker 3: through this laborious inspection. So it really needs to be
• 23:15
  Speaker 3: where you know, these these tens of thousands of tiles
• 23:19
  Speaker 3: all work and don't need to be repurbished or checked
• 23:25
  Speaker 3: one by one.
• 23:26
  Speaker 2: That was the case with the shuttle.
• 23:28
  Speaker 4: Can we maybe switch now, think, I mean, who who
• 23:31
  Speaker 4: else were you talked about Tesla? Then you go to
• 23:34
  Speaker 4: SpaceX Yeah, now I'd like to ask you some questions
• 23:36
  Speaker 4: about rock and x AI. You want to just give
• 23:42
  Speaker 4: us an update. I think you you kind of talked
• 23:43
  Speaker 4: about where the next gen model is and you said
• 23:45
  Speaker 4: something incredible. I still don't think people really understand it,
• 23:49
  Speaker 4: which is, you know, there's going to be a next
• 23:50
  Speaker 4: training run where you expect, you know, not to start
• 23:55
  Speaker 4: from the you know, common web and common crawl, where
• 23:59
  Speaker 4: you expected an enormous amount of synthetic data. Just tell
• 24:03
  Speaker 4: us about how the evolution of GROC is going and
• 24:06
  Speaker 4: this innovation and why it's so important.
• 24:12
  Speaker 3: Yeah, So we're running a lot of using a lot
• 24:15
  Speaker 3: of inference, compute and reasoning to look at all of
• 24:20
  Speaker 3: the source data, which is really the corpus of human knowledge,
• 24:25
  Speaker 3: and then thinking about each piece of information and then
• 24:30
  Speaker 3: adding what adding what's missing, and correcting, correcting mistakes and
• 24:37
  Speaker 3: removing falsehoods from the from that training data. So it's
• 24:41
  Speaker 3: it's if you take data Wikipedia as an example, but
• 24:44
  Speaker 3: this really applies to the books, PDFs, the websites, every
• 24:50
  Speaker 3: form of information. The GROC is using heavy amounts of
• 24:59
  Speaker 3: inference can put to say to look at, as an example,
• 25:03
  Speaker 3: a Wikipedia page and say, what is true postially true
• 25:08
  Speaker 3: or false.
• 25:09
  Speaker 2: Or missing in this page?
• 25:12
  Speaker 3: Now rewrite the page to correct the remove the false foods,
• 25:21
  Speaker 3: correct the half truths, and add the missing context.
• 25:26
  Speaker 6: Well, elong, by the way, could you just publish that?
• 25:28
  Speaker 6: Could we create like a Grockipedia?
• 25:30
  Speaker 1: I mean yeah, especially for our bio pages, which a disaster.
• 25:34
  Speaker 6: Wikipedia is so biased and it's it's a constant war.
• 25:38
  Speaker 6: You know, if something gets corrected five minutes later, there'll
• 25:41
  Speaker 6: be an army of people trying to I mean, it's
• 25:43
  Speaker 6: become hyper partisan and there's hyperperactivists.
• 25:46
  Speaker 5: All over it.
• 25:48
  Speaker 6: So if you do fix, for example, Wikipedia is a
• 25:51
  Speaker 6: source of truth, it'd be great to publish that just
• 25:54
  Speaker 6: so the world has.
• 25:55
  Speaker 3: It all right, I'll about that, so total team about that,
• 26:01
  Speaker 3: like grock Fedia or whatever.
• 26:03
  Speaker 2: Let's here's the grock Fedia.
• 26:04
  Speaker 3: Version interesting and then just have it out there.
• 26:09
  Speaker 5: Where in terms of people here like it?
• 26:13
  Speaker 4: Uh?
• 26:13
  Speaker 6: In terms of training, Grock five, you're you're scaling up
• 26:17
  Speaker 6: your supercluster in Colossus in Memphis.
• 26:23
  Speaker 5: Second one, Yeah, could you give us an update on that?
