The Real Reason Crypto Wallets Are Racing to Integrate AI with Ram Kumar & Eowyn Chen

Transcript

Welcome back to the Rowup. We are your hosts. I'm Andy. I'm Robbie. And this is your home to learn from the biggest guest, the brightest minds at the tip of innovation. Our goal is to help spread the collective knowledge of everyone in our community. We are breaking crypto out of its echo chamber and taking it to the real world. All righty, guys. Welcome back to the Rollup TV. We've got Ram and Owen on today to talk about the intersection of consumer applications, wallet interfaces, and of course AI. Looking forward to this conversation coming off the the news yesterday from Coinbase's big push. Uh there's a lot of competition, a lot of fun stuff happening in the consumer wallet, consumer app space. Uh and I think a lot of people are are getting onboarded to

the space. Uh as of now, market's hot. So guys, excited for you guys to be here. just do a quick round of intros um and how you guys are feeling about the markets. Uh Ram, why don't we start with you, man? Welcome back. >> Thanks, man. This is Rah. I'm one of the core contributors at Open Ledger. Market is pretty wild. Uh glad to see uh some green charts. Uh I'm one of the core contributors here. We build um enabling basically AI to be much more transparent. We're a blockchain specific for enabling users to contribute data and build AI models. and we want to reward people for contributing data and uh we've been building the space over the last 2 years. Personally, I've been in the space from 2017 and uh started as

a blockchain R&D company, worked with a lot of enterprises including like Walmart, um many more brands as well and then open ledger was an incubation which now is an end toend protocol that is enabling more than 20 plus projects to build AI models. We do work with again enterprises as well. We work with Walmart. We work with um Dubai tax authority and many more brands as well. And um we are also going to work with Trust Wallet. So I'm glad to be here and talking with you guys on this. >> Sweeten Awen. Uh so I'm from Trust Wallet. Right now we're the largest mobile wallet at the moment. We have about more than 200 million downloads globally uh for the

past seven years. And then from the active wallet perspective on mobile, we're also right now the biggest um globally out there. So it's definitely right now a very interesting time. There are a lot more experiments happening in the consumer side um for how to further onboard uh more users uh into our crypto space. Uh general macro environment are very much supported and uh glad to see the green charts as well. Uh hopefully every elements are pointing to to the right environment that we can get some uh good growth um for the uh mass adoption in our industry.

>> Definitely. And uh I've been loving the uh Pudgy Penguins that you guys have been rocking on Twitter and uh promoting. Those are quite the engagement hack. It's good to have you guys both here. Um and I actually I want to start this conversation with a with a a phenomenon that's been happening that I think a lot of people are starting to uh become accustomed to which is this idea around uh you know kind of distribution in our space being very very uh important and ultimately king of uh you know user interactions user relationships and I think this also ties into some of these you know chatgbtes interfaces and kind of loss that uh I think we'll be able to get into today And so, um, Owen, when you guys have built Trust Wallet, you guys have been

around for quite some time and you've seen, you know, the progression from this kind of fat protocol thesis, right? A a lot of value accurencies, different L1s using the same type of wallets, trust wallets, etc., uh moving towards applications and now even moving all the way out to the end of the stack to what some are calling this kind of fat wallet thesis uh where you know those who own the the user onboarding can then kind of uh verticalize bring swaps bring per trading NFT different types of social interactions all into the place where people on board originally um and so this has really be become a prominent view of a lot of investors um and believers in in this space. And so, you

know, I'm curious from your persp perspective, from Trustwalt's perspective, how you guys are are are seeing this play out from your user data, um, you know, and just kind of what you're seeing from the, you know, ownership of users, why that's valuable, um, and how you've kind of grown your your product suite to be able to accommodate more and more things to keep people in trustw um, you know, and and be a part of your core product suite. >> Yeah, that's a really good question. Um I I think that the narrative is often taking different sides but to be honest by being the industry so much you realize that different value chain does create different values and they take the different values. Uh it's it it looks very promising that if you own the

