Latest / Elon Musk Thinking / Elon Musk Reveals xAI & SpaceX Masterplan!!!

Transcript

Elon Musk Reveals xAI & SpaceX Masterplan!!! · Elon Musk Thinking

0:00
Speaker 1: If you like this episode, just like, share and follow
0:03
Speaker 1: this podcast. Thank you, Back to the show, Welcome.
0:06
Speaker 2: To the Xai All Hands. We've been a very exciting
0:10
Speaker 2: presentation for you.
0:11
Speaker 1: We're going to start off by recapping the incredible progress
0:14
Speaker 1: that the Xai team has made in just two and
0:17
Speaker 1: a half years. It's really remarkable in pursuit of our
0:20
Speaker 1: goal of understanding the universe. So, just going of our
0:23
Speaker 1: accomplishments since inception, it's important for bear in mind that
0:28
Speaker 1: XI is only two and a half years old, basically
0:30
Speaker 1: a toddler, and we've nonetheless achieved an incredible amount in
0:34
Speaker 1: a very short period of time. So our competitors are
0:38
Speaker 1: five ten, some cases twenty years old, they have much
0:42
Speaker 1: larger teams they started off with for more resources, and
0:45
Speaker 1: yet nonetheless we have achieved number one in many arenas
0:50
Speaker 1: in just a few years. So we've achieved number one
0:54
Speaker 1: in voice, in image and video generation. I think we
0:57
Speaker 1: now at this point are actually generating more images and
1:00
Speaker 1: video based on the last numbers I saw then all
1:03
Speaker 1: of our competitors combined, we are winning in terms of forecasting,
1:08
Speaker 1: which is one of the key metrics of intelligence. So
1:11
Speaker 1: up the grock four to twenty forecasting model beat all
1:15
Speaker 1: the other AIS and forecasting, and we've talked many leaderboards.
1:19
Speaker 1: We've got now a great app with the Imagine, with
1:25
Speaker 1: the core Rock, we've made radical improvements to the x app,
1:28
Speaker 1: and we've launched a Grokipedia which is on its way
1:31
Speaker 1: too far exceeding Wikipedia and ultimately be Audersmatitude are more
1:36
Speaker 1: comprehensive and more accurate and have more information as well
1:40
Speaker 1: as video and image data that simply isn't there on Wikipedia.
1:45
Speaker 1: So it's it's intended ultimately to be Encyclopedia Galactica, a
1:49
Speaker 1: distallation of all knowledge, of all knowledge. And we're the
1:55
Speaker 1: first to achieve one hundred thousand, h one hundred GPU
1:59
Speaker 1: training clusters, and we're now about to achieve the first
2:03
Speaker 1: one hundred I should say, one million, h one hundred
2:06
Speaker 1: GPU equivalents in training. So really an incredible amount of
2:10
Speaker 1: work in a very short period of time. And it's
2:13
Speaker 1: important to consider for competitiveness of any technology company.
2:17
Speaker 2: What matters. It's not the position at any point in time, but.
2:20
Speaker 1: What is your velocity and acceleration And if you're moving
2:23
Speaker 1: faster than anyone else in any given technology arena, you
2:27
Speaker 1: will be the leader, and Xai is moving faster than
2:30
Speaker 1: any other company.
2:30
Speaker 2: No one's even close. So let's go to our team.
2:34
Speaker 1: As we grow as a company, a natural thing that
2:37
Speaker 1: happens is you reorganize the company as it scales up.
2:40
Speaker 2: So when you first have a startup, you.
2:42
Speaker 1: Might have just a few dozen people and they'll just
2:45
Speaker 1: chat amongs themselves. As you grow to several hundred people,
2:49
Speaker 1: you have to then add more structure, just like an
2:51
Speaker 1: organism that grows from a single like we all just
2:54
Speaker 1: grow from a single cell and then to a blob
2:56
Speaker 1: of sales. Then you get organ differentiation limbs, you grow
3:00
Speaker 1: a tail, hopefully the ptail disappears, and then you become
3:04
Speaker 1: a baby. You go through these stages, and so we're
3:07
Speaker 1: uh organizing because.
3:09
Speaker 2: We've we've reached a certain scale.
3:11
Speaker 1: We're organizing the company h to be more effective at
3:14
Speaker 1: this scale. And naturally, when when this happens, that there's
3:17
Speaker 1: some people who are better suited for the early stages
3:20
Speaker 1: of the company and and less suited for the later stages.
3:23
Speaker 1: And so uh and for the people that have departed,
3:27
Speaker 1: I'd just like to say thank you for a contribution,
3:30
Speaker 1: thank you for getting us this far. And we wish
3:32
Speaker 1: you very well in your future endeavors. So now going
3:34
Speaker 1: on to h the new structure of the company. Uh,
3:38
Speaker 1: the companies organized in four main application areas. There's there's
3:42
Speaker 1: Grock Main in voice, which is really the main Grock
3:45
Speaker 1: model as much as called Grock Main. Then there's a
3:48
Speaker 1: coding specific model. There's an image and video model which
3:52
Speaker 1: is Imagine, and then Macrohard which is intended to do
3:55
Speaker 1: full digital emulation of entire companies. And then we've got
3:59
Speaker 1: the Impress ructural ayers. So I'd like to invite members
4:02
Speaker 1: of the teams come up and talk about each of
4:03
Speaker 1: their areas.
4:09
Speaker 2: Hip. Thanks you long.
4:10
Speaker 3: So Grock, Maine and Boys are going to be merged
4:13
Speaker 3: into one team. And you know on Boys when anecdote
4:16
Speaker 3: is September twenty twenty four opening, I had this product
4:19
Speaker 3: you could talk to advanced voice mode, and we had nothing,
4:21
Speaker 3: no model, of course, in no product.
4:24
Speaker 2: We started much after.
4:25
Speaker 3: That and in a span of few months, six months,
4:27
Speaker 3: we developed the model in house from scratch without a
4:30
Speaker 3: bunch of people who knew audio, and had a product
4:32
Speaker 3: that was surpassing open air. And six months fast forward
4:35
Speaker 3: six more months and now we have Grock in more
4:37
Speaker 3: than two billion test las. We have a Rock Voice
4:40
Speaker 3: Agent API. You can do all kinds of amazing things.
