← This Week in AI
This Week in AI
Data Centers in Space, AI Excavators & Fixing AI Slop | Philip Johnston, Boris Sofman, Spiros Xanthos
01:10:02
Loading summary
Transcript152 lines
- [00:00]
A
Sam doesn't have access to space. And if he wanted access to space, he would have to bend the knee to Elon to get access.
- [00:06]
B
He very carefully said, this decade and not in the next decade, how many
- [00:10]
A
satellites are going to be out there and how do you account for all the collisions and possibilities?
- [00:15]
C
Oh, my gosh.
- [00:15]
B
We've just filed for a constellation of 88,000 with the FCC.
- [00:18]
A
You said 88,000?
- [00:20]
B
88,000. Elon's just filed for a constellation of a million.
- [00:23]
C
This Week in AI is brought to you by Notion. Bring all your notes to docs and projects into one space that just works with AI built right in. Try notion with notion agent@notion.com twist and squarespace. Turn your idea into a beautiful website. Go to squarespace.com twist for a free trial. When you're ready to launch, use offer code TWIST to save 10% off your first purchase of a website or domain.
- [00:50]
A
All right, everybody, welcome back to this week in AI, the new roundtable show that, you know, hey, I've brought to you because I need an excuse to talk to three smart people a week who are building the future in AI. And the vehicle for me to get smarter is to invite smart people to have a conversation for an hour. It's episode four. If you're new to the show, go to thisweekinai AI. Sign up for the email link to YouTube, link to Spotify. Whatever you're into. Man, what a roundtable we have right now. Philip Johnston is here. Philip is the co founder and CEO of Star Cloud. He's building data centers in space, something you may have heard a lot about recently. And that obviously Philip solves a number of problems, including heat, including energy, maybe. And obviously Elon has now pivoted Tesla and SpaceX Starlink. All of this has made him believe the future is data centers in space. You were obviously onto this for a couple of years as well. Explain to folks. And then we had Sam Altman say, hey, this is a pipe dream. So explain to folks what you and Elon know and maybe why somebody who's in the know as well, like Sam Altman would say, not possible. Ridiculous.
- [02:11]
B
Yeah. Well, firstly, thanks so much for having me on. Actually, I actually think Sam does know it. He was very careful with his choice of words. What he actually said was, data centers in space are not going to matter at scale this decade. And I think he very carefully said this decade and not in the next decade. So if what he means by that is in 2029, maybe less than 1% of all compute will be in space. He is correct about that. And I think probably Elon agrees with that. I agree with that. The difference is, in 2029, space computer is going to be growing at like 500% per year, whereas Terrestrial Compute will be going at like 5% per year. And there's going to be an extremely rapid takeoff in Space Compute. What it solves is the fact we're very quickly running up against constraints on where we can build new energy projects, particularly in the US in space. We can deploy these things very rapidly. The marginal cost of adding an additional data center in space goes down over time, whereas the marginal cost of adding an additional data center on Earth goes up over time because we use up all of the easy and cheap places to build them. So there's definitely a crossover point coming. I think that's the end of this decade, but we're building for that already because Starship's coming online in two years. And yeah, it's going to be massive also.
- [03:22]
A
And we have a lot of questions for you. I'm sure some of your compatriots today on the program also have questions about this vision. Next up from Bedrock Robotics is Boris Softman. He is the co founder again and CEO. You always got to get that right when you're talking about company, because if they're the founder and they're solo, that's one thing. But more typically, it's co founder. We always want to get that perfect. And they're retrofitting existing construction equipment with AI. Now, this is really important work. Excavators, bulldozers, all these autonomous machines, if they can be controlled by AI, not just autonomous, not just remotely, because remotely is happening today. Boris, I think you can maybe explain to the audience because they don't even know that's happening, but this is the next logical step. So the industry is currently using remote operators. I understand in some situations. But autonomy obviously unlocks a number of big things. So explain to us what's happening in the remote world and if that ever took hold and why. And then what's your vision here for?
- [04:26]
C
And thank you, Jason. Real pleasure to be on. Yeah, you're right. So remote operation oftentimes is called teleoperation. You'll see companies try to arbitrage cost of labor. Maybe they're in a remote location and therefore they can hire people in a easier location to staff. Or in a country outside the U.S. this has its own challenges because you still have a significant cost. You have complexities with latency capability, the nuances. And so it isn't it maybe has local applications, but at the extreme, full autonomy is the way to go, just like you see in autonomous driving on public roads. And so the construction space in general, there's this pretty fascinating divergence where the demand skyrocketing. So, you know, to the data center conversation, at least in the 2000 and twenties, the spend is astronomical. So just this year, there's $700 billion of construction spend on data centers. Just this year, the entire industry is typically about 2 trillion. So this is like a gigantic emergence that just didn't exist. And so, and this will continue for a number of years. And at the same time, the labor pool is just at a crisis point where there's already a shortage of hundreds of thousands of people. And there's a retirement cycle happening where a lot of our partners are seeing up to half the workforce retiring in the next seven years.
- [05:55]
A
And so it's this divergence, which is crazy. We had a whole generation stop doing this type of work. And what that led to was based on my research in the space and hearing from a lot of different startups trying to even handle education here, or autonomy. Those folks hit 60, 65, 70. They've made a fortune in the last 10 years because they're so in demand. Electricians, construction workers making hundreds of thousands of dollars a year, maybe even more. If they go on location to build a data center somewhere in Abilene, Texas, they get paid even more. Is there a wedge or an initial customer profile? I'm just thinking out loud here about visualizing dump trucks and those cranes and everything, and the different jobs they might do that you're starting with, Is it making roads, is it building foundations? What works first? What's the first application for AI and construction vehicles?
- [06:54]
C
Yeah, great intuition. So we're starting with excavators. This is about a quarter of all construction machines. It's also really highly utilized, one of the hardest to learn. And we're starting with the heavy industrials. So data centers, factories, warehouses, where exactly, as you said, there's an astronomical amount of earth to move. You're loading dump trucks 12 hours a day, 18 hours a day. And it's really hard to find this labor. And oftentimes these projects go on for 10 months, 15 months. And so that's our starting point. And so data centers become one of the really nice tailwinds in sort of a sector. And then that generalizes into just moving dirt.
- [07:31]
A
Moving dirt is hard and arduous. And these things will run 24 hours a day, seven days a week. If the construction zone allows it.
- [07:39]
C
That's right.
- [07:40]
A
And I'm guessing you'll have human in the loop to start. And is this in the field already and you have humans like monitoring it or they saying like, hey, I'm going to just draw a box here, work here to take out four feet and then level it, or is it, I'm going to just watch it work and pause it like maybe FSD rolled out over the last five years.
- [08:01]
C
Yeah. And so a lot of the roots of the company come from Waymo, where we're able to help launch this and really see how this scales on public roads. Similar pattern where we've been doing supervised autonomy testings on real sites since last summer. We're continuing to ramp autonomy and safety and we're going to go full driverless operator out later this year. And so that'll be the first entry point. And then you start to snowball and leverage the ML learning curves to jump into new capabilities. Trenching, demolition, other areas, other machines.
- [08:31]
A
All right, and finally from Resolve AI, which is building agentic AI software for engineers and site reliability. Because hey, when you start putting this software out into the world, it's got to work. It can't just be some crazy gen AI images that are being shared on social media. People are building software and it needs to be reliable. Is Spyro Xanthos Spyros. How are you, sir?
