At the NYSE Wired + Crypto Trailblazers event, theCUBE’s Dave Vellante speaks with Ram Kumar, core contributor at OpenLedger, about how blockchain and AI are converging to create new economic models for data ownership. Kumar explains OpenLedger’s vision for “payable AI,” where proprietary, domain-specific knowledge is contributed by individuals and enterprises, verified on-chain and rewarded based on real-world model impact. This approach addresses the growing demand for specialized AI that goes beyond generic LLMs by incentivizing the creation of high-quality, domain-relevant datasets.
The discussion unpacks how OpenLedger’s attribution protocol works, enabling transparent provenance tracking, daily payouts to data contributors and compliance with regulatory requirements such as the EU AI Act and GDPR. Kumar also draws parallels between tokenizing intelligence and tokenizing real-world assets, outlining how on-chain auditability can safeguard against “black box” AI risks while enabling enterprise adoption. From enabling AI agents in DeFi to supporting niche, decentralized model development, the conversation sheds light on why now is the inflection point for merging blockchain infrastructure with AI innovation.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Crypto Trailblazers. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Crypto Trailblazers.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For theCUBE + NYSE Wired: Crypto Trailblazers
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: Crypto Trailblazers.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Crypto Trailblazers. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Crypto Trailblazers.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: Crypto Trailblazers
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: Crypto Trailblazers. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Ram Kumar, OpenLedger
In this insightful episode of the Crypto Trailblazers series hosted by theCUBE, Mike Cagney of Figure Markets sits down with analysts from theCUBE Research to discuss groundbreaking advancements in blockchain technology and their implications for the finance sector. This video is part of the NYSE Wired digital event, aimed at bridging the gap between Silicon Valley and Wall Street by integrating technology and finance.
Cagney, an eminent figure in fintech, shares expertise on the transformative role of blockchain in financial markets during this interview. Conducted by seasoned analysts at theCUBE, the discussion delves into Figure’s innovative contributions, including their blockchain-native loan origination and securitization process. He outlines how Figure leverages blockchain to achieve cost reductions, enhanced security and improved liquidity in financial transactions.
Key takeaways from the interview highlight insights on the evolution of the Web3 ecosystem, such as the emergence of stablecoins as pivotal to transaction processes and the rise of decentralized finance (DeFi). Oltsik states these developments signify a shift towards democratizing finance, wherein truth and transparency are foundational. The conversation concludes with a look at Figure’s pioneering efforts in creating a new financial marketplace utilizing blockchain technology.
#CryptoTrailblazers #FigureMarkets #BlockchainInnovation #Web3 #NYEWired #BlockchainFinance #DecentralizedFinance #Fintech #Stablecoins
Find more SiliconANGLE news and analysis https://siliconangle.com/.
Follow theCUBE's wall-to-wall event coverage https://siliconangle.com/events/
Learn about the latest theCUBE events https://www.thecube.net/
00:00 - Intro
00:05 - Emerging Innovations in Financial Technology and Market Dynamics
02:45 - Key Elements in Financial Ecosystem Dynamics
06:20 - Blockchain: Truth and Transformation
09:39 - Shaping the Future: Innovations in Financial Markets and Stablecoin Integration
13:15 - Enabling the Future: Navigating Disruptions in Banking and Lending
16:51 - Exploring Opportunities and Building Confidence in the Blockchain Ecosystem
>> Hi there. Welcome back to the New York Stock Exchange. You're watching Crypto Trailblazers, the NYSE Wired plus CUBE series. We've been going all week, it's been fantastic. Ethereum has hit a new high this week and we're super excited about that. Ram Kumar is here, he's the core contributor at OpenLedger. We're going to talk about blockchain meets AI. Ram, thanks so much for coming on.>> Absolutely. It's good to be here.
Dave Vellante
>> Why did you start this project?>> That's a good question. So we've been in this space from 2017. We started as a blockchain R&D company and also machine learning services, and we worked with a lot of enterprises. This includes Walmart, Cadbury's, Viacom, and many other brands as well. What we understood, we built a service to provide solutions to enterprise, but there's no product out there, which is really scalable, which can bring together the aspects of data and models. We were seeing ChatGPT happen. It was a huge rage, everyone was talking about ChatGPT, but we understood that AI is not going to be just with ChatGPT, it's not just going to be generic, it's going to become much more specialized. It'll go into very real-world use cases, then you might need AI models which power them to be much more specific and domain-centric to the domain that you're building for, and that's how OpenLedger was started. We understood that, okay, there's a lot of need for proprietary data, but there's no protocol to bring people together, to get rewarded for the data that they provide, and there is no marketplace for all of this to happen, and that's how we started OpenLedger.