4:43
Speaker 3: In a span of one year, we went from nothing
4:45
Speaker 3: to being leaders. That kind of stuff is only possible
4:47
Speaker 3: in a place like XCI. We have small teams, committed
4:49
Speaker 3: mission focus, lots of compute, and we really really want
4:53
Speaker 3: to keep pushing same story on the chat models. You know,
4:55
Speaker 3: we've always been at the forefront of reasoning, starting from
4:59
Speaker 3: Rock one point five two Doc three, and we want
5:01
Speaker 3: to really move to a world where it's no longer
5:04
Speaker 3: about this question answering. We want to build and everything
5:07
Speaker 3: up so you should be able to come to it
5:08
Speaker 3: and really get done whatever you want. You know, ask
5:10
Speaker 3: a leave question, make a slide deck, or you know,
5:14
Speaker 3: solid pizzle stuff like that.
5:15
Speaker 4: Yeah, So I really think on the product side, we're
5:18
Speaker 4: really going to see a huge transformation happening in a
5:20
Speaker 4: very short period of time.
5:22
Speaker 2: We're going to see work.
5:24
Speaker 4: The magnitude of amount of work that all knowledge workers
5:27
Speaker 4: are going to be able to produce increase tenfold in
5:30
Speaker 4: the next short period of a few months. The models
5:33
Speaker 4: that we are building out are incredibly amazing, and we
5:36
Speaker 4: have a lot on the way and we're really excited
5:38
Speaker 4: to share that with you all. And on a product side,
5:41
Speaker 4: the goal is to just build that portal that allows
5:43
Speaker 4: you to accomplish all of your work and how do
5:46
Speaker 4: we amplify everyone to achieve much much more than what
5:49
Speaker 4: they can accomplish alone. And we're building that out and
5:51
Speaker 4: it's going to be an incredibly easy.
5:53
Speaker 5: To use experience that just works seamlessly.
5:56
Speaker 2: Too.
5:57
Speaker 6: Being said, we are hiring and we're looking for intelligent
5:59
Speaker 6: and smart people.
6:00
Speaker 2: This is not an easy place to work guys like
6:02
Speaker 2: this is.
6:03
Speaker 6: It's a grind, but we have I guess like intersellar ambitions,
6:07
Speaker 6: so it's it's not going to be easy, right. So
6:10
Speaker 6: I will say, having come to x Ai, uh, it
6:13
Speaker 6: has been an opportunity for lifetime to work among really
6:16
Speaker 6: smart and really passionate people. The vibes here are amazing
6:19
Speaker 6: and it's truly an environment where if you're a smart
6:22
Speaker 6: person you want to get shipped done, you can get
6:24
Speaker 6: shipped done. There isn't like organizational overhead getting your way
6:28
Speaker 6: or kind of I don't know, like having to write
6:30
Speaker 6: docs and all this kind of stuff. You just do stuff,
6:32
Speaker 6: at least at least for me, I just you know,
6:35
Speaker 6: just you can do things here and that's amazing, and
6:37
Speaker 6: I invite more people to come here and just do
6:39
Speaker 6: it awesome things.
6:40
Speaker 1: Yeah, So with with the grockmin the sort of main
6:44
Speaker 1: foundation model, the intent is that it's it's genuinely useful
6:48
Speaker 1: in a wide range of areas.
6:49
Speaker 2: So if you're doing engineering or law or tit or.
6:53
Speaker 1: Medicine, any anything, it is useful to you in your
6:57
Speaker 1: in your job. That's essential to understand the universe and
7:03
Speaker 1: make things as useful as possible, like when God gives
7:06
Speaker 1: you an answer that you can.
7:07
Speaker 2: Count it right, all right, thank you, Thanks.
7:14
Speaker 7: Everybody. I'm Macro.
7:15
Speaker 8: So the world changed a lot recently in terms of coding,
7:19
Speaker 8: the coding models. I was always complaining people are trying
7:22
Speaker 8: to convince me to use a coding model, and I
7:24
Speaker 8: was like trusting it, and I wasn't really convinced. But
7:27
Speaker 8: as of recently, the models they actually produce good, decent
7:31
Speaker 8: quality code. I mean, you still need to review and
7:34
Speaker 8: give feedback, but you can. It's easy to see how
7:37
Speaker 8: they can accelerate you quite a lot. So it's not
7:40
Speaker 8: only about coding. It's like they understand your intuition like
7:44
Speaker 8: much better than before. Like now when you are. When
7:47
Speaker 8: I describe a problem, I only have to phrase it
7:50
Speaker 8: like I would to another colleague engineer who has already
7:52
Speaker 8: seen the code base. That's a huge change before you
7:54
Speaker 8: kind of need to handhold a toddler to make a change,
7:56
Speaker 8: and they don't only write your code, but they also
7:59
Speaker 8: can bugio code.
8:00
Speaker 7: So now we have I do like what we.
8:03
Speaker 8: Do, like hours of grog code running continuously to make
8:07
Speaker 8: sure that a more complex changed the training system actually
8:11
Speaker 8: works in production. So it's easy to see for us
8:14
Speaker 8: that this is not only about accelerating us ourselves writing
8:18
Speaker 8: code and making us ten x more productive, but we're
8:20
Speaker 8: really on this path for recursive self improvement where the
8:24
Speaker 8: current generation of grock code is training the next generation
8:27
Speaker 8: of Crock code. And we see that this path and
8:30
Speaker 8: on exponential takeoff here this path will continue. So we
8:33
Speaker 8: are doubling down on coding and making coding one of
8:36
Speaker 8: the highest party efforts in the company. So if you're
8:38
Speaker 8: out there and you're excited about coding, and you're inner
8:41
Speaker 8: either very good at training modeling, or you're a really
8:44
Speaker 8: good low level software engineer interesting in systems design, this
8:48
Speaker 8: is the place to work. Like we have a million
8:52
Speaker 8: h one hundred equivalents to train the best coding medal
8:55
Speaker 8: in the world right.
8:55
Speaker 2: Now, so please join us.
8:58
Speaker 9: Yeah, I'm going on a compare with Macro on coding,
9:01
Speaker 9: so it'll become one and more obvious to us, like
9:03
Speaker 9: you know, over time, like we are on the past
9:05
Speaker 9: to singularity at least on coding. So we decided, like
9:08
Speaker 9: you know, have our best engineer in the company in
9:11
Speaker 9: Macro to lead the coding, and we are building the
9:13
Speaker 9: best coding model for everyone to empower everyone to build
9:16
Speaker 9: and for me, like the man like limiting factor is
9:19
Speaker 9: probably computer and energy where they can run the best
9:23
Speaker 9: model to support everyone, to empower everyone and respects.
9:26
Speaker 7: Now we are one keeam and we will win on
9:29
Speaker 7: the compute and we are.