• 26:26
  Speaker 6: And then also as part of that, where are we
• 26:29
  Speaker 6: in the scaling laws? If you scale a bigger cluster,
• 26:33
  Speaker 6: do you get a more powerful AI model?
• 26:35
  Speaker 5: Is there a point of diminishing returns? Or like how
• 26:38
  Speaker 5: much more.
• 26:39
  Speaker 6: Compute if you throw twice as much compute at it,
• 26:41
  Speaker 6: do you get a ten better model?
• 26:44
  Speaker 5: Do you get one hundred percent better model?
• 26:46
  Speaker 2: Like?
• 26:47
  Speaker 5: Is it log linear?
• 26:47
  Speaker 4: What?
• 26:48
  Speaker 5: What? I guess? How much more juice is there left
• 26:51
  Speaker 5: in scaling hardware?
• 26:52
  Speaker 6: Do you think?
• 26:55
  Speaker 3: I think, I think there's a natural logarithmic function associated
• 26:59
  Speaker 3: with the amount of cute. So then like say, for auguments,
• 27:03
  Speaker 3: take like panex more compute will double the intelligence. Maybe
• 27:08
  Speaker 3: that's that might be a rough rule of thumb, but
• 27:12
  Speaker 3: you know that still means that you know, you go
• 27:13
  Speaker 3: from hundred IQ to hundred i Q still pretty pretty
• 27:17
  Speaker 3: big deal. So I and I think I think we'll
• 27:23
  Speaker 3: see intelligence continue to scale all the way up to
• 27:27
  Speaker 3: where you know, most of the power of the sun
• 27:30
  Speaker 3: is hardness for compute, and then ultimately most of the
• 27:33
  Speaker 3: power of the galaxy, you know, sort of Kardashev two
• 27:36
  Speaker 3: quartus show three scale compute. So I guess I want
• 27:42
  Speaker 3: to think about official intelligence not as sort of this.
• 27:47
  Speaker 2: You know, a destination that you reach, but.
• 27:48
  Speaker 3: Really as part of the overall escalation of intelligence.
• 27:55
  Speaker 2: That that that we are aware of.
• 27:59
  Speaker 3: You know, human intelligence is also scaled, as you have
• 28:02
  Speaker 3: as the population has increased and we've been able to
• 28:07
  Speaker 3: store more and more information. Human intelligence is scale now
• 28:10
  Speaker 3: human Because of population declines and low growth rate, human
• 28:15
  Speaker 3: intelligence is somewhat plateauing and will actually decline. And my
• 28:21
  Speaker 3: guess is that I think that we might have AI
• 28:27
  Speaker 3: smarter than any single human at anything as soon as
• 28:31
  Speaker 3: next year wow, and then probably within five, like I say,
• 28:37
  Speaker 3: twenty thirty. Probably AI is smarter than the some gold humans.
• 28:42
  Speaker 7: Do you think do you think humans are on the
• 28:44
  Speaker 7: decline because the AI is evolving? Do you think there's
• 28:48
  Speaker 7: this evolution of the ecosystem on Earth that's out our
• 28:51
  Speaker 7: way that we don't really understand the structure of what's
• 28:53
  Speaker 7: going on, but.
• 28:59
  Speaker 2: Maybe yeah, maybe we implicitly know that it's coming.
• 29:12
  Speaker 3: Yeah, I mean, I hope the birth rates soon around.
• 29:15
  Speaker 3: I'm I'm a big proponent of increased birth rate.
• 29:19
  Speaker 1: Obviously, Well you doing anything about it or no?
• 29:26
  Speaker 2: A good example.
• 29:28
  Speaker 7: You know, we had a.
• 29:29
  Speaker 1: Big conversation at this conference we didn't expect, which is
• 29:33
  Speaker 1: suicidal empathy the West. This declining birth rate. I noticed
• 29:41
  Speaker 1: you've been pretty active about it.
• 29:42
  Speaker 7: And open borders and open borders the Tucker talk. Could
• 29:46
  Speaker 7: all three of those be the same thing?
