The Rise of DePIN: Redefining the Compute Landscape - TOKEN2049 Dubai 2024

This team is clearly ready to work. Hello everyone. Let's get started right away. My name is Richard Muir. I am, as already mentioned, the managing partner of Fabric.

I have been working for ten years. I support great founders in this field. Unfortunately, not enough in deep circles, but, um, great projects like Near, Ocean, Polka Dot, etc. Look at Fabric, we've always supported the superpower of Web 3, um, which is the ability to coordinate people and deploy their activities and resources. Um, hmm, if you give a token, everyone will achieve more on the team.

Of course? I came up with this this morning. You can get it for free. Um, this is an area that I, um, particularly love, although we haven't made any investments yet. Um, we've been looking at how to allocate, um, network bandwidth since the last century.

Um, and we know that the granddaddy of use cases in this area is how Proof of Work works. We're proud that it's, um, in place. Proof of Stake is clearly in right now, with the fantastic success of the distribution merger a year ago. Hmm, and, you know year ago. Hmm, and, you know , that's fantastic.

But now is more than ever an important time to try to move to the next level. Because, um, as we shift into high gear toward artificial intelligence, and artificial intelligence, and maybe even artificial intelligence ( artificial intelligence ( AI) after that, we find bottlenecks, and those bottlenecks, those limited resources, become incredibly expensive. And, if we want to avoid the centralization of power and profit, and perhaps propaganda, with companies run by capitalism, we need to find a new business model that is, um, decentralized, that has the same economic impact as a centralized model, and I think if we do that, we can all create a system where the outcomes are much better than a fragile centralized model. It can be adaptive. It can be emergent.

It can serve the needs of individuals, communities, and the world at large. So, I mean, the stakes have never been higher never been higher . So, with all of the above in mind, allow the panelists to introduce themselves a little. Let's start with Rand from Ioteex, and I want to say that he seems to connect physical devices of almost any type. Given his experience with Uber, we would like him to give us a decentralized Uber and an open, permissionless Uber, and they could all work together, and we could get around these bottlenecks more efficiently.

But I'll pass the word on to you, Ran. Of course. Hello everyone Of course. Hello everyone . So, this is Ran.

I am the CEO and co-founder of Ioteex. We started back in 2018 We started back in 2018 . We've actually been focused on the Internet of Things and blockchain, you know, since day one. Now, after several years of iterations, it's called "beeping," but the basic spirit is the same in all of them. So Iotech is about creating deep infrastructures for builders, right?

So, in this release, creating deep projects should be as easy as, you know, creating D5 or N-type projects. A small team with a big heart can contribute to the work, so this is our foundational release. Hmm, yes, we have our own first level, like option calculations, some things with identification, as different as version 2.0. It's as if the entire roadmap is actually being implemented.

It's so nice to talk to investors, builders, people who are passionate about DP. Thank you. And now about you, Trevor, almost the same age as me. Hmm, I have a story about making our dreams of a metaverse a reality with Render. Fantasy.

Hello everyone. Trevor from Render Network. Hmm, for those of you who don't know Render, Render was founded by creators for creators to provide fast and affordable rendering in cinematic quality. And today, thousands of motion graphics artists can provide their consumer computing to run AI and rendering tasks in parallel for significantly less time and cost less time and cost . Thank you, Trevor.

And once again about Ahmed from io.net. net, which I've heard of net, which I've heard of , has more GPUs GPUs than Zack. I think I have about a million of them. I think this is about the right number.

Yes, of course. Tell us a little more. My name is Ahmed. I am the CEO and founder of IO.net.

Hmm, Ionet—the creator of the internet of GPUs. Essentially, a new network that started with a profound question. What if access to GPU computing power for AI was as easy and seamless as access to the Internet? Hmm, and from there, um, we created the I/ created the I/ O cloud, all the I/O coins, I/ I/ O workers, and all the multiple components that essentially created a compute layer on top of multiple networks in the network, uh, in the crypto industry, like rendering and so on. Um, and from there we essentially created an infrastructure layer that allows anyone who wants to run large-scale AI workloads to seamlessly deploy a cluster of 5,000 or 10,000 GPUs that you won't find anywhere else on the planet.

This is Ionet. I am very happy about this. Thank you. Last but not least, Ariel, hmmm, I'm interested to hear about mapping the world and the honeypot you created for people. Good.