9:30
Speaker 9: Wing with space compute and also like for every engineer
9:34
Speaker 9: right so, if you're like writing kernel, if you're writing
9:37
Speaker 9: a compiler, to think about like whether is still all
9:40
Speaker 9: worth it, maybe you should join us, you know, for
9:41
Speaker 9: coding effort to automate yourself alt the like, to speed
9:44
Speaker 9: it ourselves up.
9:46
Speaker 7: Yeah, I think it's like really a million.
9:48
Speaker 9: Year basically with the year to be alive, and I
9:51
Speaker 9: can already feel the AGI, feel the hi at least
9:53
Speaker 9: for coding.
9:54
Speaker 1: Yeah, yeah, I think actually things will move, maybe even
9:57
Speaker 1: by the end of this year, to where you don't
10:00
Speaker 1: even bother do during coding. The AI just creates the
10:04
Speaker 1: binary directly, and the AI can create a much more
10:07
Speaker 1: efficient binary than can be done by any compiler. So
10:11
Speaker 1: just say create optimized binary for this particular outcome, and
10:15
Speaker 1: and you actually bypass even traditional coding. There's there's no
10:19
Speaker 1: that that's an intermediate step that actually will not be
10:22
Speaker 1: needed probably if by i'd say the end of this year.
10:26
Speaker 2: And we do expect.
10:29
Speaker 1: Code to be stead of the art in two to
10:31
Speaker 1: three months, so it's happening very quickly.
10:38
Speaker 9: Also, do Imagine, so you know, I mean, what will
10:41
Speaker 9: you do right after post aga?
10:43
Speaker 8: Right?
10:43
Speaker 9: You probably do, like digital life, So that's what we
10:45
Speaker 9: are doing here as well. We have the Imagine team
10:48
Speaker 9: like started pretty much from Squash like six months ago.
10:50
Speaker 9: We have a few people we decided will to do
10:53
Speaker 9: the image and we'll do the video gen like, yeah,
10:55
Speaker 9: look at it. What do we achieved today, Like you know,
10:57
Speaker 9: like two weeks ago. We'll recently Imagine we actually thought
11:00
Speaker 9: of a leader ball across like many of them, and
11:03
Speaker 9: people really love our product, love our model. And we
11:06
Speaker 9: have many more releases actually this month and next months.
11:09
Speaker 9: So yeah, to me, there's like a very high chance,
11:13
Speaker 9: like we actually made view the Metaverse before Mada. Yeah,
11:16
Speaker 9: I'll also pass to try to to talk Abaul, like
11:18
Speaker 9: you know, the metrics we have, the products.
11:21
Speaker 10: Yeah, yeah, Like like Godong said, it's only been six
11:24
Speaker 10: months since we started working on Imagine, we had we
11:26
Speaker 10: had no code internally for Diffusion at all six months ago,
11:30
Speaker 10: and basically now we've launched Imagine on every product surface
11:34
Speaker 10: that we have, including seamlessly integrating into X so you
11:37
Speaker 10: can open the except right now. You can long press
11:38
Speaker 10: on any image, you can edit the image, you can
11:40
Speaker 10: make a video out of the image. We also ran
11:42
Speaker 10: a contest recently where we had some really funny submissions
11:45
Speaker 10: that I'm sure many of you have seen.
11:47
Speaker 5: So Imagine it's growing extremely.
11:48
Speaker 10: Extremely fast, and it's because of the speed at which
11:51
Speaker 10: we iterate. Basically, we do multiple product updates every day,
11:54
Speaker 10: we do model updates every other week, and effectively what
11:57
Speaker 10: this has led to is now users are generating close
11:59
Speaker 10: to fifty million videos every day using Imagine.
12:02
Speaker 2: And just to.
12:02
Speaker 10: Reiterate what Elon said earlier that to the best of
12:05
Speaker 10: our knowledge that is more than every other provider combined,
12:08
Speaker 10: which again is an astonishing place to be compared to
12:10
Speaker 10: where we were six months ago. We are also generating
12:13
Speaker 10: six billion images in the last thirty days nano You know,
12:17
Speaker 10: Google recently posted that, you know, one billion images were
12:20
Speaker 10: generated using nano Vana in thirty days, So you know
12:23
Speaker 10: we're six times that, right, And really the goal is
12:26
Speaker 10: it's not like we don't just want to win. We
12:28
Speaker 10: want to win like like over a long period of
12:30
Speaker 10: time and have sustained greatness. And so the goal with
12:33
Speaker 10: Imagine is to take anything that you can you know,
12:35
Speaker 10: imagine and turn it into reality. And so that's that's
12:39
Speaker 10: what we're gonna you know that we're going to speed
12:40
Speaker 10: run that basically is the goal.
12:42
Speaker 5: Yeah, Hey, I'm haatin.
12:44
Speaker 11: We As we keep scaling our model capabilities puwing visual
12:48
Speaker 11: worlds that's indistinguishable from reality, we're also puling systems that
12:53
Speaker 11: onlocks much more possibility than what we have right now,
12:57
Speaker 11: never be able to generate the videos that's much longer
13:00
Speaker 11: and what we have right now with stories or with
13:02
Speaker 11: souls of your imagine, And by the end of the year,
13:06
Speaker 11: we likely will be having models that allow you to
13:09
Speaker 11: generate videos of ten minutes or twenty minutes in one
13:12
Speaker 11: shot without any intervention. You just need to give your
13:16
Speaker 11: imagination and our model, our agents.
13:19
Speaker 2: Will do it for you.
13:20
Speaker 11: And moreover, those are the videos we generate, and we're
13:25
Speaker 11: also going to allow rendering those. We're already the fastest
13:29
Speaker 11: in generating the videos, and we're going to keep pushing
13:32
Speaker 11: the extreme where we're going to render those videos in
13:35
Speaker 11: real time and you will be able to imagine, build
13:38
Speaker 11: and interact with your own world, and the world will
13:42
Speaker 11: respond to you in real time, and it is exciting
13:45
Speaker 11: future that we are going to build with ourself.
13:47
Speaker 1: Absolutely, my friction is that most of AI compute is
13:52
Speaker 1: going to be real time, but your understanding and real
13:54
Speaker 1: time video generation and we expect.
13:57
Speaker 2: To people leaves in that.
13:58
Speaker 1: It's really emphsizing these points that you know, six months
14:01
Speaker 1: ago we didn't even have. We had basically nothing in
14:05
Speaker 1: very weak in video and image generation and editing, and
14:09
Speaker 1: we're in six months to number one spot and in
14:12
Speaker 1: fact generally more more videos and images than everyone else combined.
