Knowledge Assets and AI-based Search in Web3 - Brana Rakic | OriginTrail

all right so uh thanks for the intro actually Tiny correction origin Trail is the decentralized knowledge graph so um yeah uh just kind of a sentence two or more about uh who we are so we as Trace labs are where like your uh your parity to you know the polka dot or whatever so Trace Labs is the the core developer of origin Trail decentralized large graph I'll explain it a little bit today uh we we basically started um the company almost 10 years ago and uh back then we started as a proprietary centralized company that was building essentially knowledge sharing Solutions with what we knew at the time and we didn't really know much about decentralization when we started um and over time we actually went fully decentralized fully open source and standardized in many ways so today I'll share a little bit about that and what is the the actual vision for origin Trail and how you can use what we're building so I'll start with the problem as I guess usually is the best way to start and um essentially I like to point the one fact out that we're in the middle of a new knowledge Revolution essentially the first knowledge Revolution was uh in a way the revolution that kick-started with the the invention of the printing press so essentially we we all of a sudden got this amazing tool to replicate knowledge uh and essentially we solved the problem of knowledge scarcity um with that I'm sure you know many really cool things came about in the world feel free to come in man it's open uh all right so um essentially you know all great things Renaissance and everything I think everybody kind of knows about that um and then um after this revolution came the the next one so we had the whole internet Revolution knowledge sharing became kind of a very common thing and we solved the problem on fragmentation essentially knowledge fragmentation was what we had before we had the connectivity as the key way to solve it and through the the Network's ethernet internet and all the other networks that were kind of now building all together here we're tackling this problem effectively for a long time and with that problem we actually have abundance of information now uh and in the the problem with abundance usually comes with also a problem of you can call it problem of spam or you know generally problem of when there's a lot of something there's usually some of that that is crap or actually wrong so the next knowledge Revolution I'd like to make the case today is happening right now it's actually happening with AI and we already basically have a huge problem of misinformation happening all around I'm sure you guys know about like all these things like fake news and deep fakes and these are just some manifestations but generally it's very hard to you know sort of decouple information from misinformation and it's going to become harder because now we have ai that can go ahead and hallucinate all kinds of things people are going to be using it like crazy so we believe that the solution to this particular problem and what is going to come about with this new knowledge Revolution is really decentralized so I'll try to explain the vision of origin trail behind that in the next slides just a little bit more about the problem I'm sure you guys heard about this guy The Godfather of AI Jeffrey Hinton recently left Google he's the guy behind the llms basically um one of the the researchers that that's why he was named the Godfather of AI in the first place and he basically left Google so he can talk about this problem freely freely and he's very afraid of um of well llms and and the AI that comes out of it um and I'm sure you're hearing all kinds of news just yesterday we heard this scary thing about um I don't know if you heard about it but some Army AI uh thing did a simulation where it actually uh killed the the guy who was running the guy the the imaginary guy in its world was running the simulation or you know something along that sort um also funny thing that was a little bit more um real happened just recently the couple of days ago I took a screenshot um some some lawyer actually used chat GPT to actually well basically come up with uh with the documentation needed for for the case and uh they used chat GPT and they got a bunch of quotes to some legal documents that never existed so you know is this uh benevolent mistake or intentional probably like the guy didn't want to get in trouble so it was probably you know as we all are in a rush he was trying to you know come up with something and um ended up creating misinformation and probably not even you know being a bad guy behind it in a sense not best practice so definitely not say you should do this but more and more of this will happen and this is just like a kind of a minimal example it's already happening right now so um so with origin Trail we're trying to build fundamental trustless open technology web tree based to to tackle such problems and I'll try to explain in the coming slides how that works there's a lot of stuff to cover so maybe since we actually have quite some time 30 to 40ish minutes I'm totally open if you guys want to just like randomly jump in with a question somewhere just raise your hand I guess we can do that right maybe you can help bring uh awesome thanks man because maybe some of the slides are not so clear um or just the way I talk so feel free to jump in and we can make it a conversation as well um so all right the basically what we're trying to tackle is the problem of missing misinformation and one of our mission State the mission statement of origin Trail is really to organize trusted knowledge or what we call Knowledge assets and to make them discoverable verifiable and they are ready and that means um what I'll show you in a bit so just a little bit about us I won't go too deep like I said we started back in the day 2013 2018 launched the first version of origin 12 decentralized Knowledge Graph on ethereum it is actually a