- [08:57]
D
Great. Good to be here, Jason. Thanks for having me.
- [09:00]
A
Yeah. This is a back to back Greeks on the program. Every week we feature one Greek on this week in AI because as Spyros, you can explain to our other guests and the audience, the Greeks. We created many things. Democracy, philosophy, you know, math.
- [09:19]
D
We have to leave our mark in AI as well nowadays. Yeah.
- [09:21]
A
So we're still innovating in the world, punching well above our weight. AI tools are making it easier than ever to run your own business, even as a solo founder. But you still need a beautiful attention grabbing website to help your new company stand out in a very crowded field. And you don't want AI slop. Nope, you want to use Squarespace. That's the easiest and fastest way to turn your idea into a real business. Because the team at Squarespace cares deeply about design and functionality. And a plain looking or generic or AI slop website, man, that's gonna be a red flag for your customers, for your investors and people who wanna come work for you and join your team. But Squarespace will take all the guesswork out of designing your first website with the blueprint AI builder, which has been finely tuned to make beautiful websites. Squarespace isn't just gonna help you make a new web page either. They're gonna be your all in one launching your business. They're also going to help you set up your email. They're going to handle invoicing, paperwork, all your needs. Go to squarespace.com twist for a free trial. And when you're ready to Launch, go to squarespace.com twist for 10% off your first website or domain purchase. Tell me a little bit you know, I had sent you in the group chat the Financial Times story about and it's maybe a good place for us to start. People are getting pretty excited about coding agents. They keep getting better and better copilots, et cetera. And hey, everybody can ship software now. And Amazon tip of the spear says, hey, you know what, let's start laying people off. Let's prepare for less people. This kind of matches their ethos over the past 25 years of just trying to be more and more and more efficient. Well, it turns out according to the ft, Amazon's commerce business has had a bunch of engineers going into deep dive code red meetings because of a bunch of instances where generative AI code made crazy changes. And they had a quote, high blast radius. Spiros, what's happening now? Because is the issue this software is not ready or is the issue people are trusting it too much and they've kind of leaned into like some people did in the early days of fsd, that keeps coming up. There were some people tragically in the Bay Area, I believe it was an Apple employee who just trusted Autopilot and another Navy SEAL in Florida. They just decided they would start watching movies in year two of the software when they were clearly told that's not what this is for, it's not ready. So what's your take here? Is it ready or are people trusting it too much?
- [11:56]
D
I think the AI works great. But you know, maybe to give you some background on how we do this right today at scale, at a system like aws, still the majority of time engineers spend is not actually writing code, right? Building new features. Actually, it's well known that AWS allocates a big percent of engineering time when they do planning for people to just be on call to ensure they're dealing with issues when they come up. Right. So that then they don't impact customers. Because obviously for an infrastructure provider like aws, that is catastrophic. So in a way the bottleneck at that scale has not been developing new code. Right? Not that that's not desirable, but this is not some new app that I write over the weekend, let's say, and published on Monday. These are systems that the planet relies on to run everything we do, basically, right? So in that sense, I think by accelerating the velocity of developing software, by having AI to create new software, we're changing the equation even more. From a point where essentially the bottleneck was already, how do I deliver these systems reliably to now moving at much higher velocity of producing code and pushing it into production. And that means that probably a bit less oversight in what gets pushed. It means probably software that developers don't understand as well as before because they didn't essentially, or they offered a lot more using AI.
- [13:16]
A
This is a key one, isn't it, Spiros? I just want to have you unpack this very important point. We saw this in aviation as well. As they abstracted the cockpit and people became very used to using the autopilot, very used to trusting the sensors. We had a series of tragedies because people didn't know how to aviate with the core metrics. They didn't understand, you know, that, you know, they trusted the technology too much, sensors would go off and they couldn't actually natively fly the plane in a robust way. Do you have that fear that we're going to have a generation of developers who are just not able to look under the hood or understand how to fly a Cessna 172 and they can only click the autopilot button?
- [14:02]
D
I mean, in effect, I think we have instances of that now, right, where essentially there is no deep insight in maybe how a particular application or infrastructure service works by many people. So when something goes wrong, and that's when you find out you don't have the deep intuition that we had before on how to do it. And that coupled maybe with people leaving these organizations, that makes it a lot harder now. I don't think the future is slowing it down either, though. And that's kind of what resolve is about. What I think is going to happen is the same way we kind of automate the creation of code. We need to automate the running and debugging of code. So we need to have AI that is at least as good as in, I guess, every stage of the way, right? Like code reviews, you know, then whatever happens in deployment, then actually monitoring it in production, and when something goes wrong, reacting very, very quickly, much more quickly than a human can and with more context than an individual can have. Right. I think we have to go through this transition, but I don't Think. I think the risk in the short term is, you know, we advanced code generation quite a bit without necessarily letting the rest of the stack catch up. Right. And that's what Resolve is pushing. Now let's have AI that is on call, let's say. Right. When something goes wrong in production, reacts a lot faster, has the full context of the whole system and helps us do that. But there is risk in the meantime. Right. Especially when, let's say vibe coding, let's call it, is applied to these infrastructure systems. I think maybe we're going too far sometimes. Just because it's possible, it doesn't mean it should happen also. Right.
- [15:28]
A
Yeah. This seems to be one of the key challenges to the industry is you get four or five good responses, Philip, and then your trust level goes up and you're like, well, I'm five for five and you know, using FSD in, using my lawnmower, that's, you know, AI enabled. So now I'm just going to extrapolate, hey, I can drive it in a snowstorm. I can drive, I can drive it any way I want. And so what are your thoughts here on the general pace that we're going and people's either adoption of it or, or. And are they overconfident in it, I guess is the real question. Are people overconfident? We need to take a pause here, Philip.
- [16:11]
B
Yeah, I mean, certainly we do see that. We see, you know, even, even where we are, where we have our. Everybody's been using it for 3D sort of CAD generation and all these kinds of physical simulation. We have to be very careful not to over presume the capabilities of it. So things like, for example, simulating thermals, we actually have to run the thing in a thermal chamber to validate that whatever the AI says is correct is right. So yeah, I think it will become less of a problem over time as these systems get more and more capable. But for now, it's certainly something to be very cautious of.
- [16:48]
A
Boris, what do you think in terms of the pace of the industry and the adoption? We had this very interesting report this week in America and we'll dovetail that with the employment story next, that Americans are not trusting AI. They have this deep underlying lack of trust. But what we experience every day, if you're in, in the industry, if you're in the room where it happens, is the opposite. An over reliance and an overconfidence in it. So maybe you could give me your thoughts on how we sort this out as an industry.
- [17:18]
C
Yeah, so we've seen this in accelerant as well. But you have to separate out the safety critical and decision critical applications from the things that are just accelerants of your day to day work. And so it's incredible for really fast data analytics, visualizations, dashboards, interface code. We're very, very careful on embedded systems hardware. You certainly have to be very careful about safety critical elements, but you know, it ends up being a facilitator when it's used in the right way. And so for example, you don't let AI tell you that an autonomous system operating a 100,000 pound machine that can kill somebody is safe. But it's incredible at analyzing historical data, creating test scenario ideas and accelerating some like mechanics behind the scenes. But at the end of the day there's a big difference between an LLM that you know, has to be generally correct, but it's, you know, not a catastrophe if it tells you something wrong versus if you're going to use it for an autonomous vehicle where you're going to use it for legal advice or medical advice. And so that's always been one of the most interesting challenges is that the way you qualify a system when you're optimizing for the worst case completely changes. And I think we've gotten comfortable with that context switching and I think much of the rest of society is still experiencing some of the failure modes and we probably need to do a better job of actually framing it in the right way because that's not currently being done automatically by these systems. It's up to the user to actually decide on how much you trust each parts of the systems outlet.