14:15
Speaker 1: We're going to do the same thing with coding, and
14:16
Speaker 1: we're going to do the same thing with Macrohart and
14:20
Speaker 1: I think people will be pretty impressed with the clock
14:23
Speaker 1: four point two models coming out. That's a it's a
14:26
Speaker 1: it's a significant improvement, and that's really just that's that's
14:29
Speaker 1: the small version of our new model. So we will
14:33
Speaker 1: have a medium and a large version that are even
14:36
Speaker 1: more intelligent.
14:41
Speaker 12: All right, Hi everyone, I'm Toby and I work on Macrohart,
14:46
Speaker 12: the most series of all product names. So arguably, giving
14:50
Speaker 12: computers to humans was a good idea, So we're doing
14:53
Speaker 12: the same thing thing for AI. It's kind of like Inception.
14:56
Speaker 12: We're giving computers to computers. So Macrohart is building fully
15:00
Speaker 12: capable digital review time, very important human emulator, so it's
15:05
Speaker 12: able to do anything on a computer that a human
15:07
Speaker 12: is able to do, including using advanced tools and engineering
15:11
Speaker 12: and medicine. So there should be rocket engines fully designed
15:15
Speaker 12: by AI.
15:16
Speaker 5: And in a sense, it's one of the last few.
15:18
Speaker 12: Remaining areas where AI is significantly worse than humans, which
15:22
Speaker 12: is why I think it's one of the most exciting
15:25
Speaker 12: areas to actually innovate in and actually change change the field.
15:29
Speaker 13: Hi everyone, So yeah, my name is John and yeah,
15:32
Speaker 13: so we're building these strong reasoning models which are now
15:35
Speaker 13: going to control our CLI.
15:36
Speaker 5: Like we're actively using these every day.
15:38
Speaker 13: They are like tremendous, like productivity bursts to the whole team.
15:42
Speaker 13: I know, the voice team is like killing on that.
15:44
Speaker 13: And you know, this is the reason why we need
15:46
Speaker 13: the compute. You know, we need a large scale computer
15:48
Speaker 13: run these models to boost our own productivity.
15:51
Speaker 5: But you know, to ninety five percent of the world
15:55
Speaker 5: world software has a GUI.
15:58
Speaker 13: So that's like, you know, great representation, and you know,
16:01
Speaker 13: to truly make people's lives easier, we need to develop models.
16:04
Speaker 5: That are capable of solving day to day tasks on GUI.
16:07
Speaker 13: So Macrohart, you know, we will emulate a company where
16:11
Speaker 13: the output is digital and so this is.
16:13
Speaker 5: The obvious next step for agents.
16:15
Speaker 13: Macrohard will and the able true end to end orchestration
16:18
Speaker 13: across the desktop, and it will lead to immense economic prosperity.
16:22
Speaker 13: So yeah, we're entering in there where we need to
16:25
Speaker 13: tackle the hardest of tech problems. But in order to
16:27
Speaker 13: solve this, we need to hire the best people. So
16:30
Speaker 13: you know, think of the smartest people that you've worked
16:32
Speaker 13: with and and put them forward for a position here.
16:35
Speaker 13: And if you can't think of anybody, like, go through
16:37
Speaker 13: your phone book, go for your LinkedIn you'll be surprised
16:41
Speaker 13: like how big your actual network is. And they just
16:43
Speaker 13: need three properties obviously that we want to optimize for.
16:47
Speaker 13: Are they clever, can we solve hard problems? And the
16:50
Speaker 13: second property is are they driven? Do they have the ambition?
16:52
Speaker 13: Do they want to win? And the third is are
16:55
Speaker 13: they a nice person? Like do you want to actually
16:56
Speaker 13: work with them?
16:58
Speaker 14: Yeah?
16:58
Speaker 5: So thank you.
16:59
Speaker 2: Yeah. The Macrohart project is over time.
17:02
Speaker 15: Actually, we'll probably be the most important project because what
17:07
Speaker 15: we're talking about is emulation of entire human companies. So
17:11
Speaker 15: when you look at the most valuable companies in the world,
17:13
Speaker 15: they are.
17:15
Speaker 1: That their their output is digital, so they don't actually
17:19
Speaker 1: make hardware. So it should be possible to completely emulate
17:23
Speaker 1: any company that where.
17:24
Speaker 2: The output is digital.
17:26
Speaker 1: And this will usher in an age of prosperity lives
17:29
Speaker 1: which we can barely imagine at this point.
17:31
Speaker 2: You need imagine to imagine it.
17:33
Speaker 1: So this is a big This is a big deal,
17:35
Speaker 1: and this is why the words macrohard are painted on
17:38
Speaker 1: the roof of the training cluster, because that's what it's
17:40
Speaker 1: going to build.
17:41
Speaker 5: It's also pretty funny, meant to be a joke. It's
17:49
Speaker 5: me again.
17:50
Speaker 12: You might remember me from macrohart in computer used from
17:53
Speaker 12: a long time ago, but I also actually work on
17:56
Speaker 12: a core product infrastructure and API. In fact, this is
17:59
Speaker 12: what I've done for most time at XAI. So anytime
18:03
Speaker 12: you use any of our products like Cork dot Com
18:05
Speaker 12: API authentication, you go to status at xot Ai. This
18:09
Speaker 12: is done by the core product infront team, and a
18:12
Speaker 12: large portion of them actually sit in London and we
18:14
Speaker 12: work with himI over there, so we keep the lights
18:17
Speaker 12: on peak our four pm every day. We get paged
18:21
Speaker 12: at night when stuff goes down.
18:23
Speaker 5: Also, thank you to anyone in paleal to getting paged.
18:26
Speaker 12: This is really important work reliability, security, core product infrastructure.
18:30
Speaker 12: So if you actually if you're really interested in solving
18:33
Speaker 12: difficult distributed problems with like messy data, this is.
18:36
Speaker 5: The team to join.
18:40
Speaker 7: Hey everyone, manasiego. Yeah.
18:42
Speaker 16: So, I think one of the main bottlenecks in this
18:44
Speaker 16: next year for these models is gonna be very high
18:46
Speaker 16: quality emails and training data. And one of the ways
18:48
Speaker 16: we solve that is by taking the world's foremost experts
18:51
Speaker 16: in these domains, bringing them here and having them a
18:54
Speaker 16: value with them up. We do this for domains like medicine,
18:56
Speaker 16: finance law with voice actors. We have video editors who
19:00
Speaker 16: contribute daily to making GROW better and uh yeah, we're
19:04
Speaker 16: going to be continuing to work on very high quality
19:06
Speaker 16: emails over the.