So, hmm, it's a very simple idea. We're creating a new global map, ahem, to compete with services like Google Maps to help people in their daily navigation business, right? So Uber, Lyft, obviously, are mass consumers of maps, hmm, and ultimately navigation for people just running around the city, etc. running around the city, etc. So the idea is pretty simple: you buy this device from us, install it in your car, and then you just do what a lot of people do, which is drive the car.

And all the images collected here are then processed so that we can understand: okay, there's a 35 mph speed limit sign. There are three lanes of traffic here. Road construction is underway here. All of this happens on this device through a series of map-based artificial intelligence processes. And then we create this map and then we sell it to different companies and organizations around the world.

We mapped about 18% of the world's road network in about 15 months. So it clearly works. And the most wonderful thing is that unlike paths unlike paths , or many of its predecessors, where the people predecessors, where the people who actually created the map never participated in the economic component of it, when you create this map with this device, you actually earn our honey token, and in this way you share in the underlying economy of this map that we all create together. Perfectly. Good.

So, we'll start at this end, but then we'll move from one stage to the next. Let it be unexpected. And, um, I'd like to raise a few questions from the audience near the end, if that's possible. I received a specific request from the panelists to make this interactive. So we can ask for about two or three and then maybe choose the best one, if that's allowed.

Um, you know, back to the key question. We talked about how this new business model this new business model , which is obviously based on tokens and the token economy, um, you know, is important if we're going to fight centralization, and um, but there's a feature, not a bug, about it that, um, is attractive to speculators. Uh, I want to start with a question about how to harness the power of these speculations, but also about people who are really going to be experienced and help. I'll move on to the second question, which is to put a card or, you know, put GPUs and so on. So I don't know, Ran, do you want to go first?

Yes, of course. I think Yes, of course. I think , um, you're talking about speculation, right? But speculation, looking at it from another perspective, is essentially like capital; they gave you, like, an initial, like, capital injection when the projects are really small. Think about, you know, you have, like, three friends who work together.

You have a great idea to conquer the world, right? Where does capital come from ? You don't want to approach a venture fund from day one, you know, trying to make a hard sell, a hard pitch. But you can try, try something on your own, maybe have a token, maybe have a small community, maybe get people to believe in you. I think that's kind of the magic of cryptocurrency, uh, in terms of depth, especially, you know, IoT networks are notorious for having huge upfront costs to develop, to manufacture, to distribute devices, and then you see if something works or not, right?

But deep trading, I think, has actually had this magic, like cryptocurrencies, from day one; they give you something you can try, right? Maybe with a small initial capital you can also prove to your investor, to the whole world, that something already works. So I definitely think this is really good for the cryptocurrency cryptocurrency defendant. It is clear that she receives capital. So for us, it all starts with abstraction, where artists never have to deal with cryptocurrency, they are brought into Web 3 from a fiat money perspective, where they use the network to pay fiat money for rendering.

And then, most importantly, it's kind of like an onion that they can peel from there, where they can contribute, become a node operator, and by doing that, they start earning and learning. Um, what we do on the cryptocurrency side is that it all comes down to choosing the right model or token model or token economics and then being able to leverage incentives for different players in the ecosystem. So a great example for us was that we recently upgraded our blockchain, and in the process, we were actually able to offer incentives to upgrade the crypto upgrade the crypto side so that they could help us transition in a way that would actually leverage it. Perfectly. Well, I mean, look.

If, if I have a data center that's about a hundred million dollars, not too big, like an H100, and if I just buy GPU power , put it in the data center, and wait for data center, and wait for my sales team to make long-term contracts to sell GPUs GPUs , then this time I'm losing out where I'm not being hired, my device is not being hired, my GPUs are not being hired for any workload, whether it's an AI workload, or any workload, then this time there's a loss of time, a loss of opportunity. But when in data centers like Amazon or Google Cloud, when these cloud providers, when you want to deploy a cluster, it takes about 15 minutes to boot a node because it's not online; they want to reduce electricity consumption, etc. It takes about 30 seconds for us and you're up and running. So imagine the difference in speed. But the reason we have this, and the reason that incentivizes this center to run on Ionet rather than AWS, if it has the option, is that on AWS, that time has passed, right?