decentralized network that sits on top of blockchain as a decentralized network so it's not a blockchain based protocol just sitting on the blockchain rather it's a it's a you can think of it kind of like ipfs but it's with Knowledge Graph Powers so like ipfs from superpowers in a way and uh this network sits like I said outside the blockchain and uh has been launched 2018 has its own token called Tracer track you might have ran into it and since then actually went ahead and expanded to because it's a multi-chain decentralized Knowledge Graph it went to polygon knows this um and uh last year also in polkadot so we launched the custom original Power chain with some custom developments there and we're we're actually running that for a year now meanwhile we've been very successful with adopting the technology in the Enterprise space so you can see here some of the companies who are using who are sharing actually security audits via using origin Trail in a privacy preserving fashion using the decentralized knowledge graph and you can see a quote from Dan Patel The Group innovation director of BSI British institution who is one of our partners um longer Partners we also work with the Swiss Railway company the Swiss Railway company has for different solutions being built on top of origin Trail they track trains train parts events actually particularly security related events for example events such as think of train tracks they can split actually there's like they get welded but if it splits literally the train goes off tracks and well oh Breaks Loose right so they're trying to use all kinds of data which they collect from different welding Partners using the decentralized network to basically Infuse trust into it so then they can actually use it for prediction so they're trying they're running different kinds of queries to understand what and why do these uh weldings break so they can understand aha this will happen I don't know a year from now in that part of the country things like that they also track different things because it is good for efficiency so those are just some of the partners we also have some Partners traditionally from the supply chain industry basically all of them using origin 12 in some of their Solutions so maybe you've seen this before if I won't go too deep into it but if you're interested in knowing how it works and there's a lot of the the information on our website and you can also approach us later there's a bunch of us from the team here uh sitting around so you can also ask us how it actually works and what what it uh what do these companies do um so all right um how do we tackle misinformation so I'm going to present a couple of tools that will help us basically approach the solution to this problem but let's first establish one big sort of point when it comes to uh trust in data or information uh trust really has two enemies one is bad character actually the more appropriate photo I think here instead of these generic people would be SBF but like I guess it's okay like this too so we have bad character and we have bad data and essentially bad character like SBF we cannot really eliminate this this is always going to be there and but with bad bad data we can do something and this is a challenge that we can we can tackle at least to some degree so bless you we're not going to be causing we're not going to be tackling bad character per se um so how does origin Trail work origin 12 basically synergizes blockchains and knowledge graphs um and you can think of it as in the simplest possible terms some three layer structure like this so you have the layer one with blockchains Layer Two is this decentralized knowledge we have this network I told you about which is its own decentralized network and uh on top of it we can Envision all kinds of knowledge applications and the knowledge coming from the knowledge graph essentially is um it's kind of a construction coming from the semantic web world if you ever heard about the original web 3.0 coined by Tim berners-lee the guy who invented World Wide Web basically they talk about the web of data or the semantic web of data basically um the kind of the difference between the original web being that they had like relatively unstructured documents on on the web while the idea was to go and essentially instead of sharing documents rather share data so that you know that vision of the semantic web gets achieved many reasons why I think it haven't it hasn't really been achieved so far big reason is webtrade didn't exist if you actually go and look into Tim berner's Lee architecture for this semantic web you'll see all kinds of things Uris you know query languages and you stack it up and on top of that there's like a small cherry on top like thing which says trust and like you know the trust was kind of the last thought that they had but that makes sense because at the time when they were coining this which was 20 plus years ago it wasn't really something that that was as as interesting as today and we didn't have the things like we we have today with web3 so um in many ways origin Trail somehow marries this old web 3.0 version of semantic web team berners-lee and the new web3 and essentially it's uh like I said it's a multi-chain decentralized knowledge graph or dkg for short the key primitive are these knowledge assets which I'll explain in a bit more essentially knowledge nfts and they are they have a very cool property so they have a verifiable data origin an information Trail see what I did there and it was like I said launch 2018 it's a permissionless system so there's no like admin or some sort blockchains are really used as a trust layer and as we believe blockchain is a really great trust metrics and the dkg is really this semantic Network or really kind of a data layer if you make think of it as as so the reason knowledge graphs really being great for data they're like your ultimate in a way database depending on what you're looking for but if you're looking for something that can map