user relationship that you can seem to be taking a bigger cut or more values with the users but people also some often underestimate how hard it is to do retail user businesses. um a lot of the customer support, a lot higher liability from the security concerns. Um and then this is a whole level different kind of uh mental stress game to some u compared to if you say hey I can be a really good niche beautiful product and I would just say that hyperlquid is a classic example right I'm going really vertically niche in this area and I can do do really well and then if we're serving the retail users and then we have to be super adaptive because people want all

different things all the time then sometimes you might not necessarily have a very uh fixed road map. Hey, I'm going to do A and then market changes, industry growing, these are all positive things. You feel like you're you're scrapping all your road map and trying to do product B. So if you're actually user focused that's you want to do but to a lot of the builders being able to change all the time is really really hard. So uh I think it's just from the practitioners perspective you know grass is always greener like infrastructure builder feels like oh if you're on the application side or the user interface side then this is a lot cooler maybe we'll feel the same way um but I the reason why I'm in this product and why I'm doing this business is I think

fundamentally it just matters to me that accuring value to users and humans is one of the most rewarding and valuable thing and in that front uh having the being the UI for the industry right like being the face being the user experiences it carries a lot of blame carries a lot of opportunities but ultimately you do we do have a direct relationship with the users and that to a lot of the extent is because of our values that we cherish it and then it's it's doing great um from the business side doesn't mean that all the values acrew to the front end it's not always the same I'll just be very honest right so and that should be the case Because if the whole value chain, just certain

parts are getting a lot more value versus the other, just like a human, if your left leg is way stronger than your right leg, then you're not going to walk very well as a grown-up. So, I actually think that being a bit fair across the value chain and do the things together so that every part of the muscles as a collective framework can grow together is is way more important. But of course, I love the thesis that, you know, it's a fat wallet like everybody would love to be feeling important and then good to see that's kind of happening to the front end side.

>> Introducing asset runes, the future of decentralized finance, stable coins natively on Bitcoin. Hedge against volatility. Diversify your portfolio. All directly from your Bitcoin wallet. Secured by ZK technology. This is asset runes available now exclusively on Starknet. >> Yeah. Yeah. Ram, I'm curious from your perspective if you feel similarly with the evolution of of some of the AI tools and how you guys view it from open ledger, right? Like there are obviously large uh platforms that have a lot of this kind of like AI usage. It's quite different from crypto usage though. Um

how are you seeing it on your end? And then we can we can kind of tie these two ideas of thought together into kind of combining these two industries. >> Yeah, I think AI has been predominantly very generic in the last 2 years. If you take a look at chat GBT, right, it's more of an app which is which is predominantly used by people to ask very generic questions. Um lately if you take a look at it over the last six to 6 months to year we've seen people actually use for products uh actually use for productivity and uh it's slowly getting into almost every other part of our life right it could be personal and professional life as well um you know we've started optimizing our uh hiring process because you know if we could have AI do that we pretty much wouldn't

want to hire someone for that right so I guess that's that's the momentum we are trying to see we want we all want to see that it's not that people are going to lose jobs. People are going to be having more better jobs to do. But I think this enables AI to become much more broader than just asking about you know where do I get the best sushi, right? So um I think now we are at a stage where like how Google search started, Google search was very generic and internet was born people started using the internet and then we also wanted to figure out okay how can we use a search for finding a job which is LinkedIn. how can we use search for you know finding content which is like YouTube and stuff like that how can you use search for like food and stuff so I think very similarly

AI is getting into that phase where it's becoming much more specialized where um we're building AI for very specific use cases real life use cases and that's what we want to solve as well um just charge and any other AI applications out there they would be there think of them to be like an OS and which is which is like the predominant place where you would find most interactions but you would also find very similar specific apps for a particular use case which are going to be like specialized models and that's what we want to enable. We want to enable people to bring in their own data figure out unique use cases and build like models for it. Like when we in fact when we were talking to trustwlet right when we talk spoke to the team they could they were trying to

explore integrating AI into their um wallet experience and the biggest problem they had was that they could not use chat GPT or any other generic AI as it is. is not really providing the solution that they required right that's one of the reasons where we had a conversation with them and then you know we understood that there is need for a specialized model to be built like this like every specific use cases that we take a look at it we can't use a generic AI model as it is we would need to customize it for our own need and we provide an you know and protocol and an application for people to do that >> yeah I mean I I I I kind of keep having this thought coming back to this idea of chain abstraction and application abstra action really just being oriented around