- [19:01]
A
Here's an interesting survey that our industry should probably think about a bit.
- [19:06]
C
The Pope is on top.
- [19:08]
A
The Pope, Pope Leo. Total negative 8, total positive 42 for a very positive score of 34. There's your benchmark. People also seem to like Stephen Colbert and Marco Rubio is not as hated as the rest of the people on this list, but at the bottom of the list is the people. The, I guess the, the previous management of Iran had a negative 53. Total negative score here, 8% positive. I don't know, I think that's their families and them. And Then total negative 61. Democratic Party right behind Iran in terms
- [19:45]
C
of their positive negative neutral's a win in this case.
- [19:50]
A
And there you go. Somewhere between the Democratic part, somewhere between ice, the Immigration and Customs Enforcement Agency in Iran is AI. Artificial Intelligence Notion is the AI powered connected workspace for teams. It brings all your notes, docs and projects into one space that just works. And with AI built right in, you spend less time switching between tools and apps and more time creating great work. And now with Notion's custom agents, busy work that used to take hours or never got done at all runs itself. Custom agents automate all of your team's repetitive workflows and they live inside Notion already. Maybe you want to keep track of what everybody's working on. Maybe you want to see which pages are getting edited. Maybe you want an analysis of what your team is working on. I'm constantly, constantly getting disturbed by pings and pings from Slack by team members with all these questions. Now Notions Q and A agent can research the answers from anywhere on the platform and get back to those people directly. In Slack you can design custom agents on your own. But Notion has a bunch of pre trained ones ready to go. Try custom agents now@notion.com twist that's all lowercase letters. Notion.com twist and when you use our link you're supporting our show and keeping it free and vibrant. Notion.com/g survey came out as well saying that companies, and I think these are ones with 500 million plus in revenue, only have a 9% chance. 9% plan to cut jobs due to AI. I think this is the key part of this survey. While 55% expect to increase hiring. And the reasons they give in this KPMG study reasons people don't trust AI include fear of misinformation, privacy, data concerns, lack of transparency, job displacement, lack of human oversight, and a general trust of corporations. So let's talk a little bit about the trust issue. Spiros, we'll start with you. Why are Americans so concerned here? And then how do you think about job displacement which obviously seems to be a big part of the uneasiness as well as what should our industry do to communicate, hey, there are benefits here for society and are we not doing that enough?
- [22:11]
D
First of all, I think I completely understand why this is a sentiment that people have around AI. It is very disruptive. It has actually a huge potential impact on jobs and I don't think the world is prepared for what to follow. I am a huge believer and a huge optimist in AI and I think the net benefit to the world is going to be significant. One of the biggest maybe advancements we've seen in technology definitely in our generation. But I think it comes with a lot of downsides and it comes with a lot of downsides for people who maybe are not involved in it. I don't think we do enough to probably explain or think through what the world is going to look like in the next three years. Models advance at the pace faster than maybe even those of us who are in expected. Right. This means that a lot of the, I guess have physical world and virtual world here. I think it means that a lot of white collar jobs are going to be highly automated and it means also a lot of the physical world jobs are going to be highly automated. And of course that's, that's desirable. Right. And probably it fills a gap like Boris was saying, but we have to think through the implications of all of that. Right. And I don't think we as an industry do much to think that through or explain to the rest of the world.
- [23:27]
B
To be honest, I was just going to ask Spiros what he thinks we should do to explain it better. I mean, yeah, it's quite a difficult one to convey to the general public. Is there anything you had in mind specifically to get the word out?
- [23:40]
D
We know we focus too much in what's happening inside the industry and not maybe enough in explaining what are the benefits of the world. Right. I think, you know, I think the story what Boris was describing to me is, is a real one, right. Where, you know, we don't, like, like maybe Jason was saying first generation people stop doing like essentially this physical world. Right. And now we're limited by that. Right. Our economy is limited by that. And I would say in the era in the world of software that, you know, I am in, I believe that, you know, the AI is actually going to result in a lot more technology, a lot more software. It's going to make a lot of things easier and simpler and cheaper. Things that are not accessible to people today are going to be much more accessible. Right. You know, like happened with every generation. Right. You know, my phone or Jason's phone is not more powerful than the phone that anybody can have anywhere in the world. Right. That wasn't the case maybe 30 years ago. And I think AI is going to become. Intelligence is going to become actually abundant and everybody going to have access to it. Right. And the rest of the world is going to benefit as a result. But, you know, that's not going to happen overnight. The benefits are going to be seen over a few years. Right. But I think we should be focusing a lot more on the benefits than,
- [24:39]
C
you know, if you look back at like some of the like revolutions that have happened, you know, go, go back like 150 years, you have, you know, everything from Internet, mobile, computer, industrial revolution, every single one of those there was a immediate fear that the disruption is going to be absolutely massive. And it's always a lot easier to identify the places that end up getting, you know, immediately shaken up and displaced. But it's a lot harder to see the secondary side effects. And so mobile phones created this astronomical economy with Uber and Instacart and, you know, and all of the other applications and services that became possible. Same thing with the Internet and computer. You go back to industrial revolution. It was massively disruptive, where almost every single type of, you know, fact, traditional factory job, or even farming job got completely kind of shaken up. And everybody thought that there's going to be the end of labor. What ended up happening is by the end of it, it was, it was disruptive locally. But when everything settled, the number of jobs actually increased, the average salary doubled, the productivity skyrocketed way more than that. And you actually had a big net positive for overall society. And so one of the elements here might actually be, like you said, spirit is like highlighting the wins where we need the like, drug discovery and the cancer treatments that end up being giant wins that were discovered for the first time by AI and not by, by scientists doing, you know, massive experimentation.
- [26:01]
A
How do you frame it at your company, Boris? And I'm curious, the folks you're selling into, I'm sure, are like, yeah, we need this because we're behind on these three projects. But then, and I don't want to stir the pot here, there might be in some countries people who have very strong union protections and say, hey, you can automate this, you know, excavator all you want, but we need to have a supervisor and we need to have somebody in that cabin. So automated if you want, but we still want our two guys in there making 75 bucks an hour. What's the vibe like when you go bring this to market?
- [26:36]
C
Most of our partners are actually employee owned, where they literally have like stock options in their own company. And so they succeed when the company succeeds. And the physics of these companies right now are that they're turning down lots of jobs because they physically can't stop the work. And so they see this as this massive expansion rate. So instead of pushing their people to work 60 plus hours a week, where you start to get fatigue and safety issues and then turning down the rest of the jobs, they're actually able to actually take on more work, utilize their equipment and their teams better. And so it becomes genuinely expansionary because the demand just exceeds supply.