19:07
Speaker 7: Next few months.
19:07
Speaker 16: We have some same stuff in you know, the frontier
19:10
Speaker 16: of useful tast and finance and law. You know, we're
19:14
Speaker 16: trying to build evls that are are are useful in
19:17
Speaker 16: training data that that represents useful work and not necessarily
19:20
Speaker 16: proxies of intelligence for a lot of the open source
19:23
Speaker 16: emails to today.
19:25
Speaker 7: Yeah.
19:26
Speaker 1: Yeah, I'd like to say like we're we're shifting from
19:30
Speaker 1: using these sort of common uh Internet emails, which I
19:34
Speaker 1: think are actually not a real indicator of usefulness, to
19:38
Speaker 1: having expert judos in each domain to every domain of engineering, medicine, law.
19:43
Speaker 2: Whatever the case may be. And the the.
19:46
Speaker 1: The actual evail is, does the expert in that arena
19:49
Speaker 1: or does our group of experts in that arena human
19:51
Speaker 1: experts agree that GROCK is extremely useful and that the
19:56
Speaker 1: results are correct. That's the that's actually the only EVL
19:58
Speaker 1: it really matters.
19:59
Speaker 5: Yeah, exactly.
20:01
Speaker 16: In you'll you'll see this in GROW four twenty. But
20:04
Speaker 16: we've made some improvements because of that type of data
20:05
Speaker 16: in truth seeking and kind of minimizing political bias. Our
20:08
Speaker 16: responses are are much more cogent. Yeah, that's exciting. And
20:13
Speaker 16: we are also working on Grockipedia. So the the goal
20:16
Speaker 16: of Wrockipedia is to create a distillation vol human knowledge.
20:19
Speaker 16: I kind of like to think of this as like
20:20
Speaker 16: a modern day version of the Library of Alexandria. And
20:23
Speaker 16: in the quest to build Encyclopedia Galactica to will one
20:26
Speaker 16: day be cult we've gone from essentially having nothing to
20:29
Speaker 16: around six million articles for context wikipedias around seven million
20:33
Speaker 16: English articles uh and uh yeah, or we're improving on
20:37
Speaker 16: hallucination uh. And our our goal is essentially for Rock
20:41
Speaker 16: five to not have to search out of the data center.
20:44
Speaker 7: So yep, I guess.
20:51
Speaker 8: So in the infrad team, we are building the training
20:55
Speaker 8: in France and touring team tooting.
20:57
Speaker 7: Software for the company. So to give a new example,
21:00
Speaker 7: we were training Rock three.
21:01
Speaker 8: We built the pre training framework for this and these
21:05
Speaker 8: are some of the coolest system in my opinion, that
21:07
Speaker 8: you can build as a software engineer. So it's like
21:09
Speaker 8: we have one hundred kh one hundreds at the time
21:12
Speaker 8: and they were just delivered and we.
21:14
Speaker 5: Didn't quite have the software.
21:16
Speaker 8: We thought we'd have the software, but then at thirty
21:18
Speaker 8: K scale we realized, actually the software is not quite working,
21:22
Speaker 8: and it took a major almost I was like halfway
21:26
Speaker 8: rewrite off the software because there's so much going on
21:29
Speaker 8: in a data center that you can't actually account for.
21:32
Speaker 8: Switch switches are flapping, little links are flapping, switches are
21:36
Speaker 8: going down, GPUs are just burning through. You have numerics issues,
21:40
Speaker 8: and it's a system where you want really one hundred
21:43
Speaker 8: k h one hundreds to behave in locksteps. So a
21:45
Speaker 8: training step is like five seconds and you're going five
21:48
Speaker 8: seconds in lockstep, but during that five seconds everything can happen.
21:51
Speaker 8: So you need to write a system that makes progress
21:54
Speaker 8: despite all these things that can happen in the environment.
21:56
Speaker 8: And we did a successfully in what's one of the
21:58
Speaker 8: coolest times in our life where the system was actually running,
22:01
Speaker 8: and it was running at the same time my son
22:03
Speaker 8: was born, so that was extra excitement. But these problems
22:08
Speaker 8: like you don't find anywhere else, like nobody has this
22:10
Speaker 8: kind of compute and also nobody has this kind of
22:13
Speaker 8: talent density. So at the time, to give you a perspective,
22:16
Speaker 8: we were like an overall team in pre training we
22:18
Speaker 8: were probably like fifteen people. Out of that, maybe like
22:21
Speaker 8: seven people were working on the actual training system, and
22:24
Speaker 8: we still maintain that talent density in the team.
22:28
Speaker 7: So if you're interested in working on these problems and.
22:30
Speaker 8: You don't want to be just like part of a
22:32
Speaker 8: bigger organization where you're one of like a thousand people
22:35
Speaker 8: working on this, then this is the place.
22:37
Speaker 7: Like we are still a very small.
22:38
Speaker 8: Team with me is Lea and Minn from the RL
22:40
Speaker 8: and inference team.
22:41
Speaker 2: Hi, I'm Lemming.
22:42
Speaker 17: So at our team we were on our reinforced Malanning
22:45
Speaker 17: training job and the production inference has a larger scale
22:48
Speaker 17: on the ours and probably as soon in space, and
22:51
Speaker 17: we are kind of already designed a lot of seeing
22:55
Speaker 17: to make it more resilient and scalable, So building a
22:58
Speaker 17: system to scale from hydro kate ships to millions of ships,
23:02
Speaker 17: and we aug every aspect of the stack like paradism
23:06
Speaker 17: pre fiell decault and make resilient to error in them
23:10
Speaker 17: and the unknown hardware failure. So if you are system
23:13
Speaker 17: hackers obsessed with extreme performance and reliability, so here is
23:18
Speaker 17: you will find the most interesting problems to work place.
23:21
Speaker 17: And I think actually like very similar to all kinds
23:25
Speaker 17: of things like you.
23:26
Speaker 2: It's it's very important for you to first see the.
23:29
Speaker 17: Problem and there you will developed the solutions that Noah
23:32
Speaker 17: else can develop.
23:33
Speaker 7: Before I'll handle to the tooling team. Hello, I might
23:37
Speaker 7: slip from the tooling team.