out the real world different types of models different types of things you'll find knowledge wraps there and where is that being used today in web 2 it's all over the place we just don't see it it's under the hood so Google invented it basically coined the term knowledge graphs they said actually the explanation of knowledge graphs is things not strings so instead of moving from like data in string form we move to things that have a specified type A Certain structure or ontology really applied to a certain knowledge domain so for example you can encode inside of the data you can encode let's say a bunch of your business logic in this these ontologies I won't go too deep into that but basically think of it as a kind of a very clever system for mapping and storing knowledge and then having a specific standardized w3c standardized query language for that which is actually called Sparkle in this case um while the blockchain in origin Trail is really used for not to store a bunch of data so as obviously that that will be very costly but also because knowledge graphs are great for that there's no reason to put blockchain like in the same function as knowledge graphs rather it's used for things like decentralized identity fingerprinting of data actually Merkle roots of the graphs and ownership so essentially it kind of looks like this we have this knowledge asset which contains knowledge which is this set of information that is structured in this standardized form either rdf but also Knowledge Graph embeddings and similar things feel free to jump in sir it has this uniform asset locator or ual which is this origin trail version of urls which basically combines also two things so the pink part being on the blockchain we have this knowledge asset nft which gets minted when you publish this knowledge and the state proofs so essentially if you create a knowledge asset you you basically put the data on the Leaky G put create an nft and a proof that corresponds with that data on the blockchain that you pick out of the blockchains that are available and that also creates kind of a combined identifier this ual which basically is an extension like I said the URL it combines a did of the record on the blockchain so of this nft did meaning the decentralized identifier also by w3c standard and the knowledge graph identifier so basically their self dereferenceable identifiers you don't need like a trusted DNS in between you can just you know look up the right nft on the blockchain yourself and then you can find the right record in in the knowledge graph or you can use the client that does both things for you um if you're the owner of the nft that means you can also manage that knowledge so you can update it you can you know do all kinds of things and whenever you do that another state proof is created basically so it's it's a miracle proof so you can also have all kinds of cool things like if it's a lot of knowledge you can show that certain component or parts is also part of that because you can construct a Merkel proof around it but long story short you can kind of think of this as like your git Branch like you have some data then because you have rights to update that Branch you create a new commit that's a new state essentially and you can basically go throughout States so it's not um it's not something that we haven't seen it's just kind of a thing that lives on a couple of networks and then on the basically all of this is indexed in the dkg what does that mean so that means that you can basically run queries um against this data this knowledge in the in the in the basic uh in in the knowledge assets all of them together so if you let's say you were to try and create one and you take the JavaScript client dkg.js you could instantiate the client and you can just like you know do your regular crud stuff so this will be how you would create it so you take the knowledge in Json LD form some options you publish it on the network then you get a ual so you can also resolve it by get just like HTTP get you can also update the knowledge asset and then because of the knowledge asset 10ft you can of course transfer it so we have this classic nft interface under the hood and obviously you can get owners you can go get all kinds of other things like proofs and so on these are just some examples so we like to think of it as a kind of a Crut rather than a crud create read update transfer so that's basically what knowledge assets are about and the cool part about it is this this knowledge part up there if I can click on this thing yeah so in there is the the secret sauce that basically gets this semantic properties so having the ability to query all of the knowledge asses the entire decentralized knowledge graph of knowledge assets comes from there and um also the obviously the blockchain part having the regulating the integrity and also the ownership of these knowledge assets so what can we put in this knowledge we'll talk a little bit more about that over the next slides but essentially the way you can sort of think of it is kind of having one decentralized launch graph which has a bunch of knowledge assets rooted in multiple blockchains around um and um obviously taking benefits of different blockchain ecosystems be that security or speed or you know just some custom logic that's implemented there or just a well you know Community or benefits that you might want to get so we're building this in uh as we like to say blockchain agnostic matter so for example if you were to think about something like an llm let's say let's try to have an example what benefits can an llm have from or just any AI solution from more Justice well we basically get the verifiable information origin because there's a transaction you can follow as well as ownership we see the update Trail of the information data all the data has Integrity because you can always check if it hasn't been changed or modified in certain way and the most cool part actually here is this semantic representation this knowledge