- [27:10]
A
Which is very different than say Uber drivers and Lyft Drivers who, you know, I was talking to Dara, a great Iranian American. Yeah, he's amazing, Amazing. And he was saying, yeah, you know, we're still growing in San Francisco, Louisiana, wherever Waymo is, but we don't need as many drivers there to match that. And we're gonna have our robotaxis. So we're just not doing driver recruitment in those markets because we, we don't want to make it harder on those folks. So then people see the writing on the wall in Austin, Texas. I think we have five people testing right now, maybe six in this tiny little million person, you know, enclave, mid sized city or even on the smaller side of cities. So it's super interesting. By the way, back to our other discussion, I just asked Claude to tell me what are the, what's the industry term for our previous discussion about automation? It's automation dependency and skill degradation. Automation dependency, right. Create skill degradation in aviation. That's what they talk about in aviation is this automation, automation bias. The tendency to over trust and defer to automated systems when a manual intervention would have been better. And so we see that in code, we'll see that in robots and we see it in cad. Right. That's going to be the skill in some ways, Philip, isn't it? Knowing when to stop trusting this, Knowing when to blindly trust the system. Hey, it's going to get it right. It's like a spreadsheet. Two plus two is four. We haven't seen it make that mistake in a long time. But hey, when we're making this more complicated thing, we then we should have human in the loop and then hey, for this, if we're making a specific bolt that goes onto this, you know, satellite, it's got to be human designed and then maybe AI tested. I'm just giving an example here. Philip, you're. Yeah, good expand.
- [28:55]
B
Yeah, yeah, 100% I think, I think that will be where the sort of niche labor is going to be. That skill set of knowing when you can use the models and when you can, when you need to intervene.
- [29:07]
A
Yeah. Having had this discussion countless times with David Sacks and he's in the administration doing AI and I'm on the streets watching startups and I see the startups just saying this is unbelievable. We don't have to hire any more people. We can get to a million dollars with just the three co founders. We're not going to hire. It's too much time to hire, too much culture friction. We'll just build it ourselves. We can do it all with AI and then on the other side, you're seeing the big companies say, yeah, maybe we'll lay out 5% of people, but. But we think AI is not ready for primetime. The startups are the most curious and they have the least amount of resources. So of course they will figure out how to apply this technology first, along with criminals, gamblers, and, you know, the people at the bar on Mos Eisley. I was explaining this to my wife last night, Spheros. I was like. And I was telling her how much I'm into this, like one specific piece of technology and I won't say what it is right now. But she said, well, why are you so into that? I was like, well, I always look for two things when these new technologies come out. What are the hackers and the tinkerers, the gadget people, the GitHub people, the Reddit hacker news people. What do they like to play with? And number two, what are the gamblers I know and the criminal element? How are they adopting this technology? Those two groups of people, they're the earliest of adopters. I was actually talking about stablecoins. Actually, I can say it. And I was like, you know, this is where I know stablecoins are like legit now. The hackers are wall trigger, figuring out how to save money. And the gamblers I know settling poker disputes for us. They're like, just, yeah, send me, send me tether. I don't need anybody to know about this driver. Send me zcash. I'm like, what's zcash like? Oh, that's really anonymous. If you really want to sell your poker debts, use Zcash or Ms. Wallen. Like, nobody will even know it ever happened. Pretty hilarious, by the way.
- [31:04]
D
I don't even remember this. The last time I saw you were playing poker. Where were we?
- [31:08]
A
We were at the Go to conference.
- [31:10]
D
Yeah, the conference.
- [31:11]
A
Yeah, that was fun. We had a good time. I mean, that was a profitable event for me. But I've been thinking about this in my debates with Sachs because, you know, steel sharpened steel kind of situation. We battled over this over and over and over again. And I think I know how to solve the PR problem. I think I figured it out, Boris. The three things that the technology industry has had the least amount of impact on because it's the most regulated here in the United States. We all know healthcare, education, construction. Boris, we have you here today. So you're in one of those. Those three are where consumers just spend so much money and if we could communicate as an industry. Hey, your house is going to be cheaper to build and faster to build. So we're going to build more houses and they're going to be cheaper. Hey, your child's education, you don't need to hire a tutor. This adaptive learning, your AI tutor is going to really level the playing field with some rich parents who have tutors there all weekend long to do better on the SATs, yada yada. And then on a healthcare basis, where are the startups saying, hey, come to, you know, this pod and this micro healthcare center and we will do all of your health stuff, you know, very easily, very cheaply. And that's happening overseas. People go to Turkey, they go to Mexico and they have these health retreats. Now. I don't know if you heard about it, but my wife, second reference here, went to one recently where she got some stem cells. She did this, she did that. All of that available, like for the same equipment but for lower price in a, you know, cash kind of way. So what do you think, Boris? If we, our industry could handle those big three and show declining prices and more availability, this, this might turn things around, I think.
- [32:59]
C
So you're hitting on like some of the most foundational things of life, right? Like what do people care about, you know, their health, education, shelter, food's the last one. Right. And that's a good punch up a lot of the foundations of life covered and particularly when the cost of living is going up. It's, this is a, this is actually pretty meaningful. But these would be, you know, big wins and, and ideally it's like it starts to impact everybody where we're feeling the impact as, you know, as, as drivers of the businesses that go and develop products. But that doesn't, you know, broaden out to, to the rest of society immediately. It has a, a delayed effect. These would be astronomically valuable particularly around education when you have such diversity of availability even within a metropolitan area. Yeah, I do worry because there's a backlash that could become bipartisan against technology over the next five, 10 years if this isn't actually thoughtfully managed in the right way, just because it's such a lumpy impact in terms of parts of society. And so it's important to actually celebrate these wins and actually try to accelerate the ones that impact, you know, the overall foundation of the country.
- [34:16]
A
Philip, we had the administration, you know, which is obviously super engaged in AI and the build out, they said, hey, we're going to get this agreement done. And I saw Sachs announced it, I think last Monday. Electricity is not going to be impacted. Your Electricity price will not be impacted by data centers. So that's kind of a reassurance of a fear. It's not aspirational, but at least I think the industry is starting to get attuned to let's lower the fear about some things and let's. I think the next thing we all agree here is maybe amplify the benefit. What's your take here on our PR crisis in AI and is it going to get worse, better, and what should we do?
- [34:56]
B
I think that definitely goes a long way to helping it. Like if people started seeing their energy prices going through the roof, that's really what it would have been a disaster for the PR thing. I think one thing we need to look at as well is sort of Andrew Yang style universal basic Income, you know, and, and in some senses we already do it. I mean, I don't know if what was this thing? Brad Gerstner has this, you know.
- [35:24]
A
Oh, the Trump accounts Invest America, which they brilliantly rebranded as Trump accounts and magically they sell right through. Very interesting, very interesting. By the way, Star Cloud is now Trump Cloud or Star Trump. Yes. You've just got all your approvals. You've just been approved to put 100,000 satellites in space. Congratulations, Philip.
- [35:49]
B
I will very happily rename it to Trump Cloud in that case.
- [35:52]
A
I mean it's just an idea. DJT Cloud.
- [35:55]
B
You never know,
- [35:57]
A
things could go easier
- [35:58]
B
for you with the Trump accounts. I mean that's essentially a form of UBI and I think there'll be more of that over time and that'll be financed by taxing essentially tech companies. I mean that's basically how they're financing these things now.