Dave Vellante
>> Makes so much sense, the power law we wrote years ago, the power law of GenAI and the domain specificity is really the action is going to be in enterprise AI. I have a note here, you were founded to bring transparency, interoperability and ownership to digital value, and you're pioneering what you call payable AI. Can you talk about what that is? Explain that.>> Yep. So if you take a look at any of these AI models or applications that you use, all of them are trained on internet data. This is the data that we contributed to the internet, we all donated that, we never got back anything out of it. So now is the time to enable users with knowledge, end users, skilled people or enterprises who have knowledge,-to contribute again to AI and to enable proprietary domain specific models to be built. As we were talking about this, there will be real world use cases that will start to emerge for AI, people will build models for trading, people will build AI models for healthcare, models for transport and stuff. For all of this, humans have the knowledge, humans have knowledge, skill on a particular subject, which is not on the internet, and the only way to get that is to increase these humans to contribute that, and only way they would do that is only if they get incentivized, only if you get rewarded for something, you would go ahead and contribute. AI has to become payable, AI has to become rewardable, and that's our motive. So everyone who's looking to build an AI model, if they need access to data, they could enable these users to contribute data on OpenLedger and then have this data to be provided to model developers. Model developers can access that, build these AI models, and use attribution to reward them back. If your data was useful in a model and the model had an inference using your data and the model made money, you should get a piece of that, that's our principle. And we've been doing this over the last two years, we've been building, and we have 20 different projects building on us, we work with enterprises. All of them collected data from users across the globe, and then now they're able to reward them with our token. Once our token is out, they'll be able to reward them, and they'll be able to reward users sitting in US, sitting in South Korea or India, wherever they are. If you have knowledge, you're going to get paid for it.
Dave Vellante
>> I love this story, because so much of the talk is around LLMs and algorithms, and it's amazing to me when I look at these models that are released, GPT-5 just came out, it's going to be months maybe before... It costs whatever, multiple billions to train, and it's only months of competitive advantage before another model comes out. And in fact, they're even struggling to go beyond where they typically have, we've seen that. And everybody's so enamored of that, algorithms and models, where the real value's in the data. We had a joke, we called it LLM communism, where everybody has the same access to the same answers and the same data. But now, you bring in agentic and you bring in the proprietary data, that domain specificity, and there's real value there. So you've called these AI models the next on-chain asset class. What specifically does that mean in financial terms? How are they going to be valued? Is it a two-sided marketplace that values them?>> Yeah, yeah. So with Bitcoin we decentralized money and Ethereum decentralized finance with DeFi and all that. I think we're about to decentralize intelligence. Intelligence is the new thing, it's going to be an era of new intelligence. People are going to build very intelligent systems, and intelligence has to go to every person in the world, everyone has to get access to it, not only people who have money to spend for it, not only conglomerates which can use that, intelligence has to get to people who needs that. It should become a need, like how water and electricity is, I think that's how intelligence has to become. And for that to happen, you need to have a system which is completely transparent, which is completely open and which is trustless, where anyone can contribute for this intelligence, anyone can access knowledge to build intelligence and get rewarded, get paid for it, and people who contributed and were part of this process should get equally distributed and rewarded as part of contributing, either as data or someone is building a model, they have to be rewarded, that's the overall idea. If you start to talk about the larger enterprises out there, let's talk about OpenAI and all these larger enterprises we're building, I'm not here to say that centralized is bad and decentralized is good. Whatever's happening with the centralized marketplaces models and all that, that should still continue to happen. I think consumers will use a lot of centralized models. But then, there should also be ways where a smaller developer who has a unique use case wants to build a model for a very smaller use case, he needs to get access to data, data which is free, he needs to get access to that, build models with that, and then if he makes revenue, then he can share back to the data contributors. That has to happen, we need to enable that. That's where the innovation really comes out, and we want to be that protocol that enables niche specialized models to be built on top of us.
Dave Vellante
>> Again, I love this story. We recently, several, earlier this year, we wrote a piece why Jamie Dimon is Sam Altman's biggest competitor, because Jamie Dimon has all this proprietary data, and then recently, we just wrote how they're going to get there, we could have laid out a roadmap for that. The reason I bring that up is because what you just said is that if I'm not JP Morgan Chase and I don't have their treasure trove of data, but I have some data and I need to enrich that with other data, I would love to get access to that data, facile access, and in return, reward folks that provide that data. That's what you guys are enabling. So it feels like you're changing the economics for people, for institutions, for data providers and for organizations to add value with data.>> I like to humanize AI. As humans, we all learn general knowledge through our schooling and then we specialize in a subject that you want to become an expert in. AI models today are these general knowledge models which have a knowledge about the world because they were trained on internet data. But these models cannot be used as it is. You need to build these models to a particular subject, to a domain, make them expertise in that particular domain and customize them to your own need. And if you want to customize them, then you need to bring in data that is among people, that is among skilled users, that is among people who have knowledge in that particular domain, and we just want to encourage them to come forward, contribute and get paid for it.