• 29:48
  Speaker 6: It seems like, there's a number of symptoms of the
• 29:51
  Speaker 6: West being suicidal, the most obvious one being the birth
• 29:54
  Speaker 6: rate is not a replacement level. So obviously if that
• 29:56
  Speaker 6: continues and definitely then the West will literally not.
• 30:00
  Speaker 5: Reproduce enough to replace itself. But there's other things too.
• 30:03
  Speaker 6: There's the fact that the borders were totally opened to
• 30:07
  Speaker 6: the point where Western culture, the social fabrics started to
• 30:10
  Speaker 6: come apart. And you see this especially in Europe where
• 30:13
  Speaker 6: they're you know, the indigenous cultures of the UK or
• 30:17
  Speaker 6: France or Germany or starting to potentially be taken over
• 30:21
  Speaker 6: by by cultures of people who are brought in and
• 30:23
  Speaker 6: aren't assimilating. You have crime where you know, we have
• 30:27
  Speaker 6: this case on social media right now, this young woman,
• 30:30
  Speaker 6: Irena who's just killed in a senseless way on a subway,
• 30:35
  Speaker 6: which is horrific enough in and of itself, but then
• 30:38
  Speaker 6: in addition to that, the elite media just for whatever
• 30:40
  Speaker 6: reason just refused to cover it like it didn't exist.
• 30:44
  Speaker 6: So you have this issue of crime that's not being
• 30:46
  Speaker 6: addressed or even no knowledge.
• 30:48
  Speaker 1: And no acknowledgment of this, Like it's almost like we're
• 30:51
  Speaker 1: trying to deny the reality of the spiral and this.
• 30:56
  Speaker 6: Yeah, so you have the you have all these data
• 30:59
  Speaker 6: points that seemed to suggest that the West is suicidal
• 31:05
  Speaker 6: or doesn't you know, doesn't seem to want to defend
• 31:08
  Speaker 6: itself or propagated itself. Look, I think everyone in this
• 31:11
  Speaker 6: room thinks that life is awesome, right, I mean it's
• 31:14
  Speaker 6: pretty great.
• 31:17
  Speaker 5: Yeah.
• 31:18
  Speaker 6: And when when Alex Karp was here earlier today defending
• 31:21
  Speaker 6: the West, that got some of the loudest applause at
• 31:23
  Speaker 6: the conference. So, uh, I guess we probably don't really
• 31:26
  Speaker 6: understand what's going on.
• 31:27
  Speaker 2: We don't really.
• 31:28
  Speaker 7: Yeah, what's your take elon because you know, what's your
• 31:31
  Speaker 7: take on the suicide of the West?
• 31:32
  Speaker 1: Yeah, what's what's.
• 31:34
  Speaker 2: I'm very worried about it. Yeah, I'm very worried about it.
• 31:39
  Speaker 2: You know. I think there's the.
• 31:41
  Speaker 3: Let's just say that the actions of the West are
• 31:43
  Speaker 3: indistinguishable from suicide.
• 31:45
  Speaker 2: It's so, but it's.
• 31:50
  Speaker 3: Look, at least in America, there's there's there's generally a
• 31:54
  Speaker 3: sense of optimism. But when's the last time you told
• 31:58
  Speaker 3: someone from Europe who lives in Europe who's optimistic?
• 32:03
  Speaker 2: Not for a while, no decades, like even one. It's rare.
• 32:09
  Speaker 3: So I think unless people have a sense of optimism
• 32:13
  Speaker 3: and purpose about the future.
• 32:15
  Speaker 2: They so suicide might be just what.
• 32:19
  Speaker 3: Happens, like like like having a child is an act
• 32:23
  Speaker 3: of optimism about the future.
• 32:25
  Speaker 2: So uh, if you're not optimistic.
• 32:30
  Speaker 3: Yeah, So I think we need to maybe give people
• 32:37
  Speaker 3: a sense of optimism and excitement about the future. And
• 32:41
  Speaker 3: I believe that the future will be better than the
• 32:42
  Speaker 3: past and they'll be more interested in having kids.
