← This Week in AI
Loading summary
Transcript151 lines
- [00:00]
Chase Lockemiller
In the future, my kids are going to have access to the workforce equivalent of millions of people worth of labor, just like at their fingertips, like on their phone. If you have incredibly high agency and curiosity to solve a problem that speaks to some other human, you're going to be able to just will that into existence.
- [00:19]
Jason Calacanis
The three or four who took to Openclaw first less 20 days, they became literally five times more valuable than the people who didn't.
- [00:26]
Anastasios Angelopoulos
Economic consequences of this technology are going to be large. I think all of us can agree. And if I were some somebody in any of those industries, I'd be thinking, how can I leverage it? Let me become the expert so that I'm at the bleeding edge of this technology.
- [00:41]
Chase Lockemiller
This Week in AI is brought to you by Notion. Bring all your notes, docs and projects
- [00:47]
Anastasios Angelopoulos
into one space that just works with AI built right in.
- [00:50]
Chase Lockemiller
Try notion with notion agent@notion.com twist and
- [00:55]
Naveen Rao
quadratic bringing the productivity boost of AI into your spreadsheets.
- [01:00]
Anastasios Angelopoulos
Visit quadratic AI twist to sign up
- [01:03]
Chase Lockemiller
and use the code Twist to get
- [01:05]
Naveen Rao
one month free of their pro tier subscription.
- [01:10]
Jason Calacanis
All right everybody, welcome back to this Week in AI. This is the new podcast from your bestie who started all in Podcast and this Week in Startups. I am absolutely addicted to talking to AI founders and building with these tools. I think it's going to be the greatest change in the world. So I said, you know, it'd be interesting if I could go all in on AI and bring together three of the top leaders each week in AI to talk about the week's news. And if you're new to the program, go to this weekinai AI and at the top you'll see a link to Spotify, to YouTube, to Apple Podcasts, and to our daily email newsletter. So, you know, you can go deep with me today on the program. Chase Lockemiller is here from Crusoe. You know them because they are one of the, if not the leading infrastructure companies building out this insane Chase Build out. I'm going to talk about it today, maybe just tell everybody a little bit about what you do at Crusoe.
- [02:11]
Chase Lockemiller
Thanks for having me, Jason. You know I'm the co founder and CEO at Crusoe. Crusoe is a vertically integrated AI infrastructure provider, which means we do everything from the hardware angle of building out this massive infrastructure boom to support the infrastructure of intelligence to the software aspects of orchestrating and running reliable large scale computing workloads on large clusters of GPUs and delivering managed services like Managed Kubernetes as well as further abstractions like our managed inference product that sells folks tokens. So we're sort of in the business of either renting data center space, renting clusters of GPUs or selling tokens.
- [02:54]
Jason Calacanis
And your biggest partnerships are?
- [02:57]
Chase Lockemiller
Our largest, most well known partnership is in Abilene, Texas where we're building a 1.2 gigawatt campus for Oracle and OpenAI. And that's I think the largest single cluster being operational and in development right now.
- [03:12]
Jason Calacanis
And that would be called Stargate I guess is the codename.
- [03:16]
Chase Lockemiller
That's what kids on the street are calling it.
- [03:18]
Jason Calacanis
Yeah, yeah, okay. And we'll get an update on how that's going. But it's incredible to have you here. Last time we saw each other we were both in Davos, so kumbaya. Davos. Had a good time running around there. Also with us, Naveen Rao, who is with unconventional AI and they're rethinking the computer from the ground up from first principles to be more power efficient, obviously for AI. So Naveen, explain what that means and what you're actually building.
- [03:47]
Naveen Rao
We're kind of thinking about the computer that we all know and love. That's something that's an 80 year old paradigm. You know, this is invented in 1940s to actually compute trajectory of artillery back then. And so we've gotten a lot of mileage out of this architecture. But now we're at this point where our application is intelligence. And how can we kind of rethink what the computing primitives really are for this application? If you think about what a neural network is, it actually just runs on the physics of our neurons. There's no numeric representation or floating points or anything like that in our brains. So can we start to build circuits that actually have similar kind of properties as those physics of neurons and actually build something that's not just 20% more power efficient, but orders of magnitude, like three orders of magnitude is what we're shooting for in the next five years.
- [04:38]
Jason Calacanis
How will that manifest and when will we see it?
- [04:41]
Naveen Rao
I guess, yeah, product wise. I mean this is a research project for now and it's products. We're definitely heading toward products. In four to five years it'll manifest as just greatly reducing cost. Basically the figure of merit we care about is joules per token. So you can think about it and I'm sure Chase can back me up on this. But when you build a data center it's about getting the power contract and then it's about how do I monetize Every one of those watts. And so we provide the path to greater monetization for each watt.
- [05:11]
Jason Calacanis
And you think you can do this 100, 1000, perhaps even more effectively than the current solutions.
- [05:20]
Naveen Rao
So there's actually some theoretical results from the 60s which suggest that we are maybe somewhere between 7 to 10 orders of magnitude away from actual physics of computing. So we think three orders of magnitude is something we can accomplish in the next few years. After that, we're going to have to rethink the actual fundamental substrate. Like, can we do this with electronics and silicon? I'm not sure. But what we're doing will open the door to new substrates, new kind of physical manifestations of computing.
- [05:48]
Jason Calacanis
It's unbelievable. I mean, Elon's always talking about, like, the amount of energy it takes to run our brains and how efficient they are.
- [05:54]
Naveen Rao
Exactly.
- [05:55]
Jason Calacanis
That's kind of. Is that the North Star, the human brain, when you build this new product?
- [06:01]
Naveen Rao
Yeah, we actually want to break past that, to be honest. So if you think about it, human brains are 20 watts. If I sum up 8 billion people times 20 watts is 160 gigawatts. The US has alone about 1000 gigawatts of. Of capacity in the world is about 9000. So we have a. We have a lot of energy in the world. We just have to use it more effectively. Just think about what we can do if we can harness that for intelligence in a synthetic world.
- [06:26]
Jason Calacanis
All right. And also joining us from arena, Anastasios Aja. Police.
- [06:33]
Anastasios Angelopoulos
Angelopoulos. Yeah, Angelopoulos.
- [06:35]
Jason Calacanis
Angelopoulos. Obviously Angelopoulos. My Greek brother.
- [06:40]
Anastasios Angelopoulos
My Greek brother, yes.
- [06:41]
Jason Calacanis
I mean, Calacanis is hard enough. Angelopoulos, actually, Angelop is not so bad. And you're obviously running Arena. And this is when people say in the arena ranking the AI models, that's actually your company. So this is incredible to have you on the program. And it makes sense to me that the arbiter of all language models and their performance would be Greek. That makes total sense. Since we created science, philosophy, theater, democracy, all these things for society, we might as well be the judge of how AI is doing is what is your business, though. We know you have the arena. People fight in the arena to move up the rankings. But where did this come from and what has it turned into in terms of a business?
- [07:32]
Anastasios Angelopoulos
The project started as a research project at Berkeley while we were doing our PhDs, Waylan I and many other students at Berkeley along with Yan. And we never really approach it with any commercial sort of ambitions at all. We just wanted to. It was at the time that chat GPT started like three years ago and we were all sitting around a table and thinking, hey, how do we evaluate the system? It's doing well on the benchmarks, but it's not doing well in real life. You try to chat with it and it's dumb. Not chatgpt but other models. And so that's how this arena paradigm emerged of hey, let's put it in front of users, have them chat with it. Two responses come up instead of one and then they can vote for the one they like better. And the platform has evolved quite a bit since then, starting from the early days and then suddenly we grew and grew and grew in terms of users until we were at Berkeley trying to run basically this industry scale project out of academic grants and support from labs and from VCs that were kind enough to donate money as gifts to try to keep this project running. And it had become the central force in the ecosystem for neutral, independent evaluation. And at some point we decided, how are we going to have the impact that we want in this world? How are we going to help measure intelligence and ensure that it's reliably and responsibly deployed? Can we do this through an academic setting? Can we do it through a nonprofit? And we decided no, because it would star a bit of resources. So we decided to start a company.