- [36:12]
A
So controversial one, Philip, that people are talking about outside the United States because there's been now paradoxically or randomly, Wuhan in China has had protests over self driving and there's been signaling from the CCP that they're going to give out a certain number of licenses and they're going to tax them in a different way for self driving cars. Then New York State said, hey, you can't give professional advice from your large language model. And I was like, okay, that doesn't make any sense, like legal advice or health advice. Why can't a large language model present some information with a disclaimer that it's up to you to check it. Just like a podcast or a book. But we're starting to see the regulations. Boston also I think was a little anti self driving. So it's bubbling up. But spirits. What do you think is some Way the proper way for the government to be involved, if at all. Or are you just, hey, free market, let it rip, we'll figure it out afterwards with unemployment, et cetera. Because we have done that in the past when we had Covid, we extended unemployment, I think indefinitely for a period of time. You know, we can react in the rearview mirror. But any thoughts on taxation and other systems for thinking about job displacement?
- [37:28]
D
Yeah. So first of all, I think that, you know, starting maybe with my perspective on this, like, I think AI is the way out of our trouble as a country, in my opinion. And I think like, you know, all other discussions aside, you know, if we didn't have AI, we would be in much deeper trouble than we are today, you know, the government included. So in that sense, I think we need to lean in and we need to. The government should, for the most part, get out of the way and let the industry work because we're playing at a global scale as well. It's not just what happens inside the US is if US falls behind, somebody else essentially wins. And that's not the future we want to some extent. So now, that said, obviously we've been very thoughtful what happens with job displacement? And I don't think taxation and slowing down is the right answer. The right answer is probably just letting the technology evolve as fast as possible and then figuring out ways to deal with what happens to people who are essentially out of the wrong side of this. And we should actually doing everything we can for those. The simple example we gave with electricity is obviously very simple, maybe thing to do and it makes a lot of sense. Right. And I think there are probably other similar things we have to think through, especially with real world automation. Right. And robot access and all of that.
- [38:41]
A
Right.
- [38:42]
D
What happens to everybody who was making a living out of that? But I don't think the answer is stop robotaxis. Right. The answer is obviously that robotaxis happen and then deal with, you know, or help all the people that maybe lost the job as a result. Right. In a very short period of time. You know, that's my perspective now. I mean, I don't know if I have the answer what needs to happen in terms of like the actual implementation of something like this. But I do know that we cannot slow down technology. Right. We're going to pay the price if we do that.
- [39:05]
A
All right, let's talk a little bit about real world models. We have seen massive disruption in a number of industries because of, you know, guess the next word, models, copilots, legal. My Lord, what You can get done with these things is amazing. And agents being able to use them to navigate the real world or to navigate the online world has been phenomenal. But Yan Lecun, I believe is how it's pronounced. Lecun, he's raised a billion dollars to bet on world models. His new startup, Advanced Machine Intelligence, raised a billion at a $3.5 billion valuation. Largest seed round ever in Europe. And he wants to build real world models. Obviously Boris are operating in the real world and FSD is operating the world. And then there was a second story figure, which is a $40 billion startup that maybe has one or two customers. BMW is a little controversy around that. And we can play the video here. The founder, I believe it's Brent. Is it Brent or Brett? He shared this video. Elon responded and said, is that remote operated or not? And it created a little back and forth. But here you see this robot just tidying up, spraying some Windex on a table and extremely slowly and lethargically wiping it in a circle. Wax on, wax off, this is. And then throwing it over his shoulder or their shoulder. It's the shoulder. I don't want to misgender this robot, but they them seem to be doing. They're the slowest housekeeper ever in the world, but this will be the worst it's ever been. Boris, when you see that, what did you think? And then just generally world models, obviously Nvidia and Jensen are building some open source ones and trying to just help everybody understand the real world so they can build AI app. What do you, what are your thoughts here? When you see this video, it's impressive
- [41:05]
C
to see the sort of progress that, you know, Humanoids in general making. But it's always very hard to know in a video like this how much has been optimized for this particular setup. And how versatile is it when you get into, you know, general households? Because that's where it actually gets hard when you're thrown into the infinite permutations of what the world actually offers. And you see incredible, you know, demonstrations like this from a number of companies that, you know, then have trouble in the last 5% or 10% that makes it like still pretty far from a commercializable product.
- [41:36]
B
What did Brett say? Just to jump in. What did Brett say when Elon asked, did he say it is teleoperated or. He said it wasn't.
- [41:41]
C
He couldn't have possibly said that.
- [41:42]
A
He said it wasn't. Yeah, he said it wasn't.
- [41:44]
B
If it isn't, it's very impressive.
- [41:46]
D
Yeah, but there is the question of is it teleoperated and has it been optimized as for this particular video also, Right?
- [41:51]
C
Yeah, because a lot of these companies have like, almost like shops of people, like teleoperating robots in order to build the data and then it can be hyper fit to a particular application.
- [42:01]
D
Yeah.
- [42:01]
C
You know, and then, and then you go and you do it autonomously. But you're learning from a lot of demonst. Like in a lot of ways it's interesting because an LLM has become so generalized when it's applications, but it's partially because there's this infinite amount of data with relationships between words that have existed in a fairly similar kind of context and feature space across all the use cases on the Internet. And so when you then tune it for law or for medicine, there's a lot of carryover when you think about the physical world, that data doesn't exist. And you also have the complexity of unique hardware, unique sensors, unique kind of applications. And so, you know, you go and you kind of tell the operator you can learn any specific thing, but it's almost like hiring a team of writers to train OpenAI versus using the Internet. Right. And so you're, you know, putting drops in the ocean which so by the
- [42:49]
A
way, there people are doing. Right? That's scale AI. We're investors in Micro One. We, we see the company, they're literally hiring really talented people to find that last 5% and iron it out and say yeah, these legal questions, we, it's hard and they're just like, hey, let's iron out this last little edge over here. But yeah, when I saw this, the first thing I thought was, you know, Robo taxis in Waymo, operating in a very narrow constrained geography is this is the same thing, right Spiros, just, let's just train it on this 7 by 7 square mile area and get it perfect before we keep extrapolating. And then even in that case with Waymo and Robotaxi, they all have remote monitors. Some people say it's 2 to 1, 3 to 1, 4 to 1. I think Baidu said it's like 50 to 1 or something.
- [43:38]
C
They're already at the one case is a bit different because even if, like just take San Francisco, the amount of infinite chaos that can happen in a city like San Francisco is almost infinite. Like the horrors that we've seen is just, you know, couldn't even begin to explain. Like we literally had like enough people jumping on the front hoods of our cars that we had rates behind it in every city there would be like a civility rating on a, on a city. Right. And.
- [44:03]
A
Oh really at Waymo.
- [44:05]
C
Yeah.
- [44:07]
A
Singapore, nobody's jumping on the cars.
- [44:09]
C
No, it's like they're very disciplined.
- [44:10]
A
Nobody's jaywalking.
- [44:11]
D
What is San Francisco borders.
- [44:12]
C
What did San Francisco rank way higher than Phoenix. And then Austin, Louisiana. Like literally it just happened.
- [44:18]
A
And it was right above Calcutta and Baghdad.