23:39
Speaker 18: Every software needs to have a great interface to be
23:42
Speaker 18: able to make it useful. So as the Tooling team
23:45
Speaker 18: were responsible for building the platform sprameworks and infrassection, which
23:48
Speaker 18: is required for humans as well as agents to be
23:52
Speaker 18: able to use our products. We started by building out
23:55
Speaker 18: the Human Data Platform. There's a place where we collect
23:57
Speaker 18: all of our human data and eventually expect, you know,
24:00
Speaker 18: to build that internal e general platform through which we
24:03
Speaker 18: basically run deployments, run evaluations or like look at like
24:08
Speaker 18: what training results exist. So if you really care about
24:11
Speaker 18: building a good interface or providing a really useful framework
24:15
Speaker 18: for researchers, for agents as well as our tutests, we
24:17
Speaker 18: should definitely join our team.
24:19
Speaker 7: So how everyone, I'm you know from the Jacks team.
24:21
Speaker 19: So now Jacks XAI is a really small team with
24:25
Speaker 19: a couple of engineers working on Jack's GPU to optimize
24:29
Speaker 19: our ultra large scale GPU training. So you can imagine
24:33
Speaker 19: that training at scale can be very complicated. Even you
24:36
Speaker 19: run hollow world at scale can be complicated. Right, So
24:39
Speaker 19: then we're actually responsible for supporting the entire companies from
24:45
Speaker 19: pre training fundation models rls and also multimodel to scale
24:50
Speaker 19: things to from from first from ten k hundred kent
24:55
Speaker 19: then probably one million, h one hundred equivalents GPO scale,
24:59
Speaker 19: and that we to implement a lot of you know,
25:02
Speaker 19: practical practical optimizations. We have to uh customize the entire
25:07
Speaker 19: jacks stack from compiler and brunt times, and there will
25:10
Speaker 19: be a lot of interesting problems.
25:13
Speaker 14: And also if you really want to, uh, you know,
25:16
Speaker 14: obsessed on optimizing uh the entire stack at scale, that
25:20
Speaker 14: we are probably the best place to go, uh because
25:23
Speaker 14: you know, we really have very large scale GIP clusters
25:26
Speaker 14: and we have a lot of interesting problems to work with.
25:28
Speaker 7: Hey, I'm branding from the Kunnels team.
25:30
Speaker 18: Basically, the Kunnel team sets at the very bottom of
25:32
Speaker 18: a training and serving stack. Our code runs inside the
25:35
Speaker 18: million current GPUs that we have and if you look
25:38
Speaker 18: inside the gp there's hundreds of thousands of threads and
25:40
Speaker 18: these threads are tend to talk to each other to
25:43
Speaker 18: multiply matrices. Computer tens scores and some of them even
25:46
Speaker 18: talk to the million other gps that we have. And
25:48
Speaker 18: this is the low level system that we have, and
25:50
Speaker 18: we like optimizing every single microsecond in this and we
25:53
Speaker 18: care deeply about squeezing every large top of performance from
25:55
Speaker 18: these gps. So if you like these low level systems, problems, algorithms,
26:00
Speaker 18: these join us.
26:06
Speaker 20: But you know, we'll try to bring in uh Heiner
26:08
Speaker 20: and Spencer who are actually at our Tubook compew cluster
26:13
Speaker 20: in Memphis.
26:18
Speaker 7: And I'm high from the computer network.
26:21
Speaker 5: Plus such a team we are maybe basic alto by
26:24
Speaker 5: the day. We're here in Memphis in the super computer.
26:26
Speaker 21: So the data center here in Memphis swum the last
26:29
Speaker 21: tribute class on the planet and it.
26:31
Speaker 3: Is still growing.
26:33
Speaker 22: Our drugs keep all this from do and running.
26:36
Speaker 21: Next Rock and the sir Ai alps us used to
26:40
Speaker 21: work well.
26:41
Speaker 2: I love to agree.
26:41
Speaker 1: It's actually just put the mic really close to your
26:45
Speaker 1: mouth because you're the ambient noises high.
26:47
Speaker 2: It's getting out that you go back.
26:49
Speaker 21: So I was saying, I'll drop the compute up and running,
26:52
Speaker 21: try the next model of ruck.
26:54
Speaker 2: And serve the so use us.
26:55
Speaker 21: So, but you used to work well and at in
26:58
Speaker 21: greens have to come together, mainly soft yeah and hotware.
27:01
Speaker 21: So there's all these tb TPUs, nicked switches and hundreds
27:05
Speaker 21: of thousands of operating systems running. Is one take supercomputer
27:08
Speaker 21: and what we need Sparks to re understand the notes,
27:11
Speaker 21: re understand that he may and re understand how computers work.
27:13
Speaker 21: Brond deep level that is, you reach album X and
27:16
Speaker 21: I'm heading over ten A.
27:17
Speaker 1: Right.
27:18
Speaker 22: So we have three hundred thousand TV, three hundred platforms
27:23
Speaker 22: us here today, still growing, still building eight hundred and
27:26
Speaker 22: forty seven miles of fiber per Data Hall twelve.
27:29
Speaker 5: Data Holls.
27:30
Speaker 22: You want to be part of the world's largest supercomputer,
27:33
Speaker 22: come join us, all right.
27:35
Speaker 23: So it's quite marvelous what we've been able to do
27:37
Speaker 23: in less than one year's time.
27:38
Speaker 2: Here we have.
27:41
Speaker 23: Once we're completely finished, we'll have north of a giggle out.
27:44
Speaker 7: Of power online and running.
27:45
Speaker 23: We'll have the largest tessil of megapack system in the
27:48
Speaker 23: world margin than Hawaii or South Australia. And Zach is
27:52
Speaker 23: really quickly going to talk a little bit about actually
27:54
Speaker 23: constructing the data is.
27:56
Speaker 24: So behind me you can see Data Hall eleven. So
27:58
Speaker 24: one of the most incredible things about what we're doing
28:01
Speaker 24: here at Macrohart's how fast we do it right, So,
28:03
Speaker 24: like they were saying before, over eight or fifty miles
28:06
Speaker 24: of fiber and every single data hall, over twenty seven
28:08
Speaker 24: thousand jews and over a two hundred thousand connections, So
28:13
Speaker 24: all of this that you can see behind me was
28:15
Speaker 24: put up in less than six weeks.
28:17
Speaker 5: We do that over and over and over again.
28:19
Speaker 24: We massively parallelize it. It's pretty much the most complex
28:23
Speaker 24: and consistent type of engineering, design and construction project you
28:28
Speaker 24: possibly imagine.
28:29
Speaker 2: So come join us. Yes, you know.