I won't give anything away right? I won't give anything away until someone comes here. But what's so interesting with the token system is that you can give him some kind of reward from the IO coin rewards IO coin rewards , small enough that he stays, you know, keeps his cool; the electricity bills are paid, you don't have to worry about your capital expenditures, etc. until a job comes your way. Which is basically, if I'm just like, like, if I don't get hired for two whole years, if it's just my GPU there, I'm essentially getting the cost of the GPU back GPU back .

Just because I'm on the network, I'm providing this capital, this computing power, to the network. That in itself creates a buzz, like I want to get this coin, be a part of something like that, and then, you know, reduce my spending. That's why I like token- token- tokonomics and deep development projects. I like it. I think it's extremely powerful, you see how it can provide, you know, cheaper but also more adaptive than a centralized model using a token economy.

Ariel. Yes, I will give you two examples. So, the first one is this guy Brad. So Brad bought, I think, about a hundred of these devices pretty early on when we were just starting out, and then he figured out what drivers were right for the card, right? So, obviously, Uber drivers, Lift drivers, that's what it is, and that's why he bought them with equity, right?

But then he distributed them to all these Uber drivers all these Uber drivers , Lift drivers, Amazon Flex drivers, and then he shared the tokens with them. So that really helped the network grow because he was identifying the right type of driver. He had capital, did n't he? Many Uber drivers, Lift drivers may not have capital. So, that was one really smart example, and he's now taught other entrepreneurs who have come online, saying, if you want to replicate what I'm doing, here's a guide.

So, that's one example. Another example now relates more to commercial fleets. So this guy who runs, I think, a thousand rental cars rental cars , uh, uh, he bought a thousand, right? And so he places them in all his different car rental locations across all the countries. And this is a bit about the commercial side.

So both of these options are really amazing because they help the map grow in a very efficient and cost-effective way, and they also reward the person who is risking a small amount of capital by buying these devices. This is great. This is super interesting because, as you know, different variations are used throughout the group. So the next kind of question, and you can take it with examples wherever you want to go, is the question of matching supply and demand. Um, you know, for example, you know, you don't want to map the whole of central London and have, you know, 100,000 versions of the same map.

And obviously there are many different layers to what you're mapping, but at the same time there are different areas that aren't mapped at all. And I think another option would be this: there might be a big demand right now for mapping puddles around Dubai. How do you react like that? And the same thing in another use case use case , but on the GPU side, how do you, you know, provide low latency for logical inference in one part, and how do you provide, you know, the high throughput that you need for training in the other? So training in the other?

So , uh, you want to go back to the first arrow in the other direction? Yes. So a really cool example. We work with this autonomous vehicle company vehicle company , and they were tracking four different cities in America, like Phoenix, Austin, Houston, etc., and they were tracking a certain type of problem.

Uh, it's related to construction zones and so on. So, take Phoenix, for example. They told us, " They told us, " Okay, here are about 3,000 locations across Phoenix that we want to see or map at least three or four times a week." OK. As for the network in the Phoenix metropolitan area, for those who are not familiar with Phoenix, Phoenix is huge.

It just never ends, does it? So the regular network covered about 80% of them, right? Three or four times a week, which is pretty good. But then, to achieve what the client wanted, which was about 98% coverage, we introduced a concept called "serious parcels," which is that if you take a five- if you take a five- or ten-minute deviation from your normal route, you'll earn a little more because you'll get to that specific intersection specific intersection that we were interested in. And it worked amazingly well.

We reduced coverage from about 80% for their three to four thousand locations to about 98% coverage. But then again , three or four times a week, right? That is, you know, a lot of coverage per week for, you know, 3 to 4,000 locations. So when you can identify a very, very, very specific location and then say, "Hey, if you're in this area, do a five-minute hookup and you'll earn a little bit more." You can get down to this micro-micro level.

And then there are also things that we do more on a macro level, like, "Hey, we want more of this region, less of that region for various reasons." Fantastic examples Fantastic examples , and we may come back to this, but I'm curious. How can this be automated? That is, the network itself decides where it needs to change incentives to fill the gaps. I mean, you can start now, or, um, I'll just briefly answer this question about how we do it.

So when customers start using the card and we're not getting good results, we actually automatically create these surges if we think there's not enough demand based on what we're seeing on the demand side. Yes. So I'm passing it on to you. think when we first figured out the proposal, it was more like a chicken and egg problem that we live in. Hmm, initially we have a few customers who say, "Hey, we need three A100s, we need four, 400, sorry, 300 A100s or 400 A100s."