users using some sort of specialized model under the hood of this kind of AI assistant per se in able to interact on chain on the user's behalf kind of on behalf of their intent or their desire. I is this like kind of what we think is going to be the the future of some of these wallet interfaces like Trust Wallet? Is it going to be uh possible to be able to have an application like Trust Wallet and be able to utilize all these different DeFi primitives and things that you'd want to do onchain just by using kind of like a natural language model or is that is that a far-fetched thought?

>> I think it depends on the users really because ultimately as a user interface uh we adapted to what the users want. Um so if users want and then I actually see quite a lot of the experiments on the market already that try to have a chat interface and then uh the the the real interesting question is that the adoption may still be slow um because if when people are already familiar with a user interface like here's where I click the button that's faster for me than if I had to go into another chat interface.

But who knows maybe the next generation of the new users will find that way more helpful because that's easier to onboard compared to the traditional analog interface that they might still not getting familiar with. Um I think that it so then from product builder perspective what are we going to do? I think the key is to leave room for experiments and see how that plays out. uh and that's what we're planning to do and that's why we're working with RAM's team um potentially how to potentially do those experiments uh in an effective and efficient way so that we can learn more we can learn users more and we can evolve with the users a bit better um so that's on the user interface part but then AI can be used below the interface level such as what you're talking about

right like can AI can do a m much better job when analyzing all the data online or onchain if we give them the access to or empower them to they probably can find the better defy strategy than what you usually are integrating with. Then can they be the workers behind the scene that actually do those work and then just presented it in a way that however humans would love it to read. I think that's highly likely possible. Uh and you will couple with other kind of tech infra to enable them to do this effectively and efficiently. Um but at the same time uh we're also working on some of the solutions that doesn't necessarily require AI to do the same thing that can also do the autobalancing finding the better solutions of the yield if it's available on the network that we're supported. So we will explore

the how uh like AI is a how other solutions backend smart contract is also a how solutions ultimately it will just coming down to be whichever approach are preferred by the users and give the best performance as an example we're in this transitional phase and that's the most interesting part right see how the users and how the industries and the society evolve with it maybe at the end of the day it's going to be a gentic society that you know AI agent does the most of the job but not yet. So how do we get to this uh and how does the process look like is one of the fun journey I feel that we have right now.

I also feel like I don't think this chat interface will be uh the way for AI like forever right like if you take a look at it we even I think even spoke about this it started with like terminals like you know uh when initially PC era started we were all interfacing on the PC using a terminal that's how AI is today right it sort of feels like a terminal where you talk to it and then it it gives you solutions I think that will evolve it'll become much more GUId driven and we would see UI being optimized using AI, right? So I think um for wallet um you know it would start with chat interface as you know as we discuss here but it also eventually evolve into much better use cases and there are two major use cases we all think right onboarding the

normies where you provide them with a better interface so that normies can access wallets and then uh the other way is basically how can we enable better usage for the existing users like how can we stop fishing how can like I've seen like even today like there are people who send tokens to the wrong network right how Can that be stopped using AI? So these are like very simple use cases which can be solved first which could save millions and billions of dollars and then probably we get into the onboarding normies and stuff like that. >> And Ram are are there specialized models that are required to be able to fulfill some of these uh cryptospecific use cases.

>> So it depends on the use case right. So it's it let's say first usually what people do is like they start with prompt engineering where they take an existing model give better prompts and then you know you have a model that is built for the prompt that you engineer it with like for example most of AIXBT and all these agents that we initially saw were predominantly prompt engineering right and then people started implementing rag which is like providing additional knowledge to your model and then broading solution the last solution usually is like basically fine-tuning the model and making it specialized Right. So it comes down to the use case that is there. Um if you want to build let's say an AI dedicated for swapping, right? So how it can optimize swapping,

obviously you can't use an existing model there. You might need to optimize the model for this particular use case. And uh if if let's say someone wants to build an AI which is going to help you lend better and provide you solutions around that um you know you might be able to you don't need to use a larger language model as you can use a very very smaller model and then optimize your lending based on that. So it comes down to the use case and uh how you want to optimize for that use case. >> Got it. >> And also it also comes down to the people who deal with it right. It comes down to actually the developer who what he chooses for the budget that is there and stuff like that. So there's a lot of factors into it.