- [44:21]
C
Yeah, exactly right there. But it's like, but you do, you can't escape the long tail safety challenges. And so anything that can happen in driving will happen in San Francisco. And so it is like a very broad solution. Just as a robotaxi, you have physical infrastructure, which is why you stamp it out by cities. That's a much broader scope in the driving domain than the household cleaning situation. But you have a very verticalized solution. You have a deep focus on driving. In this case, you have these vertical solutions that require a giant amount of data and a giant amount of focus. This is where world models are actually pretty interesting. It's not immediately applicable, but when you think ahead some number of years, the big breakthrough that can normalize physical AI development might be a giant breakthrough in simulation where you actually have a realistic representation of the world where now data is not as you can break through this bottleneck of data in a more robust way. But there's a long way to go to get there. But it's such an incredible and tantalizing unlock if you can get there, maybe.
- [45:27]
D
Jason, what we've seen happening, let's say with LLMs, is that that as we improved, like I said, the reasoning, then we applied RL to all sorts of problems. So they took a general purpose model that is very good at reasoning, gave it tools and then had it try a task as many times as it needed, math being the simplest, I guess, because you have an answer always. But we took other problems that maybe the answer wasn't that so obvious and let the models play with it. And then we made them very, very good at one particular task, one particular task at a time. And now we've allowed them to write code on our behalf or do math or all sorts of other things. I suppose that's applicable to the real world as well, right? Especially with develop models that visual, let's say simulation models where essentially we can try things. So I think the future is inevitable in the real world as well. But I'm not an expert in that to know the timeline.
- [46:12]
C
Yeah. And the thing is you just can't experiment that way in safety critical like you Wouldn't be able to do RL on a public road driving. But you can in picking things up. And there's been attempts at it. When you get into simulation, that's where you have some incredible breakthroughs. So some of the best applications of walking technologies and quadcopters and incredible balancing, that's actually simulation trained because you're able to actually simulate physics very accurately. And then you can do tricks like bore it where you may not know the exact physics of the world, but you're learning a superset of the world's physics. And then when you get out into the real world, you're superhuman in your ability to walk and balance.
- [46:49]
A
So to give an example of that, if you're a quadcopter company, you're Archer or Joby, you have the simulation, you understand the physics. Now you can do edge cases like what if we had a crosswind followed by a lightning strike followed by a bird strike and let's see if Sully can still fly, you know, the plane.
- [47:10]
D
My understanding of how self driving cars essentially get trained is this. Right. You had a model that essentially simulate the road and what others did around you. Right. And then the car had to react basically. Right. Isn't that the case?
- [47:22]
C
Yeah. And it's a lot more heavily on imitation learning where you're capturing this super complicated relationships between all these things on the road, pedestrians, vehicles, cyclists. But it's incredibly delicate and subtle because your safety is a function of your interactions with everybody else. And so you have to model all these complicated interactions. But with enough data, you're able to capture these patterns in a way that's way better than traditional kind of engineered solutions with heuristics and search.
- [47:49]
A
Philip, you must be doing simulations like this or planning to do them for when you have, I don't know how many thousands or tens of thousands of these satellites do you plan on having out there in space? And you have to now account for, I don't know, are there 7 or 8,000 Starlink satellites out there now? How many satellites are going to be out there? And how do you account for all the collisions and possibilities?
- [48:13]
C
Oh my gosh.
- [48:15]
B
Yeah, we've just filed for a constellation of 88,000 with the FCC. Elon's just far.
- [48:19]
A
You said 88,000.
- [48:21]
B
88,000.
- [48:23]
A
So you're a little suspicious. You didn't go 88888?
- [48:26]
B
Yeah, no, that would have been. Yeah, we should have. Yeah. The Chinese would be happy if we caught it 88. But no, no, Elon's just filed for A constellation of a million. And it's all going to be in this. Yeah, yeah. For a million of this frizz AI satellite constellation. And so with that you can deploy. We can deploy at least 20 gigabyte gigawatts of capacity with that. I think Elon's targeting 60 or 70 gigawatts capacity with that. So maybe 100.
- [49:00]
A
And I guess when that happens because you have so much, I don't know how. I guess you would have to be low Earth orbit or at least be mid Earth orbit. How do you think about that? Because you don't need to transmit all of them, don't need to transmit down to Earth. And you, I guess, could hitch a ride with Starlink or Bezos's and Amazon's infrastructure there. But is it low Earth orbit? Mid Earth orbit? What's the best place for these to exist? It's low Earth.
- [49:27]
B
Yeah. The reason you want low Earth is so that you can serve all inference workloads. For example, if you want to do Voice agents for AI or ChatGPT or video generation. Actually, yes, most workloads would be fine in medium Earth orbit, but you know, for video generation would be totally fine because it's going to take 10 seconds anyway to generate the video.
- [49:47]
D
Got it.
- [49:47]
B
But things like voice agents, you definitely want to have in lower Earth.
- [49:50]
A
Explain to us how large that band is. Is it 100 miles wide? Low earth? Is it 500 miles? Is it 10,000 miles? I've never actually asked that question because that would determine. Hey, there could be. If it's 100 miles, you could just give each person a mile and you could have a million in each of those hundred miles and you'd have 100 million satellites. How does it work?
- [50:15]
B
Yeah, exactly. So it's from about 400 kilometers. Usually people say low earth orbits up to about 2,000 kilometers. Anything above that to about 20,000 is medium, medium orbit. But it's a lot. You can, you can probably put at least 10 terawatts of capacity there. That's 20 times the entire UF power grid. And then once you've, once you've, filled that up, you can go to medium Earth orbit, you can go to CIS lunar orbit, you can go to the Lagrange points. So there's this enormous amount of capacity you can deploy.
- [50:40]
C
Phil, can you, like so curious about this. Can you explain the unit economics of this, where it's obviously more expensive to kind of get this out into space, but then you presumably save a lot on energy, maybe cooling, although it's not super like what is the long term kind of convergence of this in terms of the advantages over Earth data centers?
- [51:02]
B
Yeah. So the main one is on energy and infrastructure. So the best comparison is with solar on Earth because solar is the cheapest form of energy we have on Earth. So the biggest three costs there is, number one, the cost of permanent land. That's the largest cost, especially in North America. Number two is the cost of batteries and backup power. And number three is the cost of the solar cells themselves. So in space, number one, we don't need permitted land. So your biggest cost is gone. Don't need batteries and backup power because you're 24,7 in the sun. So your second biggest cost is gone. And you need 8 times less solar cells because 1 square meter of solar panel in space produces 8 times the energy of one on earth. The main additional cost we have is the launch cost. Yeah, and so there's, there's a break even point where the launch cost gets below the cost of permanent land, batteries and solar. We see that's around $500 a kilo break even. And you know, Starship is targeting launch costs of marginal launch costs for SpaceX of 10 to $20 a kilo. So yeah, it's well within range of what, what's coming.
- [51:53]
A
Just to recap that getting to space is the cost, the energy cost is, it sounds like, is it 80, 90% less?
- [52:03]
B
Yeah, 90% less.
- [52:05]
A
90% less on the. And then is the, does that include the heat dissipation cost in the energy there or is that like another level of energy that gets taken out?
- [52:16]
B
Just the infrastructure cost of what we're doing is around $5 million per megawatt. Terrestrially, if you build new infrastructure, you're talking about 15 to 20 million dollars just for the infrastructure. So that's cooling towers, chillers, batteries, backup power, none of which we need. And so that's about 4x cheaper. But then you add in the fact that our infrastructure cost includes all of our energy over the next five, six years. So that's all of the solar panels and radiators included. So in total we're talking about 10x less in terms of both infrastructure and energy.