Dave Vellante
>> Well, it's interesting, because OpenAI, before they announced GPT-5, announced an open...>> Open source model.
Dave Vellante
>> Open weight model, I'll call it, kind of open source.
Dave Vellante
>> It's entirely open source, yeah.
Dave Vellante
>> Open source, open weight. But my point being, you definitely need open weight to do what you just said, but you also need data, and everybody forgets about the data. I want to ask you what the role of transparency and auditability play. That's the inherent nature of the blockchain. Certainly, we've got a much more friendly United States public policy, but you have EU, there's the EU AI Act, there's the Markets in Crypto-Assets regulations in Europe, so everybody's going to have their own regulations. So how does auditability address the concerns that regulators have?>> There are two parts to this. One, we use transparency and on-chain verifiability to enable people to get rewarded. So every time data is contributed, if it's attributed within that model, your model got access to the data and if your data was used there, you get rewarded, that's one part. On the other side, when we work with enterprises, it's very important we use attribution to prove how data is being used within the model. Because let's assume a healthcare organization uses an AI model and you use this model to get answers, and something goes wrong, a consumer used it, you want to figure out what data caused that inference to happen, that output to happen. Unless you have attribution, you'll not be able to pinpoint to that data point that caused that and you can't remove that poisonous data point. So it's very important that models become much more open, much more explainable and transparent, and that's the other thing that we're building. Apart from using attribution to reward people, we want to use attribution to showcase how a model works, how to open up this black box that is there. And that also comes down to various other regulatory use cases as well. In the EU, there is so much need to comply with GDPR. They really want to know what really happens in these models, what kind of data goes in. So we work with a couple of larger enterprises in Europe which work with us very specifically for this attribution that we're solving because they want to know what really happens in the model. If they're going to rely on an AI, make their decisions based on that AI, they want to know how this AI works, they want to know what data is causing that AI to give their inference. So I think that's very important, making sure the models are open and transparent is very important as it gets to be used in every part of our life.
Dave Vellante
>> Yeah, the openness is critical in that. When you think about a power law, the open source nature of that, it pulls the torso up and changes the dynamics pretty dramatically. Again, the LLM business is a brutal business, and maybe if one of the big model labs achieves AGI, they'll figure out the business model around it at massive scale. But the real value is win the data and then enabling agents to access that data. So how do you see crypto-native AI agents affecting things like DAP composability? Obviously, you see it in finance, but more broadly, enterprise workflows which people are rethinking entirely.>> Yeah. So think of agents to be like actual humans who do the work, the brain being the model and the agents which does the work, you want to make sure that you have the agents on a leash. Like how humans are bound by physical contracts, agents have to be bound by smart contracts, digital contracts. Only then you would know how these agents would work. And then, all these agents are actions of a model's output.,So if something goes wrong, there is a malicious activity, you can actually trace back on us to figure out which data is causing that. You should always figure out what really is causing the impact, there's always a trail-back, and having that model to be open, having transparency on what kind of data sets goes in is very important to make sure that we have these agents in check. I'm sure that agents are going to be the biggest transaction on blockchain, more than humans, there'll be more transactions by agents. And we are at a place where I think we're discovering the potential of agents to be used in trading, especially like DeFi and all that. There is still a lot of human in the loop. I think going forward, we will start to see more of this being completely autonomous, as these models get better, as you're able to build much more domain-specific models which are less hallucinating, and then you could start using them for various use cases.
Dave Vellante
>> When you think about RWA, real-world assets, being tokenized, what are the similarities and differences between real-world assets and intelligence being tokenized?>> Yeah. So in a real-world asset, basically, you're bringing what's happening on the real world, you're bringing that providence on-chain, you're able to track who's the owner, you're able to showcase that on-chain, you're able to see how it's being traded and stuff. Similarly, in intelligence, you're basically showing who has that knowledge, who's the owner of the data, we're able to show the providence of that. We're able to show how it gets into a model and we're able to show whether they're getting rewarded. So we track the problems, how you would track in a real-world asset, and we also make sure this intelligence is tokenized. Only if it's rewarding you, it's useful. Same with RWAs, if it's able to generate yield, it's able to generate rewards, very similarly, we want to generate rewards for the knowledge that you provide. I would say those are the parallels we are trying to bring in, they're tokenizing the asset that is in the real world, we're tokenizing the asset that's in our brain.