>> Cool. Um all right guys, so kind of moving to one of the important topics that I think that users will want to know about is around some of the you know security and guard rails when using these kind of uh AI tools within something like Trust Wallet or just generally onchain. Um Owen, you you've actually recently compared self-custody itself to the Second Amendment uh around ownership of uh of guns and different types of um things in the states. Uh I'm curious, you know, how you got how you're thinking about, you know, wallets coming up with training modes, risk flags, AI guard rails, um for this kind of next step in terms of integrating AI,

you know, to take either it's to take some sort of ownership of assets to be able to use them on users behalf. just, you know, trying to curious how you're thinking about these two worlds of the importance of self- custody combining with wanting to get users a really solid uh, you know, best yields, great experience using these AI tools. How do you think about this from a security perspective for for Trust Wallet? >> Yeah. Yeah. Great question. So, if I may go back to the gong analogy, right? Like, so >> I had to bring it up.

>> Yeah. like self self- custody is not easy for a lot of people if you wanted to have pure self- custody right like it comes with huge responsibilities and then many people to be honest in this modern world doesn't know how to take responsibility and um in those cases I would say that like just like some people that they might not be able to handle the gun very well it's like just don't don't don't do it um so it's not for everybody but then it may be helpful in some extreme cases and really protect people. Um, then what does self- custody mean? I think it always needs to give people a way out, right? Go back to the first principle is that no one else can move the money for you except for that

you're fully in control of your own fund. So the key is for us to make sure that that self- custody as a custody solutions will always be available for people. But then there are additional services uh feature services on top of it that people may opt in to use. And then th those things are always have a trade-off. How much do you want to have convenience versus um security. Um of course from the product builder perspective we do our best to uh evaluate uh security control codes operations all things that can have for people. But for us the final principle is that users need to make the choice.

And I do think that this might come a little bit more tricky or difficult in the age is how much uh choice and um options that users want to delegate to agents or AI to do and then are they okay with a potential liability if that thing comes to happen that's there's always a choice behind the scene right so um I think ultimately what we can do is that we don't force any options to people but we make the the the options very clear as options to people and personally I believe that humans still need to learn how to be responsible because if we completely delegate everything else to everybody then what's

the purpose for you to do anything. Um but that's going to be a personal choice for a lot of people. But what we need to do best is to present those options and choice to people and make it clear for them and then helping people aiding them to make the best choice uh that they wanted to own in the end. Um this is a general principle but specifically on specific uh like product features or AI utilization. I think a lot of the devils might be in the details on how to figuring that out. >> Um we will see. I think right now we haven't got challenged too much with it.

I think our industry doesn't even have too much of the dilemas to present to the users at the moment. Um so we will hold on to the principles that I was talking about and then evaluating specific option situations as we uh come along with them. >> Innovating finance global institutional and onchain deep liquidity to drive capital efficiency. Now maximize yields on institutionalgrade liquid staking with M. Unlock trillion dollars of capital via F Bitcoin.

All of this secured by Mantle's treasury of billions in stables and ETH capital pools. This entire ecosystem is coordinated by a vast network of seasoned VCs that are aiming to create the next generation of banking and financial technology. This is Mantle. >> Yeah, Ram kind of building on that, right? So, you know, you you have this need within Trust Wallet to have this very secure uh way for users to, you know, have optionality to interact with different models or have different preferences for custody, for risk, etc.

And kind of flipping that back on his head towards the the you know, kind of the deeper question here is how are these models trained? Which models do you use? Which agents do you use? And you you've actually recently called AI a trillion dollar industry built on unpaid labor, which I assume is around basically all the data that this that these models have been trained on, scraping the internet, um you know, basically going around IP laws, right? All that would would come into something that a user at the end of the day who has something like like a trust wallet, like who uses chat GBT ends up being exposed to whether they like it or not.