Hmm, and we say, " We only have 300," but they want another 300 or a thousand, so we didn't have that offer. But then offer. But then , as soon as the incentive token appeared, we were able to increase the supply. What actually happened is really interesting: we had a customer come to us, tweet some startup and say, "Hey guys, we have GPUs in Lambda Cloud and AWS, but our use case doesn't fit in with hosting GPUs there, and they don't have GPU supply in Korea, South Korea, or Japan, because most of our users are there, but there are no GPUs that we can rent there." Why?

Because their app, and we're seeing this trend now, that we're going from a 5- going from a 5- second wait for an image to be generated to real-time. I want real- time artificial intelligence. So imagine it's like Sora. For example, I'm playing a game and it changes the environment in real time. This means that the industry is moving to more real-time model output, which is why I need such fast generation, and this can only be solved if GPUs are available in every country where there are users of this product.

So this company that approached us said, “We have users in Japan and Korea, and they want us to use GPUs in Lambda Cloud, and it’s going to take 300 milliseconds to transfer; geographically, it will take time over the Internet to reach this goal, but our requirement is 150 milliseconds for the image to reach and return.” So we think you have GPUs in countries where there are no GPUs GPUs . Therefore, we would like to deploy our model geographically closer to the end user. So all of a sudden we found out that we just built a CDN for machine learning models, meaning we're essentially distributing the client model to 139 countries, and then if a user comes from Saudi Arabia or Dubai, the trip doesn't have to be made... the computer request doesn't have to go to Microsoft in San go to Microsoft in San Francisco or on the East Coast, etc.

He can just go to the nearest GPU vendor here, right? And that essentially gives a latency of 20 milliseconds, which was really amazing. And they moved their entire infrastructure from such providers simply because of availability, but we couldn't achieve that kind of availability before. We kind of brought this together from the crypto community, like we all worked together to build this huge network of GPUs in many places around the world, in 140 countries. And then, with this huge margin, we suddenly found that we had actually created something that no one could actually create without 6,700, 800,000 GPUs.

It's like no one has reached this level, with 800,000 GPUs worldwide. Maybe, yeah, Zuckerberg has it in San Zuckerberg has it in San Francisco and his Facebook data centers, but it's in one place here, we have it, you know, all over the world, which is something new that we have n't done before, it's pushing the boundaries, like we're going somewhere we've never gone before. So this is really interesting. Yes, absolutely fascinating. And I was just, um, looking at Zuck's announcement in the last few days that Llama 3, you know, will be open source, which we, you know source, which we, you know , are very supportive of.

And also, I haven't tried it yet, but I guess you could, you know, talk about output and latency, you could print a query to generate images, and as you print, the image changes. So, I mean, that's the expectation that we'll have when we move to a generative model. And that's one of the things I took away from Jensen's Nvidia keynote the other day, the whole architecture, you know, the physical part of the Internet is going to have to change to adapt to this generative model. So I think that's exactly what you're talking about, Trevor. Wow.

Scaling a two-sided market is two-sided market is difficult. In fact, this is one of the most difficult challenges faced when managing one of these networks. You usually start by getting enough computing resources for people to connect and use . But at any point, if you break that, you're going to end up in a situation like we had this week where you can't find an Uber, and you can immediately tell that it's unbalanced. For us, it was really focused on getting the initial batch of users first, getting those nodes to stick, and then approaching it from an artist perspective and building that up.

And it was a phenomenal run. While we were doing this, we had a list of weights that we grew. It has about a million graphics processors graphics processors . So it was an asset that was accumulated externally without disturbing this balance. And what was great for us was that this year we were able to rediscover that as we started to move into artificial intelligence and bring in partners like Ionet who were then able to use that in different ways.

But it's a really complex problem, and frankly , you can be out of balance at any point in the system. So it requires constant attention. I just want to point out the rendering point out the rendering , because at the moment this client came to us, we were like nobody knew about us. Ionet was hidden, and I called Trevor and said, " Trevor and said, " Look, we have GPUs in Canada and Iceland, from, you know, through Falcoin SP, but we need GPUs everywhere because this customer came along." And something like, "But we have a community everywhere, let's connect them quickly."