Dave Vellante
>> I want to ask you the why now question. Some of the visions that we see, AI, crypto, obviously a lot of hype, great vision, but sometimes the execution isn't always there, why now? What makes payable AI feasible for both a technical and a market perspective?>> Yeah. I think two years before, there was a lot of need for compute, everyone wanted to get access to compute, and there was a lot of decentralized compute companies that came onboard. They actually solved a problem. For us, when we were a start-up two years back, we needed access to compute, and AWS would charge us $7 per hour, and if you use a decentralized compute company, it'll charge us probably $1.50 per hour. So the access to resources was so much high that you're able to bring down the cost so much. So very similarly, there is a need for data now, because models have reached a point where they're too generic, they can't be used in real-world use cases, you need to build models for various aspects of AI, you need to build models of various real-world use cases that you're discovering where AI can be used. Then you need access to a lot of data, not only like internet data, we spoke about this, we need access to proprietary data, which is among people, you, me, and it is among enterprises. I think now, people are getting to know about AI more, like how internet happened and people understood that if they contribute content to internet, like on YouTube and all that, they're going to get paid, the better the content, they're going to get paid more. Very similarly, if they understand more about AI, they would know that if you contribute data, if you contribute knowledge to AI, you're going to get rewarded. So there is a point where there is a demand from the model developers and there is a need for that from the suppliers as well, people have understood that they can monetize their data. So now is the time where data is very much needed, and there's no protocol out there which is connecting the model developers to the data contributors. We've been working on this over the last two years. The biggest problem we're trying to solve is the attribution side. We can build a system where you could say that, "If you contribute data, I'm going to pay you upfront. There's nothing going to be for you to get rewarded more than that." It's going to be like a one-sided marketplace. People are not going to be interested to contribute more and more. For this to happen on a cyclic manner, like how YouTube happened, you need to have a system where they can view how the data is being used, they still retain ownership to that, and they get constantly paid. So for us to crack that, we needed to crack attribution. So that's what it took us two years, and I think now, we are at the right place, there is demand for data, there is knowledge among people, they have awareness about AI, I think just things are falling into place.
Dave Vellante
>> So let's double-click on that. So I'm interested in how OpenLedger deals with the attribution layer, is it just because it's a two-sided market, the market decides, is it algorithmic, a combination?>> So what really happens here is that someone who's looking to build an AI model, they throw up a contract, like a bounty, asking for data to be contributed to that particular model, and they write a rule engine on what kind of data has to be contributed.
Dave Vellante
>> PayX.>> PayX and stuff. And data contributors come forward, contribute data, and their data is accessed, it is processed, and if their data was used in a model, they get paid a portion of the reward that can probably come to them. And then, when the model gets developed and the model gets used on OpenLedger and API is accessible, users get... Like how they access ChatGPT, they use these models, they pay for the inference, app developers get access to the model, they pay for the API access that is there. All the revenue the model developer makes is now shared between model developer and the data contributor, and it's not just shared as it is, it's algorithmically shared. Every time an inference happened, a model throws an output, on that output, we figure out whether the data that was contributed was actually useful, did it have any kind of impact? If it had an impact, we measure the impact that is there and they get paid proportionately. Let's say there is an output coming in, it got paid for $1, and the data that you contributed was 70% of the impact, you'd get probably six cents or five cents out of that, and the remaining goes to the model developer. So all of this happens on-chain, all of this is automated, and that's how attribution really works on top of OpenLedger. We use two papers, one called infini-gram, which is a research paper that was written, it is an open source research paper, another one called data-inf. So it took us almost two years to bring down the attribution time from about 20 seconds to under two seconds. And so, we batch all of this inference together and then we pay out rewards to the data contributors on a daily basis. So that's what really happens over here.
Dave Vellante
>> Okay. So you ensure that the contribution is accurate?>> Yes.
Dave Vellante
>> You essentially calculate it in near real-time, a few seconds, and then you pay out at the end of each day, like a market would?>> Yeah, very similarly.
Dave Vellante
>> Awesome.>> Because we want to be a player where we can enable this to happen in a very cyclic manner. The model developers are there, they need access to data, data the contributors have knowledge, they want to contribute, get paid for it. So we want to be a bridge which lets this happen in a trustless manner, where people contribute data, model developers uses that, they get reward, they get revenue, they get paid, then they share back with the guys who contribute for that.
Dave Vellante
>> So you obviously have exposure to some large enterprises and they're putting big investments into AI, what's your pitch to get them to tokenize their intelligence for their models, their data pipelines, what's the value-add for them to go on-chain?>> Yeah. So there are 100 other fine-tuning companies sitting in Silicon Valley that you're fighting for. The biggest difference between us and them is that we provide attribution, we provide openness of the model. If you're building an AI model on top of OpenLedger, you get access to see how the model works based on how the data is getting used in the model. You get to see the entire provenance on-chain. You're able to view that this is the data that went into the model, you're able to view, if the model throws an output, which particular data caused that. That openness and auditability, no one else provides that, we're the first guys to do that, and enterprises need that. If something goes wrong, they need to trace back and see what really happened there. We work with one of the largest e-commerce players out there, which we'll announce soon about, and we're building a customer support model for them. This AI model is going to power their agent, it's going to replace their chatbot. Now, when customers chat with this and say, "Hey, I lost my item, I couldn't get it, I did not receive it," and stuff, and the model gives out an answer and it is not very accurate and the guy is going to sue you, he wants to really figure out, what really happened here? Why did this AI provide such a bad answer? What was the poisonous data that caused this to happen? They need that auditability. If it's a human, I could see what really went wrong. But with an AI, if I don't have clarity of what's really happening inside, I would not be able to audit this. So for enterprises, making sure this model is open, making sure the data is transparent is very important for them, and that's why we work with large enterprises as well.