- [09:11]
Jason Calacanis
The customers are the language.
- [09:13]
Anastasios Angelopoulos
And since then the company's grown quite
- [09:14]
Jason Calacanis
a bit independent AI executives, what's the product?
- [09:26]
Anastasios Angelopoulos
And traverse the trade offs between performance, latency, cost and so on.
- [09:35]
Naveen Rao
All right, Anastasios, actually a question. How many users were on El Marina? We were big users of IT at Databricks when we built our models. Yeah, as you as you know, just kind of curious how many people were actually using it, providing feedback sort of on a daily basis.
- [09:51]
Jason Calacanis
If you want to be a data driven founder, and trust me, you do, you're going to spend time in spreadsheets. But a lot of these spreadsheets, they're just stuck in the 1990s. Thankfully. Now there's Quadratic. Finally bring in the productivity boost of AI into your spreadsheets. But this is not the same little chatbot in the corner who answers random questions. No, it's an AI native platform that handles all the number crunching and organization for you. You just describe what you want to do with your data and Quadratic makes it happen right there in the spreadsheet. Now you can get instant insights about your business without fighting formulas and trying to remember different ways to where to put the comma and the Parentheses it just works and it instantly shares the results with your collab. No setup or payments are required up front. So you can start using Quadratic right now and experience the magic. Just visit Quadratic AI Twist to sign up and use the code Twist to get one month free of their pro tier subscription. That's Q U A D R A T I C A I slash Twist.
- [10:51]
Anastasios Angelopoulos
Well, these days it's tens of millions. We have tens of millions of mao and we have like on the order of half a billion or more conversations. And so it's funny at the time that we were sort of chatting with you guys at Databricks in the early days of arena, in the very early stage arena, we were actually going around Berkeley and giving people Amazon gift cards to vote. Is the platform was that small? It really. And it was really supposed to just be an academic paper. And then it sort of eventually spun out. I think probably with. Basically we just got lucky that the space got so competitive and people were looking for a battleground. And then we became that default battleground and we took that and sort of ran with it and said, okay, how can we continue to retain these users and how can we, you know, retain the users that are giving us the valuable feedback that's going to drive our understanding of real world performance? And so that's been sort of our North Star. That's why our user base is heavily skewed towards the prosumers. It's about 28% software engineers.
- [11:56]
Naveen Rao
Well, I noticed, I mean you guys would actually provide that way of testing models out before they were released. So you sort of get some sense of where that model will be. And it was a secret model. Right? We kind of did that with you guys as well.
- [12:08]
Anastasios Angelopoulos
Absolutely.
- [12:09]
Jason Calacanis
Topic 1. This will fit right in your wheelhouse anesthesis. We have looked at openclaw now becoming. This is all like in the same week. Openclaw goes to number one in terms of stars. It's kind of like votes or follows on GitHub. Additionally, I don't know if you guys saw this Quinn 3.5 small model. This is Alibaba's open source. It was. It's running on a. You know, I think this is an iPhone 17 Pro. Now we're starting and we were wondering when this would happen. Here it is, it's happening. And this is obviously on airplane mode. So this is just running in the background and reflection. AI they do open weight models there. They've hit reportedly a $20 billion valuation and the world has just become enamored with OpenAI to the point at which if you run it like we are, you eventually, Chase, get to the point where you're like, I need to run my own silicon and I need to run an open source model because it's too expensive, say to use Claude or ChatGPT. We were trending, I think with 15 people getting on it towards 500 bucks a day each, $7,000 a day, $2 million in tokens. And this is like in month one of the project, I think we literally would have exceeded the salaries of everybody. So Chase, what are you seeing in terms of the open source models and their performance and their adoption?
- [13:48]
Chase Lockemiller
I think there's this, you know, kind of constant debate between open source versus closed source and you know, you know, it's a matter of like, who's going to win. And I don't really view it that way. I think there's like sort of room for, you know, multiple, you know, multiple solutions in the future. I think we're going to have a future of both, you know, a lot of closed source and a lot of open source. It is very exciting to see a lot of the developments in open source and you know, we are certainly seeing a lot of adoption from a lot of our customers in Crusoe Cloud that, you know, are AI native applications that are building and scaling and running a lot of inference workloads on open source models. And you know, I think it's a very exciting, it's a very exciting and innovative ecosystem that's sort of taking hold. And I think a lot of the efficiency gains that you're seeing and how people are able to distill the knowledge of these larger models into smaller mini models that can run like you demonstrated on device. I think that's a very fascinating trend, you know, and sort of the quality, the outputs are still quite good. So, you know, I think all this goes towards this broader trend though is like, you know, models are getting smarter, they're getting more useful. You know, things like openclaw are actually, you know, providing, you know, these, you know, agentic platforms to go do stuff for you. Right. I mean, it's like, you know, I was talking to someone yesterday that runs a very large investment manager and he was saying how he's like gone fully all in on openclaw and he's like, you know, running all this different stuff and he's like the, he's like, I'm convinced I'm calling this AGI. He's like, he's like the only way I like, you know, I like I.
- [15:44]
Jason Calacanis
Do you think it's AGI, Chase, or does it feel like this is the year we hit AGI, I guess is another.
- [15:52]
Chase Lockemiller
I think there's like a move, you know, it's kind of like, you know, Anastasios is kind of the king of benchmarks over here, but, you know, it does feel like one of these like moving benchmarks that like, as it gets better, it gets more impressive. We sort of like move the goalposts a bit. You know, I kind of had this personal definition. Like, having come from more of a math background, I felt like to me, one of the big tests is going to be when there's this set of very difficult math problems called the Millennium Problems that are hosted by the Clay Mathematics Institute. And to me the breakthrough is going to be there's a million dollar bounty. One of them has been solved. Of the seven, the guy actually never even collected the bounty. You know, to me the, the definition is like when AI is able to solve one of those problems, that's like
- [16:47]
Jason Calacanis
Anastasios, this is a problem that humans have not solved.
- [16:51]
Chase Lockemiller
Correct.
- [16:52]
Jason Calacanis
So artificial general intelligence.
- [16:53]
Chase Lockemiller
So maybe that's like, maybe that's super intelligent because it's exceeding. Sure, sure, sure, sure, ground us here.
- [17:00]
Jason Calacanis
Because I think you're making the point, Chase, is that we are, we're like in the red zone. We're within 20 yards of the, of the end zone and we're like, you know what, let's push it another 10 yards. 120 yard touchdown Anastasios. How do you look at it defining for us here?
- [17:16]
Anastasios Angelopoulos
Yeah, I mean, so it's very funny. I had actually a similar bet with a friend of mine who's a professor at Stanford Stats that we made the bet just, it was a few months ago. So, you know, I think things are moving in my Favor that in 10 years, two of these problems will have been solved autonomously by AI. The reason why it's important that these problems, you know, to sort of chase this point are unsolved by humans is that once they've been solved and the information's out there, they're no longer good benchmarks because they overfit. And so I do think that there's a component of this that's when we create benchmarks for AGI or what have you, or just performance in math and coding and instruction following and production settings. We need to make sure that the data is constantly fresh to avoid overfitting.
- [18:05]
Jason Calacanis
What do you define it as then? Do you have like a personal definition you like and I'll go to you because I know you have a lot of thoughts here too.