• 32:46
  Speaker 7: Did religion play a role in the past? Elanta? Yeah,
• 32:50
  Speaker 7: kind of play Kate and make folks feel that way.
• 32:53
  Speaker 2: Yeah, I think so.
• 32:57
  Speaker 3: Nature abhors a vacuum, and if you take away religion,
• 33:01
  Speaker 3: then I think you actually you you get something in
• 33:05
  Speaker 3: this place which is actually worse than what was there before.
• 33:10
  Speaker 3: I mean it's like destructive basically, you get you get
• 33:12
  Speaker 3: like the white work mind virus pulling, filling the whole
• 33:15
  Speaker 3: that religion used to have, taking the place.
• 33:19
  Speaker 2: Of of of religion. You get these dystopian, de facto
• 33:23
  Speaker 2: religions that that are that are very very self destructive.
• 33:31
  Speaker 3: So I think perhaps some some sort of revival of religion,
• 33:37
  Speaker 3: or at least what we need is is some coherent
• 33:42
  Speaker 3: philosophy that people can get excited about, you know, I mean,
• 33:47
  Speaker 3: for me, it's a philosophy of curiosity. I'm curious about
• 33:50
  Speaker 3: the nature of universe, and I want to go out there,
• 33:53
  Speaker 3: and I want humanity to be out there exploring the stars.
• 33:58
  Speaker 2: That maybe meeting alien civilizations.
• 34:01
  Speaker 3: Uh maybe in some cases we see the ruins of
• 34:04
  Speaker 3: elongate alien civilization, but they were they were very strong
• 34:07
  Speaker 3: for ten million years, you know, the kind of stuff
• 34:10
  Speaker 3: that you see in Star Trek, in a non dystopian
• 34:13
  Speaker 3: and sci fi book or or movie or show. And
• 34:17
  Speaker 3: so I'm just to have a I have a philosophy
• 34:19
  Speaker 3: of curiosity of like, I just want to know what's
• 34:21
  Speaker 3: going on. And in order to know what's going on,
• 34:24
  Speaker 3: we must have and an increase in the in the
• 34:28
  Speaker 3: scope and scale of consciousness. We must we must expand
• 34:32
  Speaker 3: as consciousness, we must grow, we must grow humanity, and
• 34:36
  Speaker 3: we must extend humanity in order to comprehend and to
• 34:43
  Speaker 3: understand the universe.
• 34:44
  Speaker 2: Even what what.
• 34:44
  Speaker 3: Question should we should ask about? The answer that is
• 34:48
  Speaker 3: the universe?
• 34:50
  Speaker 2: You know.
• 34:50
  Speaker 3: Doug Douglas adams book The Hitchhrackers Guys in the Galaxy
• 34:54
  Speaker 3: is actually a deep book on philosophy disguised as humor.
• 34:59
  Speaker 3: And what the point he was trying to make in
• 35:00
  Speaker 3: that book was that the questions are the really the
• 35:05
  Speaker 3: hard part. The answer is the universe, like the answer
• 35:09
  Speaker 3: is everything you see around you. But but but what
• 35:11
  Speaker 3: are the questions that we.
• 35:12
  Speaker 2: Don't know to ask?
• 35:13
  Speaker 1: Ye?
• 35:15
  Speaker 3: Now some of the questions I guess I do know.
• 35:17
  Speaker 3: I'd like to know is the standard model of physics correct?
• 35:20
  Speaker 3: About the origins of the universe? Are we actually thirty
• 35:22
  Speaker 3: twenty ellion years old? How does the universe end? Does
• 35:25
  Speaker 3: end an heat death or in some other way, you know,
• 35:29
  Speaker 3: a black hole.
• 35:32
  Speaker 2: We might be Can you talk about the whole sort
• 35:37
  Speaker 2: of simulation question? Are we assimulation? Maybe?