And so we connected their GPUs, and that was the beginning, because they already have a huge community around the world, and without them, honestly, this wouldn't have happened. So thank them. Good job. Yes. This is quite an interesting question regarding the use of DP, isn't it?

Or usage options. Um, DP usage options. Um, DP is definitely a design for people, used by people. So we're not saying that DP replaces the entire physical infrastructure from day one, but it actually fills some small niches here and there. I can give you a few examples.

So there's a project that's doing something like a Tesla charger, something like an Airbnb charger for Tesla on Ioteex, right? The idea is quite simple. They tested this idea in a small community. Essentially, you can share your charger with other Tesla owners other Tesla owners , and they'll charge them a fee that actually works in a small environment. They then plan to scale using Ioteex as the underlying infrastructure, meaning this is one use case for how deeply Tesla charging can be made more efficient, convenient, and cheaper.

And another example that we've seen so far is called Network 3. It basically uses devices, you know? They have people running their servers, running their software actually on their computers during downtime, and they measure, for example, the Internet, or collect information from the Internet in a decentralized way, right? And all the information they create is aggregated together to create images for the web. This is also another very interesting project that doesn't have a huge market right now, but it's actually being used by a lot of people.

You know, it's kind of useful. I feel like there will be a deep dive here and there, and eventually we will replace the world. We see something like a web version of Uber one day. one day one day, yes, hopefully soon. So, um, obviously Uber had its own regulatory issues in different geographies early on when it was, you know, expanding.

Um, I'm curious, and we haven't really had any prior discussions about this , but, um, you know, we all seem to be excited about the power of the token economy incentivizing people to create this new physical infrastructure, and that it's going to be, you know, very powerful. Um, what challenges or issues do you face from a regulatory perspective, um, with the unfolding of this force, maybe in a commentary on different geographic regions? I mean, we know that Dubai and the UAE are working very hard are working very hard to make it, um, a welcoming environment from that perspective. So, I don't know, do you want to start over? Do you want to start?

Yeah, I mean we have it on both sides, so obviously there's the regulatory side of cryptocurrency that you're talking about, but you know, uh, we also have it because we operate in the physical world and these devices are driving around in all these different places. Uh, we also have privacy rules, and also national security rules national security rules , right? For example, South Korea has very specific, uh, regulations related to their national security and mapping and where all this data can be stored and so on, right? So, this is a good example. Europe, as many of you know, has very, very, very strict, uh, privacy rules, right?

So there's a whole set of things that we need to do in terms of processing all the data on this device, you know. So we need to effectively mask privacy to blur out all the, uh, names of specific objects, uh, people's faces, bodies, cars, license plates, all that kind of stuff. And that's why we actually do a lot of the map processing with artificial intelligence, detecting, for example, a 35 mph speed limit sign. There is a construction zone there. So in some locales we don't even upload images, right?

Because we are not allowed to. So it often depends on the country, and even within a country, on a specific state or region. So you have to be able to balance all these different privacy rules on top of all the crypto rules. So we have it on both sides we have it on both sides . Do I understand from this that you are not commenting on crypto regulations?

I'm not commenting on what? Cryptocurrencies crypto Cryptocurrencies crypto rules. Yes, crypto rules. Yes, crypto rules...I leave...

I'm American. I have to...I have to leave that to someone else. Good. But it's interesting from a privacy perspective.

I think you're obviously think you're obviously , in particular, not just coordinating physical infrastructure and activities. You move on to the data issue. Um, and I think all of this innovation that's happening at that level is going to be key to trying to make sure that we can do well at the data level as well , where one of the problems is that the system can be manipulated, for example, when you're trying to determine what's true. I think in your case, you know, an image is an image, and it's deterministic. And in Trevor, in your case too, it is so.

Um, but you know, this issue of privacy and truth is going to be important on a level beyond the physical. Yes. I mean, um Yes. I mean, um , there was an image circulating that the Pentagon was attacked, right? And they said it looked pretty realistic, right?

Um, and the markets responded to that. It turned out that this image was clearly fake, that the Pentagon was not attacked, that it did not explode, etc. But the main concern is concern is if someone hacks into these systems and says, " systems and says, " Hey, Hivemappers found that some part of London exploded or was destroyed." This will have serious consequences. So yes, we're thinking a lot about thinking a lot about how to verify that the data coming from this device is actually accurate.

You become a news agency or a source of source of alpha for hedge alpha for hedge funds. Maybe we already are. Maybe you already are. That's right. Um, okay.