- [52:45]
C
Yes, we designed the chips to be radiation hardened as well.
- [52:49]
B
No, we're using bog standard. So we launched an H100 in November last year and it's working remarkably well so far. We just trained the first modeling space and did a bunch of other things like that.
- [53:00]
A
And those prototypes cost low tens of millions I think to make and put up.
- [53:06]
B
We did this one for like two and a half million dollars. The one you see on the screen here is two and a half million dollars. So that was about 300 grand for launch for SpaceX and thank God for Elon and SpaceX because we would never exist without them. About a million dollars for the bus and then another few hundred grand for the computer hardware inside.
- [53:24]
A
And what will these cost at scale? If you put a thousand of these up and then your 10,000th up, what are they going to cost at scale, do you think?
- [53:31]
B
In five years we'll be launching these 200 kilowatt 3 ton blades that fit on starship, so we can launch 50 of them per Starship. So it's about 10 megawatts of compute per starship. Just the infrastructure on that is about $50 million. If you were to do that on the ground, it'll be about $200 million. And then obviously we don't pay for energy over the life of them. The chips are by far the most expensive part of what we're doing. So the chips on that would be, you know, another few hundred million dollars. But yeah, it's, it's, it's.
- [54:00]
A
Now for the crazy question. Yeah, Is it possible to fabricate some of this on the moon eventually?
- [54:08]
B
100%. 100%. Elon is, I mean, come on, factory
- [54:12]
A
to make H1 hundreds on the moon. I mean now it's, we're kind of
- [54:16]
B
not stretching, not H1 hundreds. You probably would not do the, you do, you do the H1 hundreds probably on the ground for quite a long time because they're, they're, they're, they're pretty light anyway. But the things which are, you have a lot of aluminum and silicon and you need both of those for the solar panels and the radiators. That's the biggest mass. And also the satellite bus, all of that is like 80 or 90% of the mass of the satellite chips is quite a small proportion. So you'd definitely manufacture the solar panels and radiators and satellite bus on the moon and then shoot them back to earth.
- [54:44]
A
Do the elements to do that exist on the moon? Sorry for not going to graduate school, but it literally exists on the moon.
- [54:53]
B
Yeah, literally. Silicon is about 30% of the lunar regolith and which is, you know, necessary for the solar panels and then aluminum is the rest.
- [55:01]
A
So, so this isn't spiros as crazy as it sounds, you know, and I think you're taking a part of the. I loved your investigative Colombo analysis of Sam's quote because Sam doesn't have access to space. And if he wanted access to space, he would have to bend the knee to Elon to get access to Genuflex. So that's untenable at this point since they're still in a lawsuit. But he didn't say a decade, he said this decade. This decade. So that was, I think, very cool parsing on your part.
- [55:36]
B
Hey, you know, until, so just say until a few months ago, Sam was actually quite bullish on space data centers. For example, when he released O3, the first task he got it to do was to design a radiator for a 1 GW space data center. That was the video.
- [55:52]
D
Oh, interesting.
- [55:53]
A
I want to hit on this Andrew Kaparthy weekend project. He's got a side hustle, Spiros. And maybe you could explain his post, his GitHub repo and what he incepted in the world because it has inspired everybody from Toby Luckey, from Lutke, from Shopify to start playing with this. And all the open claw people are like, oh, I set up an agent myself, I might as well set up my own language model and start doing experiments again. It's like the distance between science fiction and our reality just seems to compress every day or every week.
- [56:35]
D
Yeah, I think in simple terms what did is to essentially build an agent that, you know, tries experiments on his behalf. And these are simple experiments in tuning a model basically. But I think obviously this is a prototype to show how the future looks like maybe. So essentially in this case, I think the agent, the research agent that he developed can modify essentially the code to essentially try an experiment. Then he can run it, evaluate the output, runs for five minutes, I think, see if it improves or not based on some test he has. But essentially this can run on its own overnight. It tries simple things. It's not going to lead to any breakthroughs yet. But I think it's mostly a vision for the future where maybe even these harder problems that are all human intuition based, let's say, and long years of experience can be delegated to agents. I think what we've seen happening with OpenCloud or even maybe coding agents, we can let them run for a while. A single individual can probably consume in tokens a lot more than their own salary. Right?
- [57:41]
A
Yes, I've seen this.
- [57:43]
D
I think it's happening, by the way, I think we're going to see it a lot more this year. Maybe if, let's say coding agents is the canary here, we see that engineers probably consume more in tokens than their salary, but then they produce 10 times what they could on their own. So that equation works out. I think what we see here, maybe with Karpathi, is that can be applied to many other things. Eventually it's going to be applied to all types of maybe virtual work. So in a way, I completely understand the excitement around this because it proves that maybe even very hard problems and research can be delegated to agents, that maybe they're not as smart as we are, but they can work non stop.
- [58:22]
A
And the next thing Carpathy did, Boris, was to say what would happen if we had 1,000 of these as essentially graduate students running experiments and then had them talking to each other. And this is the idea that I think is getting particularly interesting with open source and this hacker movement in AI Boris is. We're seeing from Shanghai to Brooklyn to the Bay Area to Austin, meetups for OpenClaw and everybody from knowledge workers to moms, you know, homeschooling. We had a homeschooling mom on who's using openclaw to, you know, do, you know, tasks to help her with her homeschool, with her kids. I mean, it is infected everything. And now we have this, you know, model building going out to the edges. And here Andrew Karpathy says, I tried a few setups. Eight independent solo researchers, one chief scientist giving work to eight junior researchers, et cetera. Each research program is a git bench. Each scientist forks into a feature branch, git work, trees for isolation, et cetera, yada, yada, yada. This is like, Boris, we're down the rabbit hole now. We're like firing off entire graduate school teams. There's only 3,000 PhDs, I think, of note, in AI in the United States. And you all are battling selling for them to be on your teams. This could go to 30,000 or it could be 300 million if this continues.
- [59:54]
C
Yeah, it's like you start thinking about like modularity of your agent systems versus your software and architectures. Up to a point. What's fascinating about these is that it almost more abstractly just points to the fact that the scientific method is being automated and you have a fairly well structured problem in terms of like optimizing a ML model where you have dimensions you can push on, you have a very clear way to evaluate it and kind of iterate on it. It makes me excited for the fields we talked about, like medicine, where it's exactly what is done at large scale in universities and research labs. And at the end of the day, these models have an ability to interpret giant permutations of dimensions of the data coming back and potentially be a lot better on the next experiment than independent, kind of like isolated students or researchers might be. And so to me, it feels like a next step of automating, like the, you know, just the scientific method of running experiments, accumulating those results and taking next steps. But now you can paralyze it in a way that's shocking, and that's what's probably most exciting about it.
- [60:58]
A
All right, let's end on the Claude, or I should say Anthropic V. Anthropic versus the military and the government. Obviously, you gentlemen have been probably watching this back and forth. I'm curious, Philip, if you have a take on what's occurring here. You know, you have some conscientious objectors or concerns. Hey, don't use our software to build murder bots. Don't build our software to build a police state. We had Emil Michael from the Department of War on all in last week. I'm unsure if you guys saw it or not, but. But he sort of explained, like, we need to use these tools how we want to use them, and we already have a system of law. We don't need Dario to be this, you know, interpreter of the law. We. We have the law and. And just sell us the bullets and we'll point the gun in the right direction. What was your take on this whole back and forth? I'm curious.