• 35:39
  Speaker 1: Where does the Where do you think we find the
• 35:42
  Speaker 1: answer first in AI or in the stars? Because you're
• 35:47
  Speaker 1: pursuing both obviously.
• 35:51
  Speaker 3: Yeah, I don't know if I hope more people can
• 35:57
  Speaker 3: get behind a philosophy of curiosity because I think it's
• 36:05
  Speaker 3: very exciting. Yeah, and and and and inherently optimistic, like
• 36:13
  Speaker 3: because there's there's this amazing, amazing sense.
• 36:16
  Speaker 2: Of wonder about the nature of the universe.
• 36:20
  Speaker 3: And when you just when you uncover some secrets to
• 36:22
  Speaker 3: the universe, that's amazing, and you're like a whole world
• 36:27
  Speaker 3: of understanding is opened up. I mean we used to
• 36:31
  Speaker 3: not even know where all the continents were. You know,
• 36:35
  Speaker 3: used to be like just the map would be, there'd
• 36:37
  Speaker 3: be dragons, and like all we know is that when
• 36:39
  Speaker 3: they sailed in that direction, they didn't come back. I
• 36:44
  Speaker 3: mean the moon base, that's that's all they knew.
• 36:47
  Speaker 1: I kind of feel like the moon base, or just
• 36:50
  Speaker 1: going to the Moon for real this time would be
• 36:52
  Speaker 1: a big step in the right direction. You still have
• 36:55
  Speaker 1: the Moon planned, what's the status of that? Is that
• 36:58
  Speaker 1: still on the agenda?
• 36:59
  Speaker 3: Yea, yea, I think having I think we want to
• 37:02
  Speaker 3: try to reach new heights as a civilization, so I
• 37:07
  Speaker 3: think that's it's fine to go to the Moon, but
• 37:09
  Speaker 3: we should go to the Moon in order to establish
• 37:11
  Speaker 3: a lunar base, like a lunar research base. I mean,
• 37:16
  Speaker 3: there are parts of the Moon that are perhaps older
• 37:19
  Speaker 3: than parts of Earth, and we might understand more about
• 37:23
  Speaker 3: the nature of the universe if we had a science
• 37:25
  Speaker 3: base on the Moon, that would be very cool. And
• 37:28
  Speaker 3: then we obviously want to go beyond the Moon to Mars,
• 37:32
  Speaker 3: and well they self sustaining city on Mars. I do
• 37:37
  Speaker 3: think that that there is a folk in the road
• 37:41
  Speaker 3: of human destiny where if we can establish a self
• 37:44
  Speaker 3: sustaining city on Mars, with the key test being if
• 37:49
  Speaker 3: the resupply ships from Earth stopped coming for any reason,
• 37:53
  Speaker 3: does Mars continue to prosper or.
• 37:55
  Speaker 2: Does it die out.
• 37:56
  Speaker 3: At the point at which Mars is able to asper
• 38:00
  Speaker 3: and grow on its own, the probable lifespan of consciousness
• 38:04
  Speaker 3: is dramatically greater because we are no longer dependent on
• 38:08
  Speaker 3: everything going right on Earth. You know, there's there's always
• 38:12
  Speaker 3: some possibility of self annihilation on Earth with the World
• 38:14
  Speaker 3: War three, or or supervirus or or or a media
• 38:19
  Speaker 3: or like extinct. But you know that destroyed the dinosaurs.
• 38:22
  Speaker 3: We know from the fossil record that there have been
• 38:24
  Speaker 3: many massive mass extinction events. So the question that I
• 38:29
  Speaker 3: sort of always wondering about is will civilization can Will
• 38:33
  Speaker 3: the civilizational arc continue to ascend such that we can
• 38:37
  Speaker 3: make Mars self sustaining before the civil civilizational arc descends,
• 38:44
  Speaker 3: because the window of opportunity to make life faulty planetary
• 38:49
  Speaker 3: exists now for the first time in the four and
• 38:51
  Speaker 3: a half billion year history of Earth.