So, Trevor, do you want to comment on this fact-checking issue? Yes. Yes. I mean, it's about regulation. I think decentralized GPUs really help solve this number one problem because, um, you can do it for each country individually.

You know, we're in the middle of a real GPU battle right now, so being able to choose just a few of them in certain places, I think, is a real advantage. Um, but when it comes to the data side that you're talking about, um, it's really critical for us and our artists . They are angry because they have centralized companies that teach on their data, on their artwork. Um, and that's what we really want to try to change. We're actually moving towards creating synthetic data in the Render Network to be a base model for training people without stealing data in that model, and then layering on top of that, and using blockchain in a way that it can be verified, and that really gets to the core problem that we have here when we talk about images.

We're going to move to a state where creating an image or video is trivial, and there will be tons of them everywhere. And in that case, it's very, very difficult to distinguish between what's generated and what's created. And so using blockchain to me is fundamental to demonstrating proof of creation, proof of rendering, um, and, you know, that's one of the reasons that really drives Render Network in solving these problems. Yeah, I think that's a good point that, you know, providing a truly decentralized physical infrastructure helps with the argument about crypto regulation being , um, not security. But it also helps us create an infrastructure that we can, um, maybe we can, um, maybe , trust, a trustless infrastructure that we can trust to produce, um, um, data around which we can reach a consensus about the truth.

Um, you know, this is a foundation that we definitely need, and again , on your side, I'm trying to entice people to join the conversation about crypto regulation, certainly regulation is extremely important. I think DP is, um, least viewed as a safety, right? Because everything we talk about is about utility, right? For example, for transportation, for your data, like for UBI, even these things are useful. But you mentioned one very important point–this is the DP check.

This is a level that doesn't really belong in blockchain right now. So this is kind of what we're working on overall, like a deep protocol stack, you know, you need an opportunity or another network outside of the blockchain that collects all the data or information from the GPU GPU , from sensors like that, right? So you need people to implement the rule here, at this level, say, "Okay, here's the rule: if these data points show that they can pass these rules, that counts as valid data points." This can be done using, for example, zk snark; this could be with the help of, for example, TE. There are a variety of technologies we can use to provide this verifiable proof.

This proof essentially goes to the blockchain, says, “Oh, so this device did its job, or this GPU did this calculation or rendering. That's why you should give this guy, like, 10 tokens," right? So you can't do this level using any blockchain because it's currently resource-intensive. That's why we work with a few teams, you know, here and there they do ZK, they do, for example, a GPU. We're trying to bring all of that together to provide this clear, uh, DPM verification capability.

I think deepen has a very complex approach to regulation because you have to fight, like, it's a two-front battle it's a two-front battle . The first front is the fight for normal regulation of cryptocurrencies, whether it's a security or not, and so on, and normal regulation of cryptocurrencies for storage, and so on. Then another front that you're fighting on is the new executive order and the regulation of infrastructure as a service. So once, you know, the US started regulating trade or, um, the NSA and so on, once they noticed that some people were creating a level that was actually taking monitoring and censorship and control out of government control and making it more private, it became really, really sensitive, especially given the geopolitical tensions between China and Taiwan and all the other things related to GPU bans and exports. So all these embargoes against China are forcing any infrastructure as a service provider to service provider to soon comply with the new regulation that is coming out, which involves not only the usual crypto usual crypto stack, but also a tax on infrastructure as a service.

For example, this new regulation is an executive order, and if you provoke it, if you, you know, if you do this, you provoke the Patriot Act, that's how serious it is. For example, what happened , and something happened to us. When we launched, we had between 40,000 and 80,000 GPUs GPUs . It's not like we launched two months ago, but in March/ ago, but in March/ February, we reached this number. Then there was a workshop between the NSA, Render, Solana Foundation, Falco, and they brought together all the participants in the Deep ecosystem.

Immediately, all the participants in the Deep ecosystem were brought together in a way that could have gotten out of hand, in another way, and frankly, you don't want to give that computing power to China, period. So, there were 80,000 GPUs then , and now there are 800,000. That's capable of a 51- 800,000. That's capable of a 51- pound attack on every GPU network in the cryptocurrency market, that's how dangerous it is. So imagine how much attention hackers are paying, they want to access this computing power and carry out a 51-pound attack on Tongcoin or other GPU- based mining algorithms .