Dave Vellante
>> Thank you for that. What's your North Star? At the end of the decade, where do you see this going, what's the vision?>> I think about 10 years, we want to build a platform where every person in the Earth has access to intelligence, and if we're able to even solve 10% of the problem to get to a place where we're able to access people across the globe, that's a very great achievement. I think that in the era of intelligence, the productivity and the things that we're going to achieve is going to be humongous, and we want to be a player where we can make sure that every human gets rewarded or every human gets access to intelligence because of us.
Dave Vellante
>> Well, the TAM is->> Is huge, yeah....
Dave Vellante
>> it's infinite, and it's all intelligence. Ram, thanks for coming on.>> Thank you. It was lovely talking to you.
Dave Vellante
>> Congratulations on getting the project off the ground.>> Thank you.
Dave Vellante
>> We'd love to have you back and watch your progress.>> Thank you.
Dave Vellante
>> Appreciate it. And thank you for watching Crypto Trailblazers, the NYSE Wired plus theCUBE's ongoing series. My name is Dave Vellante. John Furrier is also here. We'll be right back from the New York Stock Exchange right after this break.
>> Hi there. Welcome back to the New York Stock Exchange. You're watching Crypto Trailblazers, the NYSE Wired plus CUBE series. We've been going all week, it's been fantastic. Ethereum has hit a new high this week and we're super excited about that. Ram Kumar is here, he's the core contributor at OpenLedger. We're going to talk about blockchain meets AI. Ram, thanks so much for coming on.>> Absolutely. It's good to be here.
Dave Vellante
>> Why did you start this project?>> That's a good question. So we've been in this space from 2017. We started as a blockchain R&D company and also machine learning services, and we worked with a lot of enterprises. This includes Walmart, Cadbury's, Viacom, and many other brands as well. What we understood, we built a service to provide solutions to enterprise, but there's no product out there, which is really scalable, which can bring together the aspects of data and models. We were seeing ChatGPT happen. It was a huge rage, everyone was talking about ChatGPT, but we understood that AI is not going to be just with ChatGPT, it's not just going to be generic, it's going to become much more specialized. It'll go into very real-world use cases, then you might need AI models which power them to be much more specific and domain-centric to the domain that you're building for, and that's how OpenLedger was started. We understood that, okay, there's a lot of need for proprietary data, but there's no protocol to bring people together, to get rewarded for the data that they provide, and there is no marketplace for all of this to happen, and that's how we started OpenLedger.
Dave Vellante
>> Makes so much sense, the power law we wrote years ago, the power law of GenAI and the domain specificity is really the action is going to be in enterprise AI. I have a note here, you were founded to bring transparency, interoperability and ownership to digital value, and you're pioneering what you call payable AI. Can you talk about what that is? Explain that.>> Yep. So if you take a look at any of these AI models or applications that you use, all of them are trained on internet data. This is the data that we contributed to the internet, we all donated that, we never got back anything out of it. So now is the time to enable users with knowledge, end users, skilled people or enterprises who have knowledge,-to contribute again to AI and to enable proprietary domain specific models to be built. As we were talking about this, there will be real world use cases that will start to emerge for AI, people will build models for trading, people will build AI models for healthcare, models for transport and stuff. For all of this, humans have the knowledge, humans have knowledge, skill on a particular subject, which is not on the internet, and the only way to get that is to increase these humans to contribute that, and only way they would do that is only if they get incentivized, only if you get rewarded for something, you would go ahead and contribute. AI has to become payable, AI has to become rewardable, and that's our motive. So everyone who's looking to build an AI model, if they need access to data, they could enable these users to contribute data on OpenLedger and then have this data to be provided to model developers. Model developers can access that, build these AI models, and use attribution to reward them back. If your data was useful in a model and the model had an inference using your data and the model made money, you should get a piece of that, that's our principle. And we've been doing this over the last two years, we've been building, and we have 20 different projects building on us, we work with enterprises. All of them collected data from users across the globe, and then now they're able to reward them with our token. Once our token is out, they'll be able to reward them, and they'll be able to reward users sitting in US, sitting in South Korea or India, wherever they are. If you have knowledge, you're going to get paid for it.