• 38:53
  Speaker 4: Yeah, you'll also assume that we get there and you're there.
• 39:00
  Speaker 4: You know, you would be the elder statesman, you'd have
• 39:02
  Speaker 4: the moral authority of Mars. How do you run Mars?
• 39:08
  Speaker 3: H But there's this point that I think I want
• 39:14
  Speaker 3: to just emphasize again, that's that's that's more important than
• 39:18
  Speaker 3: the form of governance on Mars or who's there in
• 39:21
  Speaker 3: the early days. What really matters is that Mars is
• 39:28
  Speaker 3: self sustaining, that we are truly a multi planet species
• 39:32
  Speaker 3: and such such that we've achieved planetary redundancy, so that
• 39:36
  Speaker 3: if if something, obviously we should do everything possible to
• 39:40
  Speaker 3: make sure life on Earth is great. But there's always
• 39:42
  Speaker 3: some risk that of an annihilation event on Earth, like
• 39:46
  Speaker 3: I said, self annihilation or some natural disaster and UH,
• 39:52
  Speaker 3: and so the probable lifespan of consciousness increases dramatically as
• 39:57
  Speaker 3: soon as UH, as soon as we are a multiplanet species,
• 40:01
  Speaker 3: with the key test being can Mars survive if the
• 40:04
  Speaker 3: resupply ships stop coming. So it's getting like the first
• 40:07
  Speaker 3: missions to Mars are not that important. What matters is
• 40:11
  Speaker 3: can you get sufficient tonnage to Mars such that Mars
• 40:15
  Speaker 3: can prosper on its own, And that means it has
• 40:18
  Speaker 3: to have all of the ingredients of civilization. It's not
• 40:21
  Speaker 3: just that you need to build, for example, a chip factory,
• 40:24
  Speaker 3: on Mars or ship fab on Mars, but you need
• 40:27
  Speaker 3: the ability ability to apps.
• 40:29
  Speaker 4: Do you have a sense of the time scale, Like,
• 40:31
  Speaker 4: let's assume Starship is at a state starting in you know,
• 40:35
  Speaker 4: twenty twenty six, then there's going to be a bunch
• 40:37
  Speaker 4: of testing. Obviously there's going to be a bunch of
• 40:38
  Speaker 4: early testing. We only have certain launch windows, so there's
• 40:42
  Speaker 4: a bunch of time constraints. Is this a fifty year
• 40:46
  Speaker 4: thing in your mind? Is it one hundred and fifty
• 40:47
  Speaker 4: year thing? Is it something that is for our generation
• 40:50
  Speaker 4: or is it our children's generation? Where do you see
• 40:53
  Speaker 4: that point? If it's optimally possible, you know, things go
• 40:57
  Speaker 4: and break our way.
• 41:00
  Speaker 2: I think it can be done in thirty years.
• 41:05
  Speaker 3: If it provided there's an exponential increase in the in
• 41:08
  Speaker 3: the tonnage to Mars with each success of MAAS transfer windows,
• 41:12
  Speaker 3: which is every two years, So every two years the
• 41:16
  Speaker 3: planets aligned and you can you can transfer to Mars.
• 41:19
  Speaker 3: So I think in roughly fifteen, but maybe as few
• 41:26
  Speaker 3: as ten, but I saw ten to fifteen ish mass
• 41:30
  Speaker 3: transfer windows. If you're seeing exponential increases in the tonnage
• 41:37
  Speaker 3: to Mars with each mass transfer window, then it should
• 41:40
  Speaker 3: be possible to make more self sustaining in about quote
• 41:44
  Speaker 3: roughly twenty five years.
• 41:46
  Speaker 1: Amazing.
• 41:46
  Speaker 2: That's incredible.
• 41:48
  Speaker 1: All right, ladies and gentlemen, Elon Musk, we'll see you
• 41:52
  Speaker 1: when we're back in town. We miss you, see you
• 41:55
  Speaker 1: in person, that's time. Thank you, brother,