This is one thing. The other thing is that with such massive computing power, you can train a machine learning model that hacks into what's doing malicious things. So how can regulation, now that this is new regulation, be careful not to give D5 services, sanctioning countries, and deal-making countries access to massive computing power without KYC and lengthy processes? How can you, essentially, the computing power...you come to our platform one day and log in; it's like level one, level two, level three KYC, and each level will give you the amount of GPU computing power you can access.

It's not that I want it, it's that if I do n't do it, they'll shut me down. There are no jokes here. Like they're really going to shut me down. And they take this much more seriously than money laundering, for example, in regular DeFi, where someone makes transactions without knowing transactions without knowing who is in order, but someone gets access to 100,000 GPUs without knowing who they are. Oh no, no, this is an application of the Patriot Act.

So now we have to fight on the front lines as if we want privacy for our people. We want the internet of GPUs to be owned by the world, as if it were owned by the people, for the people, in a way that it has verified credentials and all that stuff, that at the same time preserves privacy, but also adheres to regulation that could regulation that could , you know, cause us big problems. So the adjustment for deepening is coming, and it's happening very quickly, and it's happening very quickly. It is faster than SEC. It's faster than that.

These guys are very serious, no joke. Well, okay. So, we thought we were struggling with politically politicized regulation. Good. Applause.

Yes. Um, you thought we had a hard time with regulation, which is sometimes co-opted and politicized. We're, you know, around DeFi and, um, you know, securities and so on. Um, I think we've now marked something like technological sovereignty, intelligence services intelligence services , um, you know, KYC, we have data privacy everywhere. We have problems in every, every quadrant.

We are surrounded, we are surrounded. And it looks like it's going to get at least more complicated at least more complicated , not less complicated. So that's good. We are here, we are here for a cause and for a passion as a community. We're going to, you know, do it slowly, the right way, owning our data, making it accessible to everyone.

Just like on the internet, we all need to have access to computing power; That's all. Everyone should have access to computing power. No one should limit me in what computing power I can access. So access. So , how can we do this in a way that the network checks that the user is not doing anything harmful or trying to be dangerous to others, but still gives people this freedom ?

This, I think, is the future of deep computing. 100%. Of course, if I were to draw on the examples given by my two daughters, access to computing power in the form of their iPad is a human right and they need to have it, but that goes for everyone else as well. No, just in terms of regulating cryptocurrencies, it's probably not worth talking about it too loudly, but, you know, six months ago, a year ago, there was something in Washington like, " was something in Washington like, " Cryptocurrency has no use cases, right?" "These are just meme coins, like this, like this, like like this, like this, like this, like this, right?"

And suddenly Hivemapper appears, Render appears; these are real use cases in huge markets, and suddenly, at least in Washington, you know, that part of the argument seems to have died down, right? And so, like, you need to come up with real use cases. We still need regulation, obviously, that's understandable and all that, but I think we're better off educating these people by saying that there's a real benefit that goes after really huge markets against existing players. So if you just want Google Maps to own the maps for the rest of the world, you know, forever, then of course that's what you'll have. If you want competition, this is one way to get it.

Yes. No, that's why I said that. It helps to have a good use case to prove the utility, which is a good conversation starter. Trevor, do you want to comment you want to comment ? Yes.

Yes. So, um, I think we're almost getting to the last five minutes last five minutes . What intrigued me was that , you know, you know, as you just commented, this has started to gain momentum very quickly recently. I think, in part, obviously, due to the maturity of web-based maturity of web-based tri-stacks, layer-one blockchains , we can discuss which one is your best choice. Some transitions to Salana here, um, I; um, I don't have the Salana packages, um, but, you know, we could definitely comment freely on whether there are other, um, L1s that you might want to move to.

I know there was a lot of activity on the memecoin side of things on Base, like em and ta. I, and I, I'm a big fan of NI, like...but how do you define now how do you define now , "like," is when the activity is increasing, when you feel like you've reached a certain level of maturity in your core model? And, for example, in these last few minutes, where are you going to move on? You know what, what's the next level of innovation, from your perspective, beyond the basic utility that you provide?

Yes. So I think one comment on meancoin...I was thinking about why you're asking, is meancoin actually, you know, desirable, for people, right? If you think about it, it's so simple, and also carries a certain emotion that people want to express, essentially. That's why it's very well understood by people and liked by people and the community, right?