- [18:12]
Anastasios Angelopoulos
Yeah, I honestly don't think about the definition of AGI much because I think that the, the basically all these debates reduce down to what's your definition? Question mark. And so to me it's much more about how do we ensure that when we deploy the system we can do so reliably and responsibly that we're helping people, not hurting people, and use it as a, as a tool. And I just, you know, the AGI aspect of things is not really part of my daily life in terms of my thinking. However, what is, is that I think that a large fraction of human labor is going to be commoditized by, by AI and that's going to have very wide reaching economic impacts and probably require a new economic system to deal with it.
- [18:57]
Jason Calacanis
All right, Naveen, I think we just hit on all seven possible issues on the docket. Let me get Naveen involved because this is kind of what happens when you have people who are in it every day. These things are turning into a flywheel in our minds. You have open claw driving Jevons Paradox, you know, then you have Goodhart's Law in action. It's. It's all kind of happening at once. So just to ground us again, AGI and then what's happening inside of enterprises and what you think of these open source models. Take it where you want to go, Naveen.
- [19:31]
Naveen Rao
Well, okay, I think there's a few things here. Talking about the definition. I mean, Chase is talking about one specific one where I would say it's kind of what we think is what smart people are. It's solving these kinds of mathematical problems that's kind of intrinsically difficult for humans. And there's only a small subset of humans that can actually do these things. What I actually think is even more interesting is that we haven't solved problems that every human and actually many animals can solve still. This is why this definition keeps sliding around because it's a little like porn. You know it when you see it kind of a thing. Many times we build a model and it solves some benchmark as Anastasia was talking about, and maybe it even solves some really hard problem, but then it falls flat on some really stupid thing that everybody can do. So then you're kind of like, is that really AGI? Right? So I think we're in this place where we're creating a new kind of intelligence that solves certain things set of problems. Now, every computer has existed for A purpose. And there's a lot of economic value, regardless of whether we call it AGI or not. But I'm a motor control neuroscientist. I studied how brains can actually solve physical problems in the world. I think physical problems in some ways are very hard, they're very high dimensional and there's so many sources of error and noise that come into play and these models do not do well in those domains.
- [20:54]
Jason Calacanis
Means 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 Notions, custom agents, busy work that used to take hours or never got done at all runs itself. And my company launch is run entirely on Notion. It is the single source of truth for our company. And custom agents are now here and they can specialize anything you can come up with. Unlike AI chatbots, which, you know, they just stick around, they wait for you to ask a question. You can assign your custom agent to ongoing tasks so you can have them do regular updates across the entire system, like, hey, alert me to what people are working on or keep track of the docket for the next podcast episode and even checking when I got new ads to read. We're leveling up the this Week in AI newsletter and I assigned hire a new writer to our task routing agent. Now Oliver has started this project the moment he opens his computer in the morning. And now producer Oliver just starts that project the second he logs into his computer in the morning. Can you imagine an AI to keep an eye on what everybody's doing, not in a creepy way, but in a productive way, like a coach and dropping a new task to the next available person. That's what this is all about. Try custom agents now@notion.com twist. That's all lowercase letters. Notion.com twist to try custom agents today. And when you use our link, you know you're supporting our show notion.com twist. Right? We're still struggling with self driving edge cases.
- [22:37]
Naveen Rao
Yeah.
- [22:37]
Jason Calacanis
You know, here we are 25 years, what is it? 20, 25 years after the DARPA challenge, like two decades of work on it, still lots of work to be done. That's right. And then in terms of open source's role and how small these get, because you're working on a whole new paradigm. Right. Chase is working on scaling this, finding new energy sources. Obviously Jensen's working on next level chips and he bought rock. And at the Same time the models are getting smaller. And then Apple Silicon, which Steve Jobs conceived of in 2008, executed on, in 2018 when the first, I guess M1 and some of these things started to come out. What an amazing prescient thing. Obviously Steve Jobs, he figures these things out early and now we have people saying, you know what, I'll just run some stuff local. And we're going to have this crazy world where things are working on multiple platforms, chase humans.
- [23:39]
Chase Lockemiller
We sort of have this like system one, system two, like thought process. Like you have sort of a, you know, really fast system that you know, is very intuitive. And then you have like your system two, like higher, higher dimensional reasoning system. And I think, you know, in, because you know, you made the comment that like models are getting smaller, they're simultaneously getting smaller and bigger, right? We have things that are bigger, more capable models that are these higher level reasoning models. And then we also have things that we're able to get better accuracy, better output, sharper intelligence with smaller distilled models. And I think what's important for that is like when you think about, you know, efficiency of, you know, how intelligence not only gets created but actually gets distributed to the world, it's really important to think about like where does the compute actually happen? And you know, that's, that's part of you, you know, you know, the point you were making around like, wow, these things, you know, this, this, this new version of Quen can run on an iPhone in airplane mode and wow, that. Holy shit. That's pretty cool. Excuse me.
- [24:44]
Naveen Rao
Wow, that's pretty cool. Actually. I kind of, I'll sort of go ahead a little bit here. Is that. So I think the bigger models, what we found over time now we're still learning how intelligence works. And these bigger models are very good at retrieving lots of, of different kinds of information. So pre training and adding more parameters basically gave us more access to information. Now I think we're moving this realm. Like you said, bigger models are reasoning models. I actually think we're moving to a realm where potentially we can separate out the reasoning capability at least partially from the retrieval part of the system. And when you do that, you actually might be able to build something that has the reasoning part that's small and then you have give it access to data in different ways. That's kind of a paradigm I think is a little bit more like biology frankly. Like our brains don't necessarily remember tons and tons of really specific information, but we have mental models that we can reason through new Data presented to us.
- [25:38]
Jason Calacanis
Interesting.
- [25:39]
Naveen Rao
Potentially where we can make it much smaller. Right.
- [25:42]
Anastasios Angelopoulos
I actually have a question Chase about the world of data centers in the future because there is this, you know, as we're talking about this kind of divide between like this on device, whatever, blah blah, blah, then there's maybe going to be massive training runs. Where do you see the data centers that you're building fitting into like that world?
- [26:03]
Chase Lockemiller
My vision for the future of like compute and data center infrastructure is sort of this bimodal distribution where you end up with these mega campuses. Like what we're doing in Abilene, Texas. You know, we have another campus we announced in Wyoming. We have a couple of other like gigawatt scale campuses that we're building that are really about enabling, you know, training the next breakthrough foundational model. These are you know, you know, hundreds of thousands, soon to be millions of GPUs interconnected on the same high performance back end network and all acting as one cohesive giant computer. Right. And what that's useful for is you know, kind of again that those next breakthrough models, training and then also a lot of this test, test time, compute, scaling, a lot of these reasoning models that you know, are given a very complex task that, that need to be thought through, that need to think more, more deeply about sort of an answer. We also have a product called Crusoe Spark. And Crusoe Spark is a fully self contained modular AI factory. We have a 500kW unit, we have a 1 megawatt unit that is frankly manufactured in a factory, you know, owned by Crusoe. And those can be rapidly deployed anywhere, you know, where we can access power. It can be on site in a parking lot for you know, an autonomous factory. It can be you know, at a, at a location where you know, we
- [27:32]
Jason Calacanis
actually have wild looking. So it's, it is basically like a trailer or like a mobile.