graph properties that that you get so for example we've just done a POC I think I have a slide on that as well a POC of something that looks very much like 10gpt but it doesn't generate the answers it actually extracts answers from knowledge assets so I'll show that a little bit actually I do have a slide which is Chad dkg actually so we're just launching a basically a new initiative and there's um this might be interesting for Builders by the way because we have uh one million track token Bounty uh or rather Grand program um where you can actually apply with tools or solutions that you would like to build that would fall under the auspices of this Chattahoochee framework so I'll try to explain it in a bit and basically how you can how you can use it but this should be a video that basically illustrates kind of the concept and I hope it plays let's see and and maybe one more time no okay maybe it doesn't want to play so I can explain it essentially think of uh think of an interface where you can ask a question rather than shoot a database query you actually ask a natural language question you have the ability to somehow understand the query from that natural language question the software can for example and then it's able to extract the answer from the whole decentralized Knowledge Graph and showing you where these answers are coming from for example if the question is are is my I don't know sneaker built from like I don't know is it built in a sustainable way you will be able to get sustainability reports coming from whoever issued those knowledge assets somebody ideally with authority so you can see okay who said something who created some report and and that answer actually has a verifiable origin verifiable data trail and uh obviously ownership so unfortunately I cannot show you the video but I can show you or maybe on on the laptop if you like to see um so what is this framework about basically the idea is to create an open framework which is relatively broad which focuses on search really utilization and access of trusted knowledge and the idea is to build components that somehow either integrate or are already AI based or integrate with AI Solutions so that you can basically bring the trusted knowledge into the AI space these are the objectives so we want to create tools that basically are based on both knowledge and AI established best practices for all of these things and really generate and that's an important Point as well a trusted open knowledge base so it's also about sourcing trusted Knowledge from different uh verifiable sources obviously now what are verifiable sources it's a good question essentially the idea is to utilize decentralized identities so you can actually find that a certain particular entity who has certain knowledge assets was actually behind some information if you can think of conceptually like about this framework the idea is that ideally you have something which is already relatively structured as data sources and that's like open data personal information something from like enterprise software like the ones we work with for example Swiss Railway company sap things like that or generally ipfs or any type of data source this could also be blockchain but generally the focus here not being to create another indexer where you're looking to let's say query the latest uh price on a certain token pair in your new swap or something like that rather it's more about potentially doing some sort of semantic query on a blockchain like you can for example ask what are the three addresses that have these uh perform these activities that could be something for which the blockchain could be the data source for that you need to turn data into knowledge somehow so this is basically this structuring element anegontologies things like that so it's not it's not a trivial task but one of the cool things that can help there is also AI if you do it well and if you you know cover the the basis you can use these tools to actually create knowledge and then we get to this middle part this AI ready knowledge the essentially having two key components here have some knowledge bases and one thing I I didn't mention previously is that not all of the knowledge has to be public the knowledge can actually remain in knowledge as we call them knowledge volts within the origin 12 system but also existing knowledge graphs or just knowledge bases that that are present in the world of whatever Enterprise or just general institutions or whoever has a bunch of data for example Google Knowledge Graph and on the other hand we have some indexing instruction infrastructure how do you somehow sift through this knowledge like you have you need some ways of um having the efficient ability to build search on top which is this next layer this age knowledge interface components and on top of that finally some AI Solutions so the point here being is that we're looking to build things on the spectrum of all of these gray elements here and if you are looking to experiment with such tools and if you're interested in the trusted knowledge assets then this could be something for you and you could actually participate in building the chat dkg open framework um one of the kind of key ideas behind it is combining this symbolic and neural approach in in the world of AI actually knowledge graphs come from this older branch of symbolic AI which is in a way you can look at an autograph looking at this kind of graph databases thing on the left and then one of the cool things is when you combine that with this new new world of neural llm data processing essentially you can create something called Knowledge Graph embeddings and that is basically different kinds of vectors from the data and Knowledge Graph these vectors represent or rather encode semantic information that comes from the knowledge on the side and you can use for all kinds of things semantic search is one so you're able to actually search by Vector similarity rather than by you know keyword search like the it used to be