28:30
Speaker 23: The other really awesome thing about this is that everything
28:33
Speaker 23: is completely vertically integrated within this team, from architecture, mechanical, electrical, structure,
28:38
Speaker 23: all the disciplines. And we also care a lot about
28:41
Speaker 23: efficiency while we're designing all of this too, So it's
28:44
Speaker 23: not just about getting the most compute online the fastest,
28:47
Speaker 23: but also achieving the highest PUE in the industry of
28:51
Speaker 23: using as much power smoothing technology as we can. I'm
28:55
Speaker 23: being really good partners in the community here in Memphis
28:58
Speaker 23: with the toss of micropark that we have going. You
29:01
Speaker 23: can check them out. XAI Memphis back to you with.
29:05
Speaker 2: All right, thank you, all.
29:10
Speaker 1: Right, So that was live from the front lines in Memphis.
29:13
Speaker 1: So fundamental to any AI company success is the compute advantage.
29:19
Speaker 1: And what we've demonstrated over and over again is that
29:21
Speaker 1: Xai can actually deploy more AI.
29:23
Speaker 2: Compute faster than anyone else.
29:26
Speaker 1: And actually, as Justin Wong of CEO and Vidia has
29:29
Speaker 1: said many times in interviews, there is no one faster
29:32
Speaker 1: at getting AI compute online than Xai. So congratulations, guys.
29:39
Speaker 2: Yeah, this is what it looks like.
29:41
Speaker 1: So that's a really phase one, which is three hundred
29:45
Speaker 1: and thirty thousand Grace Blackwells with macrohart on the building
29:49
Speaker 1: that's on an image that it actually is on the
29:52
Speaker 1: roof of the building. And then macro Hotter will be
29:55
Speaker 1: the building that you can see which has got the
29:57
Speaker 1: macro Hotter with the rockets on it, and that'll be
29:59
Speaker 1: another turned twenty thousand GP three hundreds.
30:02
Speaker 2: So all of this will be training the models that
30:07
Speaker 2: you experience.
30:08
Speaker 1: So it's absolutely fundamental obviously to have large scale training
30:13
Speaker 1: compute in order to get the best models. Yeah, I'm
30:15
Speaker 1: sort of reminded of the Jose Amian where you see
30:18
Speaker 1: one guy digging and there's like seven people watching and
30:22
Speaker 1: One of the big differences between Xai and other companies
30:25
Speaker 1: is we are actually Jose.
30:26
Speaker 25: Hello, all right, I'm Nikita. You might know me as
30:29
Speaker 25: a part time ship poster, full time customer support for
30:32
Speaker 25: x So we're now reaching over a billion people across
30:37
Speaker 25: our family of apps. Every time news breaks, it just
30:40
Speaker 25: becomes evident that this is the most important communication tool
30:44
Speaker 25: of our time. It's where the most influential people come convene.
30:50
Speaker 25: It's where truth is crystallized. Everything is downstream of x
30:54
Speaker 25: The reason they say this is going to hit Facebook
30:56
Speaker 25: in a week because it happens here, and I think
31:00
Speaker 25: we're only beginning to realize its full potential. We had
31:04
Speaker 25: a remarkable year for the app. We rolled up our
31:06
Speaker 25: sleeves and got a ton done. January was our biggest
31:11
Speaker 25: month ever for the app in terms of engagement, and
31:15
Speaker 25: then February is on track to beat that. Much of
31:18
Speaker 25: the credit lies with the algorithm team. They've been putting
31:22
Speaker 25: in crazy hours and it's clearly paying off, but there's
31:25
Speaker 25: still a huge amount of work to be done. On
31:28
Speaker 25: the top of funnel side. First time downloads are up
31:31
Speaker 25: over fifty percent every month, and we're exhibiting right now
31:35
Speaker 25: like basically the growth rates of an early stage consumer product.
31:39
Speaker 25: We also made a ton of headway and solving one
31:42
Speaker 25: of the like twenty year old problems of the app,
31:44
Speaker 25: which was ramping up new users. New users are now
31:47
Speaker 25: spending fifty five percent more time per day in the
31:50
Speaker 25: app than they were six months ago. And on the
31:55
Speaker 25: core product side, we're hitting our stride to not only
31:59
Speaker 25: did we rebuild the algorithm, we rebuilt our onboarding flows
32:02
Speaker 25: and we're seeing double digit increases on all our key metrics.
32:05
Speaker 25: We rebuilt notifications, our web browser, U x chat, basically
32:10
Speaker 25: every surface of the app has been rebuilt to be
32:12
Speaker 25: better than ever, and it's clear that if we're focused,
32:16
Speaker 25: we can move mountains and evolve this platform.
32:19
Speaker 2: Uh.
32:19
Speaker 25: Just last month we did a little push on articles
32:24
Speaker 25: and articles published are up ten x, articles read are
32:29
Speaker 25: up seventeen x. And on all other fronts, like over
32:34
Speaker 25: the holidays, we did a big push on subscriptions.
32:37
Speaker 2: We just crossed a.
32:37
Speaker 25: Billion dollars in ar R.
32:39
Speaker 2: There. I think with the X.
32:41
Speaker 25: App, you know, the there's very few unknowns, like the
32:44
Speaker 25: path for us to win and become you know, the
32:48
Speaker 25: number one app in the world. Uh, We're it's it's
32:51
Speaker 25: we we know what to do. The balls in our court. Uh,
32:54
Speaker 25: it's it's for us to win, and it's just a
32:56
Speaker 25: matter of us executing yep.
32:58
Speaker 2: And yeah, So.
33:04
Speaker 1: We've evolved the what used to be the old Twitter
33:07
Speaker 1: DM stack, which was unencrypted basically just text, to a
33:11
Speaker 1: fully encrypted messaging system that allows you to do audio
33:16
Speaker 1: and video calls. Has uh you know, all the things
33:19
Speaker 1: you'd want from any messaging app for the disappearing messages,
33:22
Speaker 1: screen screenshot blocks, like, there's a whole all the features
33:24
Speaker 1: that you'd want want in an app. And we and
33:27
Speaker 1: we will be open sourcing the code for this in
33:30
Speaker 1: the next few months. As we are open sourcing the
33:32
Speaker 1: recommendation algorithm code so people can actually see what we're doing.
33:35
Speaker 19: Uh.
33:36
Speaker 1: Nothing beats nothing beats uh, transparency for believing in a company.
33:42
Speaker 1: So we're going to be the only recommendation algorithms that
33:47
Speaker 1: actually open sources so you can see what it, what
33:49
Speaker 1: it does, and how it's evolving. With with crock Chat,
33:52
Speaker 1: it will also be open source so you can actually
33:55
Speaker 1: see if there are any vulnerabilities. There will be no
33:57
Speaker 1: hooks for advertising or anything else like that in in
34:00
Speaker 1: grock Chat, which is really intended to be a generalized
34:03
Speaker 1: communication system, and in the next few months we'll be
34:06
Speaker 1: releasing a standalone UH x chat app, So if you
34:10
Speaker 1: just want to do messaging, you can just you can
34:12
Speaker 1: do that.