- [27:38]
Chase Lockemiller
Exactly, it's a modular building and that you know, you can actually see we, we connected two together there so that you know, they're on sort of a shared fabric. That's what that you know, connection there is, is actually networking armada a little bit. Yeah, it's I guess, I guess similar. You know, I think they're, they're more focused on like you know, military applications and you know, those things being deployed in the field. But that, that use case, that picture you just showed, that was actually a fully off grid solar plus batteries that we were able to power those, those modular AI data centers. And you know, I sort of view this as like we're going to have swarms of these things, right. And you know, and I think one of the power issues that you have is that you know, it's, it's hard to go on the grid and be able to get a gigawatt of power. Right. So going to a utility and say hey, I need a gigawatt at a single location. They're going to be like okay, how's 2037 sound? And you know, but, but if you go to a utility and you say hey, I need 20 megawatts, they'll be like hey, there's actually like 100 locations where 20 megawatts of power would actually be useful to us because of transmission constraints, because of, you know, the way they're load balancing the grid. And if you can have like a fleet of these small modular data centers that get deployed in those regions, it actually unlocks a lot of underutilized power. So that's one great use case.
- [29:06]
Jason Calacanis
And so just that not every chase.
- [29:09]
Chase Lockemiller
Yeah, not, not every.
- [29:10]
Jason Calacanis
Instead of chasing, just so people understand, instead of like the power chasing the data center, the data center is autonomously driving around to find the power.
- [29:23]
Chase Lockemiller
Oh no, it's not driving around but I know, yeah, yeah.
- [29:26]
Jason Calacanis
Slightly facetious here but the data center will go find a way it goes. Yeah, sniffs out, oh, here's some power. Let me put this by this dam or hey, we found oil in this pocket, you know.
- [29:39]
Chase Lockemiller
Yeah, totally. It's incredible. You know, it's really intelligence everywhere and especially for inference scaling and inference workloads. You don't need a million cluster GPU cluster to run an inference workload. You're very happy with a rack or you know, a couple of GPUs. Sometimes there's you know, multi node inference workloads. Naveen knows about these very deeply. But you know, and, and you know one of the, one of the use cases I love about like the, when I think about the future is like as humanoid robotics and robotics scale, you know, I think about these factories that are going to be in the future and you know, the reindustrialization of the United States. A lot of those factories, they actually want to move a lot of the reasoning computing off device because it limits the battery life to power the electrical mechanical systems and you'll have kind of like a system one smart, kind of like it can move it around but then it's offloading a lot of the spatial reasoning to a locally run data center that's operating a factory of humanoid robots that are building some widget and I think that's a pretty cool thing to think about. Very cool.
- [30:48]
Jason Calacanis
That's fascinating.
- [30:49]
Naveen Rao
Yeah, it's interesting. I'm giving a talk tomorrow and what you just described, there's a slide on it. It's kind of funny because when we're successful and greatly reducing the power of inference, it'll actually result in way more smaller data centers. You're just going to have them everywhere. You're basically going to fill every niche and you don't have to transmit power. Right. You don't have all that loss. Right. I'm sure you know the numbers off the top of your head. But like transmitting power is hugely costly. It's a very expensive operation.
- [31:17]
Chase Lockemiller
So it's also like massive lead times because you need these, these huge right aways to like build like seven, you know, high, ultra high voltage like transmission lines. Like you're building giant power lines. You need everybody's approval to make that happen. That' like, you know, not, not an easy task.
- [31:34]
Jason Calacanis
So Anastasios, if we were to think about this with the Optimus, right, this is something Elon's been talking about a lot is like hey, what's the battery power going to be here? And then what's the GPU in it? Because now you have to serve two masters. The thing has to be able to move. That requires energy and it needs to be able to think. That requires energy. So now we're thinking, hey, if we're going to do a bimodal breakdown, maybe it's got some ability to operate in the world with a model physically attached to it. But in some use cases you may have it on Mars and it needs to have the whole kit and caboodle and be offline and be in airplane mode. How do you think about how these models might become bimodal and live in multiple places for inference, for training the model and then local because you want privacy. And now you've got this, you know, breaking it up by what does the actual physical nature of the self driving car or robot need?
- [32:34]
Anastasios Angelopoulos
Well, I think the system 1, system 2 analogy is probably the right one in the sense that you know, there's just to reiterate what was previously said, there's a well known sort of cognitive system that humans use when thinking which is that when they need to make quick decisions that are not so high stakes and maybe easier, you just sort of immediately react to that system one thinking and then system two thinking. Is that more deeper reasoning. And of course it's somewhat asynchronous. It can happen over Hours. It can happen over days, weeks. And so you might be able to potentially offload that to a different system that's, you know, a different brain or data center somewhere else. That's, that's thinking about it. But you know, with the speed that things are moving for systems like an Optimus, my bet would be that in the, you know, medium term almost all of that is going to be handled by a very small model that's just running on the robot. If you look at for example like Quinn with these recent like small model releases, the rate of improvement and the rate of compression is so high. Like Quinn 3.5 27B, the like medium scale model, it has performance on par with the 10x larger last generation model, the Quen 235B, at least on our leaderboard. Same performance, approximately 10x smaller model, few few months difference in release date.
- [34:21]
Jason Calacanis
Extrapolate that out. Where are we going to be at the end of the year, end of next year in terms of running these models?
- [34:28]
Anastasios Angelopoulos
Well, I mean, I'm not sure how much extrapolation I even need to do. It's moving faster than we can even predict. Like these small Quin models that were released yesterday, these four B9B models, the AT least, you know, they haven't been evaluated by an external entity. So we don't yet know what the real performance is. But based on the benchmarks that they self reported, it again happened that those models are on par with and maybe even slightly better than the 27B version from the, from the previous generation. So it's moving so fast that within the year we could have, you know, models that are a billion parameters or hundreds of millions of parameters and have the same performance as a model that last year might have been hundreds of billions of parameters.
- [35:15]
Naveen Rao
It's kind of interesting because just talking through this, it just became apparent that this kind of worked in inverse to biology in a way. Like biology sort of started small, figured out some basic principles and those principles scaled so the efficiency came first. What we've done is actually the inverse, we've brute forced our way through it, throwing everything we possibly could and now we're understanding, oh, actually I didn't need to do all that. There's a lot of things I could keep chipping away at and I can actually go smaller and smaller and smaller. So I mean, just to kind of put in perspective like biology through this process of kind of the bottoms up like a squirrel runs on 10 milliwatts of energy. Your cell phone runs on about 1 watt, 10 milliwatts and it can do things at precision levels that we cannot do in a megawatt. Like, I can't make a robot jump between branches in the wind and hit the branch perfectly. A thousand times out of a thousand. Right now, I can't do it.
- [36:09]
Jason Calacanis
But biology's had billions of years, I guess, to get this right. And we've done this all in 50 years of modern compute.
- [36:20]
Chase Lockemiller
I guess a lot of this, Jason, it kind of reminds me you opened up talking about your team using OpenClaw and running up this massive bill and saying, this is going to be our whole payroll. It's like, just spend on OpenClaw. And I think there's an important thing to remind yourself of during this moment is that there are so many exponentials happening, whether it's models getting exponentially better. Anastasio's just talked about a model one tenth the size getting similar performance only a few months later. Compute is getting exponentially better throughout. Just sort of the classical trend that, you know, we're going down with GPUs as well as hopefully step function breakthroughs and things like what Naveen's working on. You know, power systems are getting more. More efficient. So you just always have to remind yourself that this is the worst these things are ever going to be. It's the most expensive intelligence ever going to be. It's only going one direction and that's down. And, you know, it's. It's. The cost is only going down and the performance is only going up.