I'm curious what you think one concrete thing that crypto wallets should do to make this kind of attribution back to this training data real. How should they go about uh providing this this attribution to what you know where these models had been trained on? And kind of how do you think about that from you know if we're taking what what Owen's saying with regards to her perspective broadly and then specifically you know kind of going deeper down the stack to where the these models are actually coming from. Um I think I remember having this conversation with Yan um while I think 3 four months back right we were discussing about how wallet is your identity like wallet is your face for you to go across any kind of digital inter intersections that you have right

>> uh I totally believe that right even like in this case right what we want to enable is to make sure that everyone who owns data have to be rewarded and only way I can get rewarded if I have a digital wallet that will enable me to accept any kind of rewards across the globe So that's what we want to enable I think like as as I spoke about right um there was massive opportunity that was taken away from us you know there's so much data that we consumed on the internet so much data that we provided which has been scraped by the AI uh you know larger angle model that is out there so much information has been given we can't do anything about it because you know internet is free you can scrape the internet you don't have to actually reward people right but but this AI is

going to as we spoke about before AI is going to be much more specific AI is going go into very specific niches where human knowledge um skill labor is actually would be become useful, right? Very similarly how like when the internet needed content, people who had skill, people who could create videos, people who could write provided the content on the internet and they got rewarded it either by users or they got paid. Right? Very similarly in AI we need to have an interface where if you have knowledge you can you know basically infer to that uh knowledge. Basically what you can do is basically you can contribute your knowledge and you should get paid for it. U that's the thesis that we want to drive because where would you get the knowledge like if uh today scale has been invested about $14 billion by meta

and the reason for that is because scale is getting into proprietary data sets right they're the ones who actually label all in data sets that's why large language models today exist because there was data better data to get access to now scale is working on proprietary data sets and all of the AI companies need access to that and what better way to do that is to go ahead and acquire them very similarly I think uh if scale is going to do that in a centralized manner it's going to probably collect data from you and then reward you once. We want to provide a platform where you can contribute your data and constantly get rewarded like how would it happen on the internet right even though internet was supposed to be decentized it did not happen but it was at least a two-way street you contribute something you get

get benefited for that AI is not like that it's not a two-way street right you contribute you don't get anything rewarded for that right so we want to change that we want to make sure that people if they contribute let's say any kind of knowledge it could be um you know Andy knowing about web 3 what really happened in the last 6 plus 7 years that you were in the industry um how you saw the industry from your eyes. How you saw saw about crypto sentiment, right? You could contribute that to a model. Uh someone who has been a developer, let's say a solidity developer could contribute his knowledge about solitary model. So he needs to get rewarded for that, right? He only his knowledge should not be taken. He needs to be part of the process. He needs to be part of the revenue that comes in.

It's not about getting paid once. You have to be constantly getting paid because your knowledge is being constantly used and it's being made to improvise the model. it's being made to give an inference and output to let's say someone asking question to an AI model. So every time that AI makes a revenue, you should be part of it because your knowledge is what made it possible. Like it's very similar to like how we how humans work, right? We've learned so much um but we never kind of like um know reward for the knowledge that we got. But AI is not like that. It's it's getting its knowledge from the people that we have. So that has to change. That's our overall idea. Um you know if you take a look at hugging face which is uh one of the largest platforms out there which has humongous of data

and models out there is not really making much bigger change because 90%age of the data set that is there is useless right because no one gets rewarded is just open source contribution. Open source is great but it has to be monetized. Uh that's what we want to enable. We want to have better contributions to AI so that better models can be built. If better models are built then you get paid than just be providing like a free source. So um that's what we believe in and the interface for that should come from a wallet right um it cannot be let's say myself let's say as a human I can let's say I go contribute I cannot be paid in fiat if someone is going to use my data and build a model in South Korea right the only way to access that is to have a

wallet a crypto wallet which will enable me to get rewarded from anywhere whichever part of the world right and I think that's truly what will be like a digital identity for a human for him to contribute any kind of uh knowledge. It could be um you know some contribution to an AI model or it could be a feedback to an AI model. The way he gets rewarded is using tokens and the interface for that is a wallet. And that's that's what I that's what we believe in and that's why this protocol enables anyone to just plug in using their wallet, be part of our network, contribute something and pay for that.

