Emiliyan Enev, ReCheck: What will happen when blockchain intersects with AI? | ETHSofia

[Music] uh hi everyone my name is Emil NF and really nice to be here today and actually I'm quite excited because this is like the biggest blockchain event in Bulgaria as far as I know and uh so happy to see some so many people coming from abroad I'll be speaking today about what will happen when uh blockchain marries AI when they meet or intersect uh and I will start first with giving some practical examples and then I'll present one product that we are building it's called Data Pond uh that visualizes uh the the concept and uh presents how this could be implemented actually in practice uh so um first of all uh we all know that uh Ai and blockchain they to some extent they follow the similar Road like adoption like challenges like um uh potential botton X and here the the example that I can give is that blockchain had this rapid growth then some kind of a crash and then at some point it started to grow in a in a sustainable way based on some practical Solutions based on uh projects that deliver value and I believe that with AI we might follow similar pattern with enthusiasm with hype and then at some point things start to go down because U uh this is um um exhausted as a as an emotion uh as an enthusiasm and uh there is inevitable crash or some kind of a crash it happened also with internet uh so maybe with AI we will see something similar but speaking again about blockchain and AI uh what are the challenges that can be addressed by uh blockchain in the realm of AI one of the things that uh we can see is that currently AI functions like a blackbox uh so you you don't know what kind of information is used by Ai and um you are not sure what uh what data is uh being uh sourced in order to generate some results and it is a kind of a blackbox uh so here uh we believe uh and this is my personal opinion and something that really motivated me to build the solution data Pond is that blockchain can provide this very secure and transparent environment for tracking what kind of information is used by AI so imagine this data sets documents registered on chain and then AI agents consuming this information in a way that there is an utmost traceability and transparency uh so uh in that way every decision or every uh result that is generated by AI can be traced and you can see what are the sources what information is being used and of course this means that you can use this information generated by AI with a high level of trust and uh conf confidence especially uh for business decisions especially for uh more regulated markets um another thing that uh is really exciting uh this is the autonomous agents and uh here uh I believe that uh they will act in a way uh that uh will leverage this strengths of blockchain uh so uh here the the examples could be around um managing the agents based on smart contracts uh increasing the the level of security and uh even uh enabling these agents with wallets and uh uh making them pay for services or data that is being used or even they uh can uh can distribute this um this incentives uh they can act in a way that is um auditable uh with the with the traceability provided by blockchain so uh I believe that here we we can see a very interesting advancement and uh Solutions uh in the future because for me personally uh I have been developing a smart contract since 2016 one of the Pioneers in Bulgaria and we we always looking for the C use case what you can do with these smart contracts how we can apply them and uh uh what is um the most practical way uh to to promote them to the market and with the adoption of AI agents I believe that this could be one of the use cases uh that will uh prove and show the value of smart contracts okay um something that is very much into this realm something that we believe um uh with my company uh re and with the team uh that could open up huge opportunities how to create decentralized marketplaces for data consumed by AI uh and um of course um now we have this AI tools we have the AI agents but they need some verified content they need some high quality data because in the future uh AI will be very much data driven now it's about the hardware about the large language models about uh this this capacity uh but in the future many things will be uh related to whether AI have access to data is this data uh reliable and um high quality and how to automate the whole process uh so here uh blockchain can play very essential role uh because we can incentivize people to contribute with meaningful information um I'm speaking about business users I'm speaking about um any kind of um ENT users like individuals as well and then uh people can participate in the process of curating data uh making this information really verified and um they can get incentives of course this could be talk organized incentives of course um they can participate in the process of Distributing rewards uh and uh we are not so far away from uh this moment where you can say okay uh for example this is my expertise this is my important um data regarding uh what I have learned throughout the years so I would like to expose this information to AI I would like to make it available for AI consumption and each time uh AI agents use my information I want to get some rewards so this is totally doable and um I'm excited to to see some initiatives already into this direction um so this is the point that I already touched but um the overall process of um uh alignment of um Distributing this incentives will create new business models and um the the most exciting examples that I can see for example with Netherlands where my company is also active is uh how certain institutions um have this um let's call it deep web information so this is an information that is not really uh indexed