Dave Vellante
>> I love this story, because so much of the talk is around LLMs and algorithms, and it's amazing to me when I look at these models that are released, GPT-5 just came out, it's going to be months maybe before... It costs whatever, multiple billions to train, and it's only months of competitive advantage before another model comes out. And in fact, they're even struggling to go beyond where they typically have, we've seen that. And everybody's so enamored of that, algorithms and models, where the real value's in the data. We had a joke, we called it LLM communism, where everybody has the same access to the same answers and the same data. But now, you bring in agentic and you bring in the proprietary data, that domain specificity, and there's real value there. So you've called these AI models the next on-chain asset class. What specifically does that mean in financial terms? How are they going to be valued? Is it a two-sided marketplace that values them?>> Yeah, yeah. So with Bitcoin we decentralized money and Ethereum decentralized finance with DeFi and all that. I think we're about to decentralize intelligence. Intelligence is the new thing, it's going to be an era of new intelligence. People are going to build very intelligent systems, and intelligence has to go to every person in the world, everyone has to get access to it, not only people who have money to spend for it, not only conglomerates which can use that, intelligence has to get to people who needs that. It should become a need, like how water and electricity is, I think that's how intelligence has to become. And for that to happen, you need to have a system which is completely transparent, which is completely open and which is trustless, where anyone can contribute for this intelligence, anyone can access knowledge to build intelligence and get rewarded, get paid for it, and people who contributed and were part of this process should get equally distributed and rewarded as part of contributing, either as data or someone is building a model, they have to be rewarded, that's the overall idea. If you start to talk about the larger enterprises out there, let's talk about OpenAI and all these larger enterprises we're building, I'm not here to say that centralized is bad and decentralized is good. Whatever's happening with the centralized marketplaces models and all that, that should still continue to happen. I think consumers will use a lot of centralized models. But then, there should also be ways where a smaller developer who has a unique use case wants to build a model for a very smaller use case, he needs to get access to data, data which is free, he needs to get access to that, build models with that, and then if he makes revenue, then he can share back to the data contributors. That has to happen, we need to enable that. That's where the innovation really comes out, and we want to be that protocol that enables niche specialized models to be built on top of us.
Dave Vellante
>> Again, I love this story. We recently, several, earlier this year, we wrote a piece why Jamie Dimon is Sam Altman's biggest competitor, because Jamie Dimon has all this proprietary data, and then recently, we just wrote how they're going to get there, we could have laid out a roadmap for that. The reason I bring that up is because what you just said is that if I'm not JP Morgan Chase and I don't have their treasure trove of data, but I have some data and I need to enrich that with other data, I would love to get access to that data, facile access, and in return, reward folks that provide that data. That's what you guys are enabling. So it feels like you're changing the economics for people, for institutions, for data providers and for organizations to add value with data.>> I like to humanize AI. As humans, we all learn general knowledge through our schooling and then we specialize in a subject that you want to become an expert in. AI models today are these general knowledge models which have a knowledge about the world because they were trained on internet data. But these models cannot be used as it is. You need to build these models to a particular subject, to a domain, make them expertise in that particular domain and customize them to your own need. And if you want to customize them, then you need to bring in data that is among people, that is among skilled users, that is among people who have knowledge in that particular domain, and we just want to encourage them to come forward, contribute and get paid for it.
Dave Vellante
>> Well, it's interesting, because OpenAI, before they announced GPT-5, announced an open...>> Open source model.
Dave Vellante
>> Open weight model, I'll call it, kind of open source.
Dave Vellante
>> It's entirely open source, yeah.
Dave Vellante
>> Open source, open weight. But my point being, you definitely need open weight to do what you just said, but you also need data, and everybody forgets about the data. I want to ask you what the role of transparency and auditability play. That's the inherent nature of the blockchain. Certainly, we've got a much more friendly United States public policy, but you have EU, there's the EU AI Act, there's the Markets in Crypto-Assets regulations in Europe, so everybody's going to have their own regulations. So how does auditability address the concerns that regulators have?>> There are two parts to this. One, we use transparency and on-chain verifiability to enable people to get rewarded. So every time data is contributed, if it's attributed within that model, your model got access to the data and if your data was used there, you get rewarded, that's one part. On the other side, when we work with enterprises, it's very important we use attribution to prove how data is being used within the model. Because let's assume a healthcare organization uses an AI model and you use this model to get answers, and something goes wrong, a consumer used it, you want to figure out what data caused that inference to happen, that output to happen. Unless you have attribution, you'll not be able to pinpoint to that data point that caused that and you can't remove that poisonous data point. So it's very important that models become much more open, much more explainable and transparent, and that's the other thing that we're building. Apart from using attribution to reward people, we want to use attribution to showcase how a model works, how to open up this black box that is there. And that also comes down to various other regulatory use cases as well. In the EU, there is so much need to comply with GDPR. They really want to know what really happens in these models, what kind of data goes in. So we work with a couple of larger enterprises in Europe which work with us very specifically for this attribution that we're solving because they want to know what really happens in the model. If they're going to rely on an AI, make their decisions based on that AI, they want to know how this AI works, they want to know what data is causing that AI to give their inference. So I think that's very important, making sure the models are open and transparent is very important as it gets to be used in every part of our life.