- [37:20]
Jason Calacanis
Okay, so that leads us to the obvious next question. Jevons paradox. Openclaw seems to have opened up collectively in 30 days. We're sitting here, like, literally 30 days into this kind of expanding and actually hitting pop culture. I don't know if any of you have had a family member say, like, hey, tell me about Open Claw. Can I set this up for my company? I'm assuming I see nodding across the board, my wife's like, how do I get Open Claw? And you know, and I'm like, okay, here we go. So I hooked up openclaw to Instacart, and I now have my opencart going, my openclaw going into opencart saying, hey, from H Mart, which is like the Asian Korean market, and looking at the last six orders, figuring out what we ordered the most, and then preparing a basket, and then my wife approves the basket. Then the week after that, we're going to go for just order a basket and add three items you think we'd enjoy, and we'll give you feedback on it. So this could change everything. This is hours of work a week is like feeding a family. Putting that aside, it seems to have cracked open the floodgates in token usage. So then we need to ask the question and pair that this incredible open source project opening up. I don't know, maybe 5% of the workforce is into this right now in America, maybe less, maybe it's 2%. But let's pick a number. 5% of people are really into this. It's going to max out what's available. It's already giving challenges to Claude and they're having to time things out and block your usage of it. If you're on a max plan, push it towards the API, et cetera. And then you dovetail that with employment. And I'm putting it conservatively at making our employees 10 to 20%, offloading 10% conservatively a month of their work right now. Now I don't know when it hits diminishing returns and flattens out, maybe it's 50, 60% of their chores, but it does feel like then what we saw from Jack last week at Block, formerly Square, laying off some percentage of the employee base. Amazon saying they're going to lay off continuing laying off and not higher. So how do we think about it's 40%, I think of the workforce Jack did. Obviously he was probably a little bit bloated post Covid, but how do we think about the tension here and employment in knowledge work?
- [39:51]
Naveen Rao
The block one I have some insight on. I was on the board of Square Block right as Covid was hitting. I think there was definitely a bit of a overhiring. It's like woohoo. Now we can move everything to Slack and you know, all the electronic media and I can just hire like mad. And I think that's what happened. I'm sure there is automation, but I think what I see over and over again is that it's not that people don't want great engineers that are building. I think this becomes an excuse to kind of cut the fat that accumulated for whatever structural reason. And what I'm seeing more of is not that hey, I want to see fewer people doing the same amount of work. It's that I want to see the same amount of people doing more work. I want to see more output, I want to see more software releases, I want to see more features. I want to see because my customer base is going to expect more. It's not that you're going to say, oh, I'm just going to. I'M happy with my shitty 2 year release cycle of enterprise software. It's like no, I want you to fix the bug now. I want you to update this thing now.
- [40:56]
Jason Calacanis
Okay. Anastasia, what do you think? Are you doomer at this point and do you flip flop like I do from oh my God, there's so many problems to solve. Yeah, just we'll create more startups or create more. And then sometimes I'm like I wonder what young people are going to do because this is the first 10 or 15 years of your career is doing this grunt work. Where do you sit on this? Is it making you stay up at night?
- [41:19]
Anastasios Angelopoulos
Well, maybe. Like many founders, I tend to be optimistic about everything. So I think the future is very bright. It's filled with plenty that we're going to solve problems that we never thought that we would be able to solve in our lifetimes. They're going to be solved by AI that it's going to make us so much more productive, free up our time to do things that we really care about doing. At the same time we have to be honest with ourselves that the majority of economically valuable work in this country and in the world is done by training people to do repeatable tasks. Okay, today that's the foundation of work Today that is the foundation of work today. Training people to do repeatable tasks. Whether that's working in a factory, whether that's delivering food, whether that's writing software and a lot of that repeatable work is no longer going to, we're no longer going to consider that work moving forward. And so the question is then what remains? Maybe there's some use for humans in the coming decades for doing work that isn't so repeatable. Maybe requires some more creativity or context that's difficult for the models to understand. But I do think that we are going to enter a crisis of labor where you know, the, the so called where basically labor does not equal value anymore and that you know, something else will be potentially valuable that humans will be able to do and a lot of the what we now call labor will be done in you know, Crusoe's data centers or done on, you know, Naveen's chips and that humans will play a different role in society.
- [43:16]
Jason Calacanis
Chase, do you, where do you fall on it now and do you honestly worry about it a bit? Does it keep you up at night or you just think human creativity is always going to figure out a way and we're going to move towards a Star Trek world where work and passion are optional and there's so much abundance. Nobody would worry about something like food, energy, water, school, et cetera. It's all free anyway.
- [43:39]
Chase Lockemiller
You know, I sort of, I'm definitely the optimist in how I view the future. And to me AI and the harnessing of this digitally native intelligence and labor is going to be the greatest catalyst for energy abundance, for abundance of intelligence and for driving economic growth. And if you look at it just through the lens of economics, I think what's fascinating is for the first time in history we're able to manufacture intelligence, right? That's like the, that is the breakthrough that we're sort of seeing. And you know, the units of intelligence that we're seeing are these tokens, right? These, these, these output tokens of querying, you know, large language models or you know, these big, big AI models. I think Naveen was actually the first person that gave me this, this, this analogy that, you know, an LLM is a bit like a statistical database where you sort of, you know, it's just doing a next word prediction in terms of like, you know, you, you, you send a set of input prompts and then it sort of gives, you know, these, these, these output tokens. But I think what's FAS, what's happening with OpenClaw, with Cloud Code, with a lot of these agentic platforms that are actually wrapping those models and actually utilizing, you know, how tokens are generated to actually create a useful task actually producing digital labor, not just digital intelligence. You know, it's like you give it a task and it goes and accomplishes that task. That is another step, function, breakthrough. And when you look at like, you know, I think the primary KPI that we sort of look at in terms of like economic prosperity, it's gdp. And when you, you know, going back to your introduction to macroeconomics class, you know, the growth in GDP is really the sum of three things. It's the sum of change in capital, change in labor and change in technology. And when you look at this massive acceleration we're seeing, you know, obviously it's, it's a huge delta in technology and the efficiencies we're able to gain. But for the first time history, we're actually able to massively accelerate that delta l the change in labor with this new digital labor force that's creating massive growth in what we're able to do and what we're able to accomplish. So I think we're going to see GDP growth that we've never seen before because you've been able to amplify it through digital agentic labor.
- [46:06]
Jason Calacanis
Yeah, I, I was talking to my team and I said, you know, if we ever hire people, I think we'd be hiring people for the company as like a luxury. Almost. Like we just want to have more people around because it would be fun to have a new person hanging out with us.
- [46:22]
Chase Lockemiller
And then I was thinking, your interview's based on vibes. You know, that's.
- [46:26]
Jason Calacanis
Yeah, just like, you know, we'd like to have another member join the Grateful Dead. And yeah, they weren't going to play bongos. We would like to have a sax player in. And Dire Straits went from four or five people and then Mark Knoffer's like, what if we had two pianos? What if we had a saxophone in here? And some of the great Dire Straits songs had a saxophone.
- [46:45]
Chase Lockemiller
You just added.
- [46:47]
Jason Calacanis
Wasn't like an economic decision, it was an artistic one. And I was trying to fill the metaphor and I'm a foodie and I was just thinking, can you imagine taking somebody from 200 years ago in an agricultural environment and then taking them on a food tour of, you know, like a modern supermarket? Erewhon H E B the what's the place in Barcelona? The market, you know, or skiji fish market. You just take them to these markets and show them what's available and they would be like, why? And then take them to a bunch of Michelin star restaurants and force them to go through a 20, you know, coarse meal, you know, one little, you know, 50 calorie, 150 calorie morsel at a time. They would be like, what are you doing? This is such wasted effort. You don't need this many types of potato chips or cereal. Now. Take us as a society and you could probably do that with entertainment choices. You take somebody from 1950 and having watched TV that was on for four hours a day and the rest was just an American flag testing signal and put them in front of the Internet and YouTube and just watch their brain explode. Like, why does this exist? So if we start to think about that for what's about to happen, Naveen, I mean, what could society look like in a. And I guess then for the people who are on the bottom in these up and coming jobs and knowledge workers coming out of college, I think there has to be some message for them in terms of what would you advise? What do we advise? Anastasios of the Berkeley student coming in now you're going to be 400k in debt or 250k in debt. Here's what you do when you graduate. So maybe answer both questions, the abundance one and the oh my God one.