by uh search engine it's not really available on Google but then this information can be also uh distributed with AI and um you can get um reward and you can get compensation about it uh so yeah this is uh this is something where uh we're going to see a lot of collaboration a lot of um new ways to uh to make uh Solutions community-driven uh based on mutual curation and verification and uh I believe that this will be one of the key word use cases for uh for the intersection point between blockchain and AI so now I'm switching to a solution that we're building uh it's called Data Pond and data Pond uh comes from um our expertise as a blockchain company uh throughout the years we have built infrastructure for securing data for tracing data signing data sharing information uh and now we are connecting this with AI uh so I just wanted to give you um some flavor about the solution that is being shaped right now and um I wanted also to to give some practical uh insight about uh how we are building this okay uh so these are some of the challenges that we identified um before starting the solution and so we already touched some of them uh So currently uh the whole process is very opaque it's it's it's not transparent and you don't know what kind of information is being used by AI uh I speak not only for the uh training of the model or for the fine-tuning but also when uh when you ask questions and when you see the results you want to make sure that there is a there are some references to to evidence something that is real something that is um reliable information and this is um this is a missing piece and also on the other hand there are a lot of regulations about transparent AI about of uh about um uh human Centric AI about U making this in a very compliant way and here blockchain can can play essential part because again uh with within this Dutch institutions with with whom we are collaborating one of the most important question is okay we have this sensitive information like business private data or um data of Institutions and we want to use AI as a distribution channel uh in order to efficiently uh share and monetize information but how to protected how to make sure that uh uh there are no breaches there are no leaks and uh everything is used according to the rules and uh what is the audit trail that that you can um leverage uh in order to protect the whole process and uh here again Solutions like data Pond and others uh can play an essential role I'm not sure if you can uh see the text or or if is too small um but um here I'm presenting uh the the main features uh behind uh data P so what is data P you can look at it as a data collection and uh data collection means that we gather information regarding a certain topic uh it could be information about uh for example a specific website it could be information about certain institution it could be uh data about a product and then uh this becomes uh a pwn a place digital space uh where you start to organize this data and to create it and uh then the next step is to use this pond uh as an end point for the a agents so they connect to to this um data collection and they start to uh to consume data they start to use it for for their inference for uh generating results and uh for uh for their functioning but each time this happens there is a evidence recorded on chain that exactly this piece of information was consumed and uh once you have this of course um you can introduce some incentivization mechanism uh there is there is a new way to to monetize your information so uh what is the vision uh behind this solution um actually we uh aim at um creating a Marketplace and Marketplace for AI data um this is like the the big picture that we have in mind but uh at this moment uh we're making a small steps and uh we are applying our technology mostly for um in-house for internal solution for businesses because uh if you want to uh to use it for your organization again um you can apply the same principles for protecting securing data tracing information uh audit Trail uh even uh uh some kind of incentivization uh mechanism behind this uh but um the the big vision is to create um uh a data feeding Hub and to to enable businesses and users uh to distribute information uh with AI in a very efficient way so this is just um a screenshot from the latest um version of the product and here on the right you can see all these uh interactions with files all this evidence that information um was consumed so this is based on blockchain transactions everything is uh verifiable uh everything is uh really transparent and now we are working on another feature for introducing web three points and tokens so once your information is being used then you actually get some rewards um we believe that now is the right moment to to speak about the intersection between blockchain and AI but also now is the right moment to shape such Solutions because we see this explosion of uh AI agents uh we see um how this can can create totally new uh Solutions business models and um many of them will need access to such high quality data uh and the emerging regulations are also pointing into this direction because um uh again I can speak about Western markets uh all these aspects related security privacy of in of data uh they're they're really critical for such Solutions and um here uh blockchain uh can also play a part I I'll not go into the direction of private or public or hybrid networks it's is a totally new topic I just want to make this uh clarification that for the beginning we are mostly leveraging public data because you need to be very careful what kind of information um you are registering on chain uh uploading to the blockchain so this is um a whole new discussion that that we can have okay okay uh I'm going to uh skip this s uh we're speaking about a market