So I feel like deep somehow carries the same spirit. It's also something that people use, and it should be understandable, because "Oh, this is like the web because "Oh, this is like the web version of Uber, this is the web version of Uber, this is the web version of Airbnb," and so on. But I think the deep problem is that scaling or scaling up happens very slowly, right? So if you see, you guys are doing great. Basically, you know, scaling, as in the case of large mappers, from scratch.

But we see a lot of projects, especially on Ioteex, you know, so they have a very steep curve to grow to a massive scale that people can understand. And how can we educate the market so that people can understand, you know? So this is real crap, right? So we have to trust this token. I think this is the biggest challenge in DP and we have to find a way to approach it.

Yes, Trevor, I think you're talking to three of us here on Salana. So I think that in a way it shows some momentum deep down. I think for us it was a technological approach that made sense in Rust. Um, Helium just switched over and had a very similar model. So that was good.

But a big part of that is also low cost and high throughput. And as I look to the future, I think a lot of our initial premise was putting proof of creation on the blockchain. And I think you'll see a lot more of that as we move forward, more in real time. Um, you know, using this, using data creation on the blockchain. Um, and taking advantage of the fact that we have a high-performance, low-cost system for doing this.

Our artists who create things create things that are hosted on every blockchain, and we would like to see that preserved. There are many great uses for this beyond just creating. Thank you. Yes. Um, so today is Hive, so we use Salana.

We love Salana. You know, the way I look at it is, it's infrastructure, so the less resources from an engineering perspective we spend on Salana, the better, right? For example, if 25-30% of our team was on Salana and working in Salana, that would be bad, right? So today it's actually one and a half people, which is great, right? So it's good that it's one and a half people, um, because it means um, because it means the rest of the team is focused on the product, the value and the utility that we can offer our customers or our members.

So today, all the rewards transactions and consumption transactions from the perspective of people consuming the card, it's all on Salon on the blockchain on Salon on the blockchain . We're starting the process of moving some of the card data to the blockchain, right? So that people can say, “Yes, it’s confirmed. This place exists in this place. There's construction work going on there, etc.

, right work going on there, etc., right ? Many of you must have had trouble getting here. I had them yesterday. And so when I got in the car, um, she said, “Yeah, there’s going to be a slight delay.

It's usually a 30-minute drive from my hotel." drive from my hotel." And it was supposed to be 40 minutes. In the end, it took me about three hours. Good thing, because the main road was flooded.

This road has not only flooded in the last hour, but I'm telling myself: well, when Hivemappers reaches full development, we'll know about this flood, right? We'll provide visual evidence of this flood, right? Because we have eyes all over the world. And then we'll really understand that no, no, no, Ariel, this is not going to take you 45 minutes. It's going to take you three hours, because boom, here's 500 meters of this road that's flooded with 60 centimeters of water.

So how do we do this and collect all this data in a chain? It will take a year or two. Uh, but eventually it will happen. Surprisingly. Here's what will happen.

Does anyone have a burning question? I'm probably not allowed to do that. Too many questions. One in the back. There are a lot of people in the hall.

How exciting is that? it's like, really, you guys were, um, you know, um, you know. I think of these as real use cases: practical, physical, close, uh, with a real opportunity to be involved, to earn and benefit from, uh, what's part of these, uh, these systems, uh. And I think that's a testament to all of you for what you've been able to create. I can't believe so many people are so shy.

Um, but, you know, I saw one hand from behind. Um, but are there any closing remarks, maybe we're spitting, we, um, jumped over you at the last minute you at the last minute . Any final remarks you want to make before we wrap up? I am. So I think AI will be 30% of coin be 30% of coin market cap very soon , and I think that should be, uh, one of the most important things that changes in real time.

This is essentially bringing the Internet to 20% by saying the Internet to 20% by saying that the Internet is 20% of is 20% of cryptocurrency. Of course? It makes no sense. That's all, but I think the market capitalization of AI coins will be at least 30% of the market within the next two to three years. Listen, it was extremely exciting, you know, to be able to do this with you guys and get to know you really quickly.

I told the people backstage that we're not really here for the money. We are here because we care about the things we create. I really feel it from all four of you. And, um, I think I want to take a selfie based on that. A whole second.

Are you ready? And I want a round of applause for everyone on stage. And thank you very much for your attention.