- [48:36]
Naveen Rao
Well, it's interesting. So you got to really kind of click into what economic value is. I think you kind of went there. It's sort of like, oh, it's vibes or I want another guy in the band. Well, that is. It's a human construct, right? Like, these things are driven by things that we want. We're moving away from a world where we buy things maybe that we have to have. Like, our subsistence will be covered, right? We will have food, we will have plenty of it. We will have electricity, we'll have shelter. Those things are going to be covered. But we still want to. We move resources around based upon what we want. And that becomes. As long as humans are still in control of those resources, if that changes, things are to going get pretty funky. But for now, let's assume that humans are in control of the spending. I actually think we end up in this world where we actually start to reject some of these digital artifacts. I actually had someone. It was really funny. I had a slide deck put together. And when you put your slide decks together and you draw some figures, you sometimes are a little misaligned in the lines and you get a little overlap. They're like, that's like an organic slide deck. You made that? I was like, damn right I made that.
- [49:38]
Jason Calacanis
That was bespoke.
- [49:39]
Naveen Rao
It's handmade.
- [49:40]
Jason Calacanis
It's acoustic.
- [49:43]
Naveen Rao
Exactly right.
- [49:43]
Chase Lockemiller
Artisan.
- [49:44]
Naveen Rao
It's artisan. It's like handcrafted, you know. And I think we actually call back that. Because when I did that, the guy who looks at it is like, oh yeah, I've done that before. And it kind of relates to an experience, the human thing. And that's why we like imperfection. And so I think that drives demand. And we can't forget that it's not economically valuable. Work is not just productivity of churning out more shit. It right? It's about what we want to buy. And I think there's that. You got to like, think about the supply of it and the demand of it. And I honestly, that's why I'm super skeptical of all these doomer scenarios. I don't think we go to this place where everyone sits there like a zombie. Like, I'm going to be like, screw that. I want to buy something that I can relate to. You know, that's where I'm going to spend my money.
- [50:28]
Chase Lockemiller
So anyway, I think, Jason, I loved your analogy because, you know, the transporting someone from 200 years ago and allowing them to see the industrialization of, you know, food, of you know, just supply chains and you know, the abundance and prosperity and you know, honestly the volume of people that that's able to support these days. When you look at like the percentage of humanity that used to be involved in agriculture versus today, it's like, you know, I don't know the exact numbers but it's you know, some very, very high percentage to some very, very low percentage. And you know, the amount of leverage we're able to get on human time. I think that's what we're seeing right now with this industrialization of intelligence and industrialization of labor, with this digital workforce that's going to be created and this booming that's going to unfold. When you asked about how you would advise kids coming out of college or even parents raising kids, I think is an important thing. I'm a father, I have young kids and when I sort of think about Naveen's got both ends of the spectrum, both young and older. But I think what's important when I think about my own kids, I think the things that I'm optimizing for are one, a high sense of curiosity, being able to take a thread and really pull on that and ask the next question and really just like get deeply curious about something. And and number two is actually a very high sense of agency which like sort of goes in line with that, you know, sense of curiosity because in the future, right, like my kids are going to have access to the workforce equivalent equivalent of millions of people worth of labor just like at their fingertips like on their phone. And you know, if you have incredibly high agency and curiosity to solve a problem that that you know, speaks to some other human, you're going to be able to just will that into existence and you're going to be able to, you know, get more leverage than you've ever had in the history of humanity. And you know, I think that's going to be a very unique and cool and abundant future that you know, my kids are, you know, it's a great pairing.
- [52:46]
Jason Calacanis
Curiosity agency. I'm going with. I have a 16 year old and two 10 year olds I'm going with and I think we'll just go around the horn here in Anastasia. So I want to get your advice to students and I don't know if you have kids. You have kids? No, no kids yet. Okay. So just you have students, you have an unlimited you could give advice to in the university. I'm going with radical self reliance and then the ability to communicate and socialize, you know, with other humans and the ability to learn new skills like problem solve. Learn new skills. Because what I'm seeing is the people in my own organization have a lot of young people, the three or four who took to Openclaw first, in 30 days, less 20 days, they became literally five times more valuable than the people who didn't. And Anastasia, at that point I said I'm calling, you know, a red, a red notice here. Like I'm calling a red, what do they call it?
- [53:44]
Anastasios Angelopoulos
Code red.
- [53:45]
Jason Calacanis
A code red. I'm calling code red. This Sunday, the five people who know how to do this, Oliver, Lucas, train, everybody else, 15 people showed up and now we have the whole organization. So Anastasia, what do you think is the best advice for young people who might be scared right now or people who are the Uber driver or work in a factory or paralegal? Like, what should they do? What should they do?
- [54:10]
Anastasios Angelopoulos
It's a great question. The economic consequences of this technology are going to be large. I think all of us can agree with that and I think it would be, it's basically naive to ignore it. And if I were somebody in any of those industries, I'd be thinking, how can I leverage it? Let me become the expert so that I'm at the bleeding edge of this technology so that I can help shape how it's used so that I can be helping the economy modernize. Because if you don't, I think that it's likely people will be left in the dust.
- [54:44]
Jason Calacanis
So if you are the Uber driver, going and figuring out how self driving cars are trained, going and figuring out how these new depots are going to work, how the new shepherds of 10 cars are going to work.
- [54:55]
Anastasios Angelopoulos
Yeah, how is this going to change? You know, let's say I'm a doordash dasher. What would be going through my head is okay, it's clear that the self driving part is going to be solved, right? There's going to be self driving cars that can deliver xyz. So what's the utility of what. So basically like what's the next evolution of that market? Is it going to be that you need somebody to coordinate with the store? Is that, is there sort of a last few steps problem of how who's going to actually deliver the food to the door is that, you know, where is it that humans are going to be fitting into that process and how can I sort of accelerate and take part in that next phase of the system?
- [55:39]
Jason Calacanis
This is such a great insight. We literally invested in a company called Auto Lane. Their premise, the Target parking lot is becoming like an airport. You have Waymos dropping people off. You have Zipline coming and picking up packages from drones. So their air traffic control is there pitch. And the idea is to have people, you're going to need people to do that last mile, as you pointed out. Okay, here's a mall, the Dominion here in Austin. There's you know, 150 stores. Somebody has to coordinate those 150 stores. And me sending my Tesla, my personal Tesla to go pick something up after it drops me off at work, there's going to be something there or going, working in a cloud kitchen and creating the niche food that doesn't exist yet in this suburb that has no food options because it's in a food desert. Should people learn to code? This is I think the big question. And we had the stock market. I mean, God, the news cycle is so crazy. Which is why I started this podcast two weeks ago. The entire SaaS space had a SaaS apocalypse. Everybody thought this is the end of days. Salesforce is over, HubSpot's over, IBM's over. All the stocks got hit. And we have this big question, Naveen, should you learn to code? And is. And one of the great breakout moments I think Anybody who uses OpenClaw has is when OpenClaw says, oh, I don't have said to me, I don't have a CRM or I don't have the ability to. I can't find a service to download YouTube videos for you. Shall I make one? And the shall I make you a CRM since you don't have one on this computer? Or shall I make you a video tool that rips TikTok and Twitter and YouTube videos instead of using this Russian website with spam spyware on it? That's like a mind blowing experience. Should people learn to code? And is everybody now a developer?