34:12
Speaker 2: You don't you don't have to go to the core product, and.
34:15
Speaker 1: We'll have desktop sharing and multi user so you can
34:20
Speaker 1: do you can do video calls with lots of people.
34:22
Speaker 1: It's really intended to be a fully functional communications system
34:27
Speaker 1: with x chat. For x money, we're UH. We've actually
34:31
Speaker 1: had x money UH live in closed beta within the company,
34:35
Speaker 1: and we expect in the next month or two UH
34:38
Speaker 1: to go to UH a limited external beata, and then
34:41
Speaker 1: to go worldwide to all x users. And this is
34:45
Speaker 1: really intended to be the place where all the money
34:48
Speaker 1: is the central source of all monetary transactions. So it's
34:52
Speaker 1: it's a it's really going to be a game changer.
34:55
Speaker 1: And the reason we say one billion user is actually
34:58
Speaker 1: over a billion users is that while our monthly users
35:01
Speaker 1: are on average around six hundred million, the number of
35:05
Speaker 1: people who have the X app and sold this well
35:08
Speaker 1: over a billion. It's just that most people only occasionally
35:11
Speaker 1: come to the X app when there's some major world event.
35:14
Speaker 1: But as we give people more reasons to use the
35:17
Speaker 1: x app, whether it's for communications, for rock, or for
35:23
Speaker 1: X money, whatever the case may be. We wanted to
35:26
Speaker 1: be such that if you wanted to, you could live
35:28
Speaker 1: your life on the X app. And as we make
35:30
Speaker 1: it more more useful, we'll obviously give people reasons, compelling
35:33
Speaker 1: reasons to use the app every day and have my
35:37
Speaker 1: expectations well over a billion daily active users now. In
35:41
Speaker 1: order to understand the universe, you must explore the universe.
35:45
Speaker 1: There's only so much you can learn from just being
35:48
Speaker 1: on Earth with telescopes and clatters on Earth. Ultimately, you
35:52
Speaker 1: have to go out there and you have to explore
35:53
Speaker 1: the universe to understand it. And that's the motivation behind
35:58
Speaker 1: the combination of space and XAI, is to accelerate humanity's
36:03
Speaker 1: future in understanding the universe and extending the light of
36:06
Speaker 1: consciousness to the stars. So, in the grand scheme of things,
36:09
Speaker 1: when you look at how much energy Earth is actually
36:12
Speaker 1: using for civilization, we're only right now using called it
36:15
Speaker 1: roughly one percent of the potential energy of Earth. And
36:19
Speaker 1: if we wanted to use even a millionth of the
36:21
Speaker 1: Sun's energy, that would be roughly a million times more
36:25
Speaker 1: energy than civilization currently uses. The only way to access
36:29
Speaker 1: that that energy, the energy of the Sun, is to
36:31
Speaker 1: extend beyond Earth. Earth is really a tiny, tiny dust
36:34
Speaker 1: mote in a vast darkness. The Sun is ninety nine
36:38
Speaker 1: point eight percent of all mass in the Solar system.
36:41
Speaker 1: So you have to expand beyond the tiny dust mote
36:44
Speaker 1: that is Earth to make any significant dent in using
36:48
Speaker 1: the Sun's energy.
36:49
Speaker 2: Like says, you'd have to expand roughly.
36:51
Speaker 1: A million times just to get to one millionth of
36:54
Speaker 1: our Sun's energy, and then going beyond that exploring, extending
36:58
Speaker 1: to the galaxy and maybe someday even to other galaxies.
37:01
Speaker 2: So the.
37:04
Speaker 1: Next step beyond Earth data centers is our Earth orbital
37:08
Speaker 1: data centers, and we'll be launching with SpaceX orbital data
37:13
Speaker 1: centers at the one hundred to two hundred gigawat per
37:15
Speaker 1: year level, not cumulative, I mean per year. And ultimately
37:20
Speaker 1: we see a path to maybe launching as much as
37:22
Speaker 1: a terror wat per year of compute from Earth. But
37:25
Speaker 1: what if you want to go beyond a mere terror
37:28
Speaker 1: wide per year. In order to do that, you have
37:30
Speaker 1: to go to the Moon. So by having factories on
37:32
Speaker 1: the Moon, building AI satellites and having a mass driver,
37:36
Speaker 1: which is the kind of thing you really need to
37:37
Speaker 1: learn about in or read about in science fiction. But
37:40
Speaker 1: we're going to make it real. We're actually going to
37:42
Speaker 1: have a mass driver on the Moon. And if you
37:45
Speaker 1: do that, you can go several orders of magnitude greater.
37:48
Speaker 1: You can go to one thousand gigawats or more per
37:51
Speaker 1: year and ultimately get to maybe a millionth and then
37:55
Speaker 1: a thousandth and maybe even.
37:57
Speaker 2: A few percent of the Sun's energy.
37:59
Speaker 1: Is simple to imagine what an intelligence of that scale
38:02
Speaker 1: would think about. But it's going to be incredibly exciting
38:04
Speaker 1: to see it happen. I really want to see the
38:07
Speaker 1: mass driver on the Moon that is shooting AI satellites
38:11
Speaker 1: into deep space. It's going to like shoo shoom, just
38:13
Speaker 1: one after the other. I can't imagine anything more epic
38:16
Speaker 1: than a mass driver on the Moon and a self
38:19
Speaker 1: sustaining city on the Moon, and then going beyond the
38:21
Speaker 1: Moon to Mars, going throughout our Solar system, and ultimately
38:26
Speaker 1: going being out there among the stars and visiting all
38:29
Speaker 1: these star systems. Maybe we'll meet aliens, uh, maybe we'll
38:32
Speaker 1: see some civilizations that lasted for millions of years, and
38:35
Speaker 1: we'll find the remnants of ancient alien civilizations. But the
38:38
Speaker 1: only way we're going to do that is if we
38:40
Speaker 1: go out there and we explore, and this is the
38:42
Speaker 1: path to making it happen.
38:43
Speaker 2: Thank you. Wow, epic end to a great presentation.
38:52
Speaker 5: Yes it's done.
38:52
Speaker 2: If that's all he wanted to say, catch you, light up.
38:54
Speaker 5: I'm going to share a few of my
38:55
Speaker 10: K type Thanks for listening, See you in the next episode.