Dave Vellante
>> Yeah, the openness is critical in that. When you think about a power law, the open source nature of that, it pulls the torso up and changes the dynamics pretty dramatically. Again, the LLM business is a brutal business, and maybe if one of the big model labs achieves AGI, they'll figure out the business model around it at massive scale. But the real value is win the data and then enabling agents to access that data. So how do you see crypto-native AI agents affecting things like DAP composability? Obviously, you see it in finance, but more broadly, enterprise workflows which people are rethinking entirely.>> Yeah. So think of agents to be like actual humans who do the work, the brain being the model and the agents which does the work, you want to make sure that you have the agents on a leash. Like how humans are bound by physical contracts, agents have to be bound by smart contracts, digital contracts. Only then you would know how these agents would work. And then, all these agents are actions of a model's output.,So if something goes wrong, there is a malicious activity, you can actually trace back on us to figure out which data is causing that. You should always figure out what really is causing the impact, there's always a trail-back, and having that model to be open, having transparency on what kind of data sets goes in is very important to make sure that we have these agents in check. I'm sure that agents are going to be the biggest transaction on blockchain, more than humans, there'll be more transactions by agents. And we are at a place where I think we're discovering the potential of agents to be used in trading, especially like DeFi and all that. There is still a lot of human in the loop. I think going forward, we will start to see more of this being completely autonomous, as these models get better, as you're able to build much more domain-specific models which are less hallucinating, and then you could start using them for various use cases.
Dave Vellante
>> When you think about RWA, real-world assets, being tokenized, what are the similarities and differences between real-world assets and intelligence being tokenized?>> Yeah. So in a real-world asset, basically, you're bringing what's happening on the real world, you're bringing that providence on-chain, you're able to track who's the owner, you're able to showcase that on-chain, you're able to see how it's being traded and stuff. Similarly, in intelligence, you're basically showing who has that knowledge, who's the owner of the data, we're able to show the providence of that. We're able to show how it gets into a model and we're able to show whether they're getting rewarded. So we track the problems, how you would track in a real-world asset, and we also make sure this intelligence is tokenized. Only if it's rewarding you, it's useful. Same with RWAs, if it's able to generate yield, it's able to generate rewards, very similarly, we want to generate rewards for the knowledge that you provide. I would say those are the parallels we are trying to bring in, they're tokenizing the asset that is in the real world, we're tokenizing the asset that's in our brain.
Dave Vellante
>> I want to ask you the why now question. Some of the visions that we see, AI, crypto, obviously a lot of hype, great vision, but sometimes the execution isn't always there, why now? What makes payable AI feasible for both a technical and a market perspective?>> Yeah. I think two years before, there was a lot of need for compute, everyone wanted to get access to compute, and there was a lot of decentralized compute companies that came onboard. They actually solved a problem. For us, when we were a start-up two years back, we needed access to compute, and AWS would charge us $7 per hour, and if you use a decentralized compute company, it'll charge us probably $1.50 per hour. So the access to resources was so much high that you're able to bring down the cost so much. So very similarly, there is a need for data now, because models have reached a point where they're too generic, they can't be used in real-world use cases, you need to build models for various aspects of AI, you need to build models of various real-world use cases that you're discovering where AI can be used. Then you need access to a lot of data, not only like internet data, we spoke about this, we need access to proprietary data, which is among people, you, me, and it is among enterprises. I think now, people are getting to know about AI more, like how internet happened and people understood that if they contribute content to internet, like on YouTube and all that, they're going to get paid, the better the content, they're going to get paid more. Very similarly, if they understand more about AI, they would know that if you contribute data, if you contribute knowledge to AI, you're going to get rewarded. So there is a point where there is a demand from the model developers and there is a need for that from the suppliers as well, people have understood that they can monetize their data. So now is the time where data is very much needed, and there's no protocol out there which is connecting the model developers to the data contributors. We've been working on this over the last two years. The biggest problem we're trying to solve is the attribution side. We can build a system where you could say that, "If you contribute data, I'm going to pay you upfront. There's nothing going to be for you to get rewarded more than that." It's going to be like a one-sided marketplace. People are not going to be interested to contribute more and more. For this to happen on a cyclic manner, like how YouTube happened, you need to have a system where they can view how the data is being used, they still retain ownership to that, and they get constantly paid. So for us to crack that, we needed to crack attribution. So that's what it took us two years, and I think now, we are at the right place, there is demand for data, there is knowledge among people, they have awareness about AI, I think just things are falling into place.