- [57:44]
Naveen Rao
Well, I think both of those things can be true. Everyone's a developer in a sense. But I actually think the ones who still understand the machine are going to have an advantage. So to me, learning to code is not a, it's not a, it's not a tradesman skill. It's a, it's a structured thinking process. Right. Like the vast majority of coding as a, as a professional engineer actually goes in before you physically write the code. Like generally you think about all the interfaces and you, you structure it, you write a lot of boxes on a whiteboard and then when you go and code it, it's like just go and execute it so that Part we can kind of just leave to the machines. But I think understanding how to structure a problem and think through it is still important, you know. But you kind of brought up like what do you do with your kids? I have a 19 year old and 18 year old, my 19 year old studying math and I'm like, just learn to think, learn to break down a problem. And agency like agency combined with learning to think is will still be valuable. You'll still have to, we'll still need to do that. And like all the things you're talking about Anastasios is okay last mile problem. Well, you got to think about how does the whole system work? If you don't understand how to break down the problem and understand how the system works, you can't apply your agency to fix something. So to me the answer is actually resoundingly yes, learn to code. But for the purpose of not being a tradesman in building software anymore, but more just to understand how machines work, how processes work, how to break problems down. So I think I'm still pretty bullish on people understanding how computers work.
- [59:11]
Jason Calacanis
I don't think learn to code or not.
- [59:13]
Chase Lockemiller
I think the volume of people that need to learn to code is going to be smaller. And you know, I think things like these coding schools that we're teaching folks how to use React and Ruby on Rails and like these things that you know of spinning up a website like that that's like not useful coding.
- [59:31]
Jason Calacanis
They've already been abstracted away.
- [59:32]
Chase Lockemiller
They've already been abstracted. But yeah, but to Naveen's point, like learning, you know, low level system architectures and how software interfaces with hardware and how memory allocation works and how, you know, you know, understanding the actual core engineering principles that go into operating the machine I think are going to be important because you're going to need to be able to think through how to solve problems with, you know, computers as a vehicle to solve those problems. And it will also help drive the innovations that are going to lead to the breakthrough computer architectures that are going to be the next generation. I mean I think about a lot about like what Naveen's working on now. It's like an inspiration to me. It's like, you know, just rewrite everything from first principles. Why, why is, you know, like why is a computer designed the way it is? Why is the transistor designed the way it is? Like do we have to do it that way? Is that the best way to do it is in the, in the age of, you know, AI, I don't know, and. But just re. Evaluating a lot of those core primitives, I think is a important exercise. And those that actually understand machines and software and, you know, what's actually happening in the underlying hardware, I think is actually a very, very critical thing to learn for a small subset of people.
- [60:45]
Naveen Rao
I think embedded in your question is actually kind of two questions. It's more like, how do you. Should you learn to code? If you want to be in the sort of elite group who's kind of leading these things? And also, what should everybody do? I think those are two separate answers, actually, to Lock's point, to Chase's point here.
- [61:04]
Chase Lockemiller
Yeah. And one other thing I'll add is that, you know, should you learn to code? Like, to me, a lot of, like, Claude, code is a new way of coding. You are coding. Right. But you're doing it through a different vehicle in the same way that, like, should you learn how to write assembly? Like, probably not. Like, should you, you know, like, learn how to write C? That's like a. Maybe a useful skill for a foundational knowledge piece. But, like, you know, languages have extracted to higher.
- [61:31]
Jason Calacanis
Like, it's the paradigm shift. And higher. Yeah. It reminds me of a big debate that happened in the 90s when digital film cameras came out. They started to make digital films. Center of the World, a film with Peter Sarsgaard that I have a small part in. Bennett Miller did one called the Cruise, a documentary. It was all because this VX1000 Sony digital camera came out and you could edit on your MacBook. And there was a big controversy at Sundance, like, well, what do we do with these digital films? Because anybody can make a film now. And they were like, yeah, that's the whole point of Sundance. Anybody can make a film. We used to have people submit them on 8 millimeter, 16 millimeter. But you had this paradigm shift. Should I learn how to cut film and splice it and edit in a film room and tape the film together? Or is this new paradigm that feels in some ways analogous to this moment in coding?
- [62:22]
Anastasios Angelopoulos
Yeah, I think Chase had, you know, was saying exactly the right thing, which is that. And you. You're sort of supporting it, which is that the history of computing has involved layered abstraction, starting from the bare metal assembly to higher and higher level languages. And the relevant question when we ask, should people learn to code? Is what do you mean by code? Because if you're asking the question, should people learn to think? The answer to that is probably yes. But if. And but is thinking the same thing as coding and communicating the same thing as coding. It might be, you know, as Karpathy said, the next programming language is natural language.
- [63:09]
Chase Lockemiller
No, that's spot on. Like Python felt that way a little bit when it, you know, when I first started using Python, it's like, oh, am I coding? I'm just sort of like writing out in English what I want to happen. But like, you know, Claude code and cursor and these things are sort of like a further layer of abstraction away from that. And sure, it gets compiled and, you know, put into all these scripts and compiled into bytecode and all this stuff. But. But it is. That's spot on. You know, natural language.
- [63:33]
Jason Calacanis
Trying to draw a circle on the screen of the computer using BASIC on your IBM PC versus, like just drawing a circle with, you know, your Microsoft Paint or Photoshop tool. Let's end with something a little spicy here. There's been a big debate over the last couple of days. Obviously we're sitting here right after this military action with Iran. Put aside the politics of that. But more importantly, AI's role in things like war is important. Things like surveillance. And you've obviously had Claude take stances. OpenAI flip flop their stances. A lot of politics in here. There's a lot of nuance to it. But I guess the two things that Dario brought up that are super relevant. Should a technologist and is this technology ready to be deployed on the front lines to build murder bots? Should we be building that? And then should technologists, because this technology is powerful in a way that other things have not been powerful. Probably nuclear power would be the best. The next best analogy. Should scientists, physicists, you know, be making nuclear bombs? Feels actually eerily similar. And then obviously the police state. Nobody wants to live in a police state. That's a 90, 95 issue here in America. And this technology would be uniquely give people the unique ability to do that. Where do you guys stand when you read these stories and you think about it? Because this feels like a moment in time. Time. Chase, I'll start with you and just
- [65:18]
Chase Lockemiller
go right, the question specifically is like, should we use AI in warfare or.
- [65:24]
Jason Calacanis
Well, like, I think maybe the. The role in the. The tool builder building this. The person who is manifesting this technology in the world and then giving it to an agency like the Department of War, you know, just like the, you know, physicists giving, you know, the atom bomb to, you know, a president or the Department of Defense back then. Yeah. How do you. What's your just general take on what we've seen in the Debate last week. I'll let you take it where you want.
- [65:55]
Chase Lockemiller
I mean, look, I think my perspective is that, you know, with any technological shift that's happened over all of human history, you know, that those techno, those new technologies are used by people to gain power, influence, control over other people. And, you know, you've seen this, you know, with the development, you know, from like World War I, which was sort of like the first war fought with like Gatling guns, right?
- [66:26]
Jason Calacanis
I mean, long guns, you know, you
- [66:27]
Chase Lockemiller
know, like, you know, versus these, these muskets that you would like, you know, manually load and, you know, couldn't, couldn't really do that much damage with like, you had these like automated, you know, industrial weapons that, you know, caused tremendous, you know, damage and, you know, were able to. They were mass killing machines. And, you know, the country or the superpower that had, you know, the best control over those things ultimately kind of like was won. Right. So, and I, I guess, you know, and then World War II with, you know, the advent of the nuclear bomb, I think, like, you know, there's, you know, still an ongoing debate as to whether the Japanese would have surrendered if we wouldn't have, you know, you know, dropped, you know, Little man and Fat Boy or Fat Boy and Fat man and Little Boy. Yeah, yeah. You know, and I think, you know, sort of independent of what I think should happen or anything else, I think any global conflict is going to be at the foundation of it. It's going to be fought with information, it's going to be fought with AI and AI is going to be embedded in weapon systems. And like, so I don't know that it's my job to opine on the should this be happening or not. I just think it more. It's like, to me, it's an inevitability that, you know, AI is going to be used by the U.S. it's going to be used by China in any sort of, you know, it's going to be used by, to the extent, you know, whatever they have access to, it's going to be used by, you know, Iran, Israel, like anybody who can get access to it. They're going to use that to try to get a leg up. So I don't know that I have a perspective on whether it should be happening. I just know that it will happen. And, you know, what do you think?