>> Got it. Yeah. Yeah, I mean the the economic models of of the pre- AI internet have drastically changed in this era of abundant content, abundant ideas, abundant kind of uh information and then now you know as these worlds of of of AI and artificial intelligence have started to interact with these incentive mechanisms and kind of uh cryp uh crypto and blockchain uh ways of creating economic value. a lot of the, you know, a lot of the ways that we've thought about building these systems in the past, just 10 years ago, have have changed drastically. And so, it's going to be it's going to be quite interesting to to kind of watch this continue to unfold. Uh, again, I think a lot of this

is going to come down to, you know, creating the the best most secure user interface for interacting with crypto. um and you know those who can you know get the you know get verifiability into these models into what you're in uh interacting with uh amongst other kind of core core ideas. So guys, as we come to a wrap, um you know, I'm I'm generally just most curious what what you guys are most excited about within your current uh roadmap, you know, what's on the uh what's on the agenda for for the next quarter for Trust Wallet, for Open Ledger. What can you share with our audience today that you know that will get them excited about what you guys are building? Download

Trust Wallet, participate in Open Ledger. What are you guys cooking for for the next couple months uh into the end end of the year? Uh Ram, why don't we start with you? >> Yeah, a lot of stuff is happening. We will go probably mainet as I think when this one goes live. Uh so we are working towards that. >> Uh we want to enable two aspects to our product, right? One is the uh data contribution which is we believe will become a gig economy going forward. How video content creation and any kind of content creation was um there's a huge need for it. We want to enable that. So that's our consumer play where we want the end consumer to contribute data to open ledger and get rewarded for that.

On the other side, we would work with a lot of enterprises. We would enable community build models on us. We will also enable enterprises uh to build models and support them and use that as a way to um have people who contribute uh to open ledger get benefited because these enterprises if they use your data they're going to make revenue and you're going to be part of that as well. So one among that is our partnership with Trust Wallet and a couple of other um you know enterprises that we work with where we enable them to build specialized models. We enable them to implement AI for their solutions and we want to make sure that whoever contributed to that gets rewarded using our protocol. So that's what we're working on and we have close to about 20 projects building on us in our community. All of them will go

mainnet as we go mainet as well. Uh some of our enterprises implementations will also be live as we launch as well. So quite excited. Um pretty nervous as well. >> And on the trust wallet side, what are you guys cooking? Yeah. So I think one of the hottest thing recently is RWA related uh and we are one of the major wallet um that we have this capital market alliance with and also there are a couple of other uh RWA issuers are working with us and then we we are very excited about it because this is where bridging traditional world with the onchain world and showcasing how strong our technology can be and most importantly is that this can leverage our uh global

user accessibility to truly test some of the hypothesis. Uh are some of the sort of markets of the assets are going to be interesting to the other parts of the markets. Um so we're quite excited about that to be at the forefront of the industry and then rolling out firsthand when things are available then you're be able to expect that to see in the next two to three weeks. Um and uh it may it may work or it may not work. But you know this is part of the fun fun aspect of being one of the first and being on the front and then trying to test out um some meaningful impact our industry trying to bring to the world. So I'm I was looking forward to it.

>> Guys, thank you so much for your time. It's been an absolute pleasure. Congrats on both of your successes. Working hard to bring this space forward, the collaboration as well. stoked on it for you guys. Hope you enjoy the turnaround in the market, the sentiment, and really what what I'm most excited about from the whole kind of turnaround is just the embracing of of all these, you know, of stable coins, of tokenized assets, of of all these different parts of the crypto stack where now uh payments, capital formation, yields, and onchain trading has gotten to a point now where it's undeniably as good, if not better, than the traditional centralized counterparts. We are making large

strides forward and I think the perception of our space is equally growing in a positive direction which I think is very very bullish more more bullish than any price can can do for the space. So guys thank thanks again and uh look forward to syncing up uh in a couple months again hopefully and catching up. >> Thanks. >> Thanks Andy. Thanks B.