Dave Vellante
>> So let's double-click on that. So I'm interested in how OpenLedger deals with the attribution layer, is it just because it's a two-sided market, the market decides, is it algorithmic, a combination?>> So what really happens here is that someone who's looking to build an AI model, they throw up a contract, like a bounty, asking for data to be contributed to that particular model, and they write a rule engine on what kind of data has to be contributed.
Dave Vellante
>> PayX.>> PayX and stuff. And data contributors come forward, contribute data, and their data is accessed, it is processed, and if their data was used in a model, they get paid a portion of the reward that can probably come to them. And then, when the model gets developed and the model gets used on OpenLedger and API is accessible, users get... Like how they access ChatGPT, they use these models, they pay for the inference, app developers get access to the model, they pay for the API access that is there. All the revenue the model developer makes is now shared between model developer and the data contributor, and it's not just shared as it is, it's algorithmically shared. Every time an inference happened, a model throws an output, on that output, we figure out whether the data that was contributed was actually useful, did it have any kind of impact? If it had an impact, we measure the impact that is there and they get paid proportionately. Let's say there is an output coming in, it got paid for $1, and the data that you contributed was 70% of the impact, you'd get probably six cents or five cents out of that, and the remaining goes to the model developer. So all of this happens on-chain, all of this is automated, and that's how attribution really works on top of OpenLedger. We use two papers, one called infini-gram, which is a research paper that was written, it is an open source research paper, another one called data-inf. So it took us almost two years to bring down the attribution time from about 20 seconds to under two seconds. And so, we batch all of this inference together and then we pay out rewards to the data contributors on a daily basis. So that's what really happens over here.
Dave Vellante
>> Okay. So you ensure that the contribution is accurate?>> Yes.
Dave Vellante
>> You essentially calculate it in near real-time, a few seconds, and then you pay out at the end of each day, like a market would?>> Yeah, very similarly.
Dave Vellante
>> Awesome.>> Because we want to be a player where we can enable this to happen in a very cyclic manner. The model developers are there, they need access to data, data the contributors have knowledge, they want to contribute, get paid for it. So we want to be a bridge which lets this happen in a trustless manner, where people contribute data, model developers uses that, they get reward, they get revenue, they get paid, then they share back with the guys who contribute for that.
Dave Vellante
>> So you obviously have exposure to some large enterprises and they're putting big investments into AI, what's your pitch to get them to tokenize their intelligence for their models, their data pipelines, what's the value-add for them to go on-chain?>> Yeah. So there are 100 other fine-tuning companies sitting in Silicon Valley that you're fighting for. The biggest difference between us and them is that we provide attribution, we provide openness of the model. If you're building an AI model on top of OpenLedger, you get access to see how the model works based on how the data is getting used in the model. You get to see the entire provenance on-chain. You're able to view that this is the data that went into the model, you're able to view, if the model throws an output, which particular data caused that. That openness and auditability, no one else provides that, we're the first guys to do that, and enterprises need that. If something goes wrong, they need to trace back and see what really happened there. We work with one of the largest e-commerce players out there, which we'll announce soon about, and we're building a customer support model for them. This AI model is going to power their agent, it's going to replace their chatbot. Now, when customers chat with this and say, "Hey, I lost my item, I couldn't get it, I did not receive it," and stuff, and the model gives out an answer and it is not very accurate and the guy is going to sue you, he wants to really figure out, what really happened here? Why did this AI provide such a bad answer? What was the poisonous data that caused this to happen? They need that auditability. If it's a human, I could see what really went wrong. But with an AI, if I don't have clarity of what's really happening inside, I would not be able to audit this. So for enterprises, making sure this model is open, making sure the data is transparent is very important for them, and that's why we work with large enterprises as well.
Dave Vellante
>> Thank you for that. What's your North Star? At the end of the decade, where do you see this going, what's the vision?>> I think about 10 years, we want to build a platform where every person in the Earth has access to intelligence, and if we're able to even solve 10% of the problem to get to a place where we're able to access people across the globe, that's a very great achievement. I think that in the era of intelligence, the productivity and the things that we're going to achieve is going to be humongous, and we want to be a player where we can make sure that every human gets rewarded or every human gets access to intelligence because of us.
Dave Vellante
>> Well, the TAM is->> Is huge, yeah....
Dave Vellante
>> it's infinite, and it's all intelligence. Ram, thanks for coming on.>> Thank you. It was lovely talking to you.
Dave Vellante
>> Congratulations on getting the project off the ground.>> Thank you.
Dave Vellante
>> We'd love to have you back and watch your progress.>> Thank you.
Dave Vellante
>> Appreciate it. And thank you for watching Crypto Trailblazers, the NYSE Wired plus theCUBE's ongoing series. My name is Dave Vellante. John Furrier is also here. We'll be right back from the New York Stock Exchange right after this break.
Dave Vellante