- [61:52]
B
Yeah, I think it's a really tough spot for. For Anthropic and Fedaria in particular. I think the. The military definitely, if they're paying for technology, has the right to use it how they want. I do think, though, the private companies are allowed to say that they don't want to work with the military and not have the military threaten to, you know, cut off all of their other customers by saying they're now a supply chain risk. I think if they go down that route, it's a very. It's a dangerous path for the military because what it means is new companies coming up will be much more hesitant to sign any contracts with the military if they know that, oh, if this goes bad, we're going to lose all of our. Everybody now we're going to be called a supply chain risk, whereas if we just don't engage in the first place, then there's no downside for us. So, yeah, I think the military should be careful threatening to call people supply chain risks. But, yeah, I definitely think they have the right, if they pay for technology, to use it how they want.
- [62:49]
A
Yeah. And to Just give you guys the definition of this statute. U.S. c. 3252. The statute defines supply chain risk as the risk that, quote, an adversary may sabotage, maliciously, introduce unwanted function, or otherwise subvert a covered system. Pretty aggressive. Spiros, how much of this do you think is performative on the two parties? Because, you know, this administration can be a bit effervescent in their responses and performative. And let's face it, Dario and Anthropic, you could make the same critique of them. So what are your thoughts trying to get through these two?
- [63:28]
D
I think so. I think there are, you know, big egos. And, you know, first of all, Anthropic was a supplier. Like, they gave their models to the government, like.
- [63:36]
A
Sure.
- [63:37]
D
Especially when they were a smaller company. Right. And I do think, like, there was a bit of. I mean, I understand maybe the dynamics inside the company. Right. Or in the valley. Right. And the balance they're trying to keep. And, you know, I respect that to some extent. Right. But I do think it is performative to some extent. And I do think also the government maybe is going too far. Anthropic has amazing models. They should be able to use them. I think punishing Anthropic with the supply chain risk maybe is extreme. It doesn't help either party. Anthropic obviously loses as a result, and they are to blame to some extent for getting to that point. But also, the government would rather have access to their models. I don't think that makes sense either. On the other side, there are great alternatives, and OpenAI, especially the new models seem to be catching up. But, you know, there's no reason to not have access to anthropic models. I think, you know, just maybe, like, being a little bit cooler on both sides maybe would have helped them in this case.
- [64:25]
C
I agree with Philip's take. It's like, you know, they're free to do business with the government if they want to, but if they do, you know, the government has, you know, can use the product however they need to. It's. It feels a little bit. It's surprising how broad and heavy handed, you know, the response is, because the unfortunate part of this is that those models are actually some of the best for certain applications. There was a lot of invested time by a lot of other agencies to be able to leverage them. And now having to replace them, even if it's temporary, is actually a giant waste of effort and distraction. And it's not just the government itself. There's a lot of companies that now indirectly do business with the government that are in an area of ambiguity on whether they're allowed to use these models if they go and do business with the government. So there's a lot of secondary impacts here that in the end will probably get unwound at some point and we'll end up with a lot of ways to effort. That's unfortunate.
- [65:16]
D
The sooner the better, by the way.
- [65:17]
A
Yeah, I don't know if you guys saw this story. I. I called Cap on this one anthropic clause from Robert Wright from NON zero News, which I've never heard of and he said Anthropis Claude helped select hundreds of targets for the opening wave of Iran strikes. There's a good chance that one of them was the elementary school where more than 100 girls died. My latest non0news piece. This can't possibly be true because the military would never use an LLM to pick targets and do that blindly. Anybody have thoughts on this? I'm trying to have my team track this down, but it got almost a million views on X so I'm not sure what the state of this is. But any thoughts on would you even consider using this to pick targets at this moment in time? Philip, it feels like not wise.
- [66:15]
B
I would be very surprised if they were using flaw to pick targets in Iran. I mean for one thing, Iran thought is not trained on data to do with Iran particularly, I imagine. And even if it was, I mean the logic gates in. For the, for the LLM in particular, maybe other models have logic gates which would work for that. But yeah, I would find that very surprising to me.
- [66:40]
D
This is like blaming Microsoft for Excel when it's used to sort targets. Right. Like it's the same thing here, right? Maybe somebody put some data into Claude to try to sort them out. Right? Or you know, figure out, yeah, they
- [66:50]
B
could have done that.
- [66:51]
D
There's no more than Excel, right.
- [66:54]
B
They could have said rank these cities by population size like. And Claude could do that for sure. But that's not.
- [67:00]
C
No way human.
- [67:02]
D
I mean maybe they have more proprietary data, right, that they had to go through, let's say, right. To make decisions. But again that's akin to Excel and bringing Microsoft then for the use of
- [67:11]
A
Excel in military also I use the zebra pen to circle on the map where to drop the bomb. So I think Zebra G750 is responsible for me marking on the map where I wanted the bomb to go. I mean just absolutely crazy. And I think, yeah, this makes absolutely no sense. All right, listen, gentlemen, a great job. Philip, Boris And Spyros, you guys were awesome. Thank you for deep diving into all these topics. Star Cloud, Bedrock Robotics and Resolve AI. I wonder, are you gentlemen hiring? And if so, for what positions? I always like to give you the ability to make the pitch to come work at your companies. Who are you hiring for, Philip? And what's it like to work there?
- [67:54]
B
We are very much hiring. We are hiring across engineering, power electronics, mechanical engineering, thermal engineering, as well as some government relations, mission operations people. If people want to come and build Dyson spheres and matryoshka brains, they should come and work at StarCloud.
- [68:10]
A
Got it. Okay, Boris, your best pitch to come dig ditches with AI
- [68:18]
C
well, and a lot more and generalized to a lot of other tasks and machines. So hiring all over. It's a giant autonomy problem. So a lot of machine learning, simulation, infrastructure, hardware, also operations, a general counsel. So there's a lot of, a lot of interesting positions, but it's a fascinating problem with a lot of dimensions to
- [68:38]
A
it and a good time to get on the rocket ship because, hey, the company, you know, could go places. Your. Your stock options might be worth something. That's my editorialize. And Spyros, what are you hiring for and who's the culture like over there and what's it like to work there
- [68:51]
D
yet in person in San Francisco? And we're trying to change the way software works. We just recently announced our series A at a billion dollars. We're hiring essentially Resolve is building agents and we're trying to also collect the right data and train models for, let's say, running production software. We're hiring in infrastructure, let's say agentic workflows and post training, but also go to market. Right. Like Resolve is now working with some of the largest technology and financial services companies in the world. Companies like Doordas, Salesforce, et cetera. So we are looking for folks who want to maybe come and help us expand everywhere, especially in the U.S. all
- [69:29]
A
right, there you have it. And we will see you all next time on this week in AI. Go to thisweekinai AI and you'll find the YouTube channel, you'll find your Spotify links. You sign up for the daily email to get inside information on. We're gonna start writing profiles in our daily email of the next wave of AI companies companies here on the program. We've got like, you know, the killers who have all, you know, have established companies. But in the newsletter we're going to start covering the very early stage companies, basically these companies five years ago this week in AI. AI See you Next time, everybody.
- [70:02]
D
Bye.
- [70:02]
A
Bye.