that is in the process of development this is the um the billions um sides the billions number s especially for uh for pitching events but here uh just pay attention to the last Point not sure if you can see it actually the Deep Web the uh the information that is not indexed by search engine is currently perhaps 95 5% of the internet so just just imagine what kind of um opportunity is this because this is information that can be uh distributed with AI consumed by AI if there is an a mechanism uh for uh for organizing this in an efficient way in a secure way but also with uh Fair rewards so some of the USPS that we have are related to the blockchain infrastructure that we have built throughout the years but um uh also um we believe that such products should be community-driven there should be some positive sum game for example you do something um that creates value for the community for the whole ecosystem and then you get a reward uh and uh this is something that is uh being organized at this moment and actually I'm open open for uh any uh suggestions and ideas so if you want to to discuss this you can always come contact me my email is in the last right okay uh I'm going to uh to skip this one because the food is already in the room as far as I can see so uh these are my contacts uh thank you and I open for your questions uh I'll repeat the question uh it's about giving wallet to the agent and how to do this in a secure and efficient way uh avoiding some scenarios for Ste killing uh uh hacking that was the question right uh okay so I don't have a specific answer to be honest but I believe that if this is organized uh with uh smart contracts and with u the security provided by blockchain and within very strictly defined rules uh you can limit the risk uh currently we experimenting uh with uh with my company re uh in this regard and um we have developing smart contracts uh about consuming data and paying for data uh I can share experience if this could be helpful uh but uh to be really uh transparent uh it is a very experimental stage at the moment there is a research and development and I believe that most of the security comes from uh uh strictly programming the rules in the smart contracts and uh also very clearly uh defining uh sources of data who can operate with uh with the agent it's is about role based taxes control Etc uh it could be also organized with blockchain accounts and identities so this is the short answer but you're touching a very important uh point which will be uh further export and developed in the coming years the question is whether we're looking for uh investors for data Pond are we raising um at this moment um uh we are preparing for a potential uh round and um we would would like to do this in um let's say um hybrid way we're still not sure because we don't want to go with the token at the first stage maybe we'll use web three points as a way to validate as a way to give incentives to to have a very specific use case so at some point you can convert these points into tokens or there could be another mechanism but yes the short answer is that uh we'll be raising in maybe two or three months the question is what are the interesting um industry Industries and um markets for such solution um so we have Pilots with media companies uh we have Pilots with um um organizations that are typically data providers they're collecting uh some uh specific data sets in a certain industry and they are quite curious how they can monetize this information in a in a secure way uh logistics companies some of them because I personally have more relations with with such companies uh for me I'm very interested in the domain of e-commerce this is one of the use cases that uh that is quite quite um potential with with a big with a big potential because if you use data point for example for a product catalog you can achieve very interesting things again very important question uh the question is uh how uh to to upload the data on chain uh register it on the blockchain uh first and the second question uh was about um how uh to ensure this um security when uh data providers expose information to AI uh regarding the first question currently we're using a blockchain protocol called ICP internet computer protocol they have a stand uh on the right when you get out of this room uh they uh have this um very interesting feature of being able to upload files really on the blockchain like have 100% storage on chain and with this solution that I showed you with data Pond we actually upload the files on the blockchain and uh this is like a separate layer with with the storage of the files so this is uh currently the way we organize it but it's not very good idea if you're dealing with some sensitive business data because uh when you upload it on the blockchain then it becomes public and um other users can access it potentially so uh this goes uh to to your question regarding the uh the security the the encryption and uh and data privacy uh my company specialized in um um exactly in this field how historically we have done this uh in a way that you encrypt information with your keys and uh then you allow other users to access the files in a very secure and a very private private way so for example I allow uh only you with your keys to access my files so here in the context of the data Pond um one of the ways to do this is to provide a very precise uh metadata and description what is in the file what is in the data Pond and then based to this um end to encryption to allow AI tools so users to access data but only within uh certain uh restrictions within um perhaps a smart contract that is executed so the the security and privacy are protected there is a lot of Technology also research and development behind this and if you wish we can continue the topic thank you very much [Music]