- [68:12]
Naveen Rao
Yeah, so when I was at intel, actually, if you remember the Uyghur Muslim targeting in China, yeah, that was running on my hardware, my group's hardware. So we developed a Responsibility framework and how to think about this a bit. And basically it came down to how close to the end application you really are at as a technologist. So if I was building, if I were building the software that was running in the camera that's specifically looking for this ethnicity, I would say I bear a lot of responsibility and I have a choice to make as a technologist, whether I want to build that or not. When I build the chip, I don't because I'm building kind of a basic, a capability. And I would actually argue AI falls into this category. And so moralistic questions I don't think should be so far out. Personally as a country, I think we should use every advantage we have. As Chase was saying, like that's how wars are won. And it's not my responsibility as a non elected person to make that moralistic judgment. I can do it if I'm building the end technology and then I can choose whether I put my time and effort into that. But as someone who's built a generally applicable piece of technology, I think anyone should be able to use that for how it can work. Now what I would have done in the scenario, if I were a Dario, like my dad actually asked me this question and I said, I would have moved the conversation to technology. I wouldn't have done it on moral grounds or ideological grounds. I would have said, look, I am not okay with you using this for targeting for weapons or whatever because the technology, it hits the limitations of the technology. So I would agree to a contract to say, we will help you assess. Is this a use case that this technology can actually do or not? Is it inbounds of what the technology is possible to accomplish or not? And that's the framework.
- [70:03]
Jason Calacanis
If they make that choice, it's basically like taking a certain airplane that you built beyond the spec and you're like, you can't take it to 50,000ft. It's not rated for that. If you do, the pilots die and the fuselage breaks up. It's not meant for that speed or height. It's out of parameter.
- [70:21]
Naveen Rao
But if you, if you want to transport, you know, rubber dog shit and within spec, that's on you, you can do.
- [70:27]
Jason Calacanis
Yes. Anastasia, you must have been thinking about this. You're thoughtful. Greeks always are.
- [70:35]
Anastasios Angelopoulos
There you go.
- [70:37]
Jason Calacanis
So let's think this through here. You know, is this, hey, we're making batteries and you're putting batteries in a drone and batteries can be put in a Walkman. And you know, in your ev, we're just battery makers, use it as you will. Or is this more analogous to, hey, we're giving you the atom bomb and this is a very unique new technology. For the love of God, we don't want you to use it to build murder bots that you cannot control because this isn't ready. You can see it if you use, you know, chatgpt and you've had it hallucinate, you understand it makes mistakes, and then apologize and says, oh, yeah, thanks for catching that I made a mistake. Which doesn't work if the murder bot turns around like on the movie RoboCop, pretty prescient and murders the executives at the company. So, yeah, take me through your thinking on this. Where do you sit?
- [71:27]
Anastasios Angelopoulos
In my view, a lot of this question depends on how you view the responsibilities of a corporation in the context of a nation.
- [71:38]
Jason Calacanis
Okay.
- [71:39]
Anastasios Angelopoulos
And how you believe.
- [71:40]
Jason Calacanis
Nation got it.
- [71:41]
Anastasios Angelopoulos
How you believe decisions should be made in the context of a company within a country. Countries go to war. It's a fact of life, as we've said. And when you go to war, there's a lot of information that the country has that the corporation doesn't have. They just need to be. If we're going to be part. If you believe the corporation is part of the same, is a part of the country and is supportive of that country, part of what that means is that the country is responsible for going to war and dealing with that. And the country therefore has certain abilities to force you to conscript you to war. They can take all of this and tell us to go to the military if they want. That's the right of the country.
- [72:21]
Jason Calacanis
And the country has.
- [72:22]
Anastasios Angelopoulos
Yes, and they have. And the country has rights to do to other things of that nature. And because they have information that we don't have and they have the right to set the strategy. And to some level, if we want. If we want the country to be successful, there's a level of disagreeing commit that's sort of enshrined into law by virtue of the capability of the country to, you know, seize your company and. And, you know, just like they've. They've done previously and make you, you know.
- [72:52]
Jason Calacanis
Yeah. You're no longer making cars.
- [72:53]
Anastasios Angelopoulos
Defense Production Act. Yeah. To make you do things in favor of the war. So the country has those tools at its disposal, and it should have those tools at this disposal.
- [73:05]
Jason Calacanis
Yeah.
- [73:05]
Anastasios Angelopoulos
Otherwise it cannot be expected to deliver on the outcome that we've all told the country to do, which is to be successful in governing in. In war and so on. And then we have the political system in order to elect people that we think are going to be good leaders on this front.
- [73:22]
Jason Calacanis
So you don't object in any way for them making their feelings?
- [73:25]
Anastasios Angelopoulos
No, I can object. Yes, I can object.
- [73:28]
Jason Calacanis
Critically important. You can object and still be a patriot.
- [73:32]
Anastasios Angelopoulos
You can still be a patriot.
- [73:33]
Jason Calacanis
Warnings to Naveen sport and say, point and say, hey, just so you know, like, this may or may not work.
- [73:40]
Anastasios Angelopoulos
Yeah, I can object. But it's ultimately. Yeah, but it's ultimately a nation with laws and we, we set a direction and we have to kind of unify together to go, you know, to, to. It is one country. At the end of the day, there can only be one choice. Yeah. And I think that companies can, of course, express their opinions and so on and so forth, and so should individuals. And that's a patriotic thing to do.
- [74:06]
Jason Calacanis
So that's. And that's a great part of it. Chase, you can speak your mind. And then ultimately, if we were ever in a situation where, God forbid, the country was under attack, again, hey, this is, this is the rules and the operating system of the corporation known as usa.
- [74:23]
Chase Lockemiller
I really liked some of those points you made. And it is sort of fascinating evolution to think about the, how the public sector and private sector sort of work together because in many ways you have these inherently like, natively international, natively digital private corporations that are building technology. And sure, they're headquartered maybe in the United States, but like, they serve a global population. And then, you know, you have this private. Or you, you. Sorry, you have this, this, this nation that, you know, has other geopolitical goals or aspirations. And it's just a very different paradigm from like, you know, the development of the atom bomb. You know, that was a, you know, government sponsored, government mandated, you know, technology and weapons development program. Right. And when you think about like this, this like, you know, Claude was like funded by independent venture capitalists, private capital, resources to serve a global population. What is their, you know, obligation to service the, you know, needs or desires of the United States, a single country and sort of their global set? I don't know. It is complicated and I don't think I have the answers, but I just know that AI is going to be part of whatever, whatever global geopolitical conflicts unfold.
- [75:47]
Anastasios Angelopoulos
Yep.
- [75:47]
Jason Calacanis
All right. This has been an amazing episode of this week in A.I. wow, what a strong panel. Chase Crusoe, Naveen, unconventional. A.I. anastasios from Arena. A.I. gentlemen, thank you so much. We will have you back. If you would like to subscribe to the program thisweekinai AI and you'll see all the links there to subscribe brand new podcast so tell your friends, like rate, subscribe all those great things and we'll see you next time. Bye bye.