kind of the the name of the game until recently but also you can do all kinds of other cool things for example knowledge prediction you can take elements in the launch web that are missing and you can figure out based on certain models and embedding strategies what data is probable data or probable outcome of certain knowledge on the other hand from classic knowledge graphs ontologies also have reasoning capabilities and this reasoning capabilities are not um generative or probabilistic in nature rather they're actually based on um essentially Computing logic so you can basically apply the principles of basic logic logic and derive certain knowledge out of your existing knowledge it's a bit of a longer topic but it's essentially it's like like I said this older branch of AI which is really good because it complements the new let's say autocomplete llm world with something which is very strict and actually is is mathematical so it's not probabilistic rather deterministic so in that sense combining those two the idea is that we can do all kinds of interesting things what do we want to do particularly for example we want to do semantic search this is a a brief explanation of a proof of concept we just delivered for for one of our partners um which basically involved creating knowledge assets out of their documents and on the right you can see basically the flow so the organization documents or the original data basically went through a certain graph Builder component which applied in a certain ontology and actually added a certain set of Knowledge Graph embeddings using Bert so to produce these knowledge assets we got basically a certain pipeline going and then from that moment on when they were there you can actually run a natural language question on it so actually we ran it through both the vector similarity search approach and the term search approach the classic you can consider this your elasticsearch if you've done something with that and then basically out of the search results coming like basically filtering these knowledge assets we had the graph reasons are applied in the end which kind of decides what is the best or the best candidate answers but the the big Point here that you get is apart from the algorithm which is not part of origin 12 per se what is basically the core primitive here are knowledge assets and on the knowledge assets apart from getting the answer content and the related knowledge assets because they're all part of a certain graph you also get a bunch of interesting information Source document for example uh who's the owner of the knowledge asset who was the issuer because it might not be the same as the owner as like we've seen the ownership can be transferred or sold or whatever we get these blockchain immutability proofs and of course you of because of search you need some sort of confidence score or a relevant score at the end um and okay ultimately you can also use user feedback and basically get some more um got some more you know information going into it and basically somehow iterate and either add this information to knowledge assets or somehow enhance your algorithm so this is an example this is not what we're trying to build not the only thing but for example you could build something like this and if it's according to the proposition of chat dkg framework you can apply for a grant so just to give you a bit of a context um so what are the high priority areas we're looking for we're looking for these four that we're looking for semantic search over knowledge assets so what you've seen sort of like what you've seen previously there was just one POC we're not claiming we know how to do this best we did this for the first time in our life so maybe you guys know better um and if you do please get in touch the other part is really natural language querying over dkg so for example here you can think of something like a GPT based Sparkle query generator so the idea is like I said Sparkle this is like this knowledge graph language which pretty much is like SQL but for another draft so it looks very similar It's relatively similar to the logic of graphql as well though it's it's more look-alike to SQL and let's say instead of doing a search over some knowledge assets and then getting drilling down into the right one you can even immediately formulate a query so if for example the query can be expressive enough so you find the results directly inside that's something that could be could be built in is one of the the high priority areas of the strategy framework um generally we're also looking for AI based knowledge publishing tools so something that could help because knowledge publishing is not such an easy task it always has this knowledge structuring component and um you have to somehow take care of identity in this particular case each Knowledge Graph entity has essentially by semantic web a URI as an identifier so it's it's not just like you know bam I published some file it's a little bit more work but that work is pays off later in these capabilities of running all of these queries and and the cool stuff that you can do so basically tools that help with that make the ux better make it easier make it easier in a certain language those things are interesting and generally AI system Integrations for example if you have an idea something doesn't fall into these categories but you have some cool idea that you can leverage knowledge assets with an AI tool that's also applicable so even outside of that if like we didn't cover it with these four if you have some cool idea just give us a ring what criteria it needs to basically withhold is then it has to be a public open source project so it cannot be like something proprietary You're Building no share with the community has to apply to knowledge asset it has to solve a meaningful problem and obviously we have to do this in basically active Community Development so what we are looking for actually is also to learn from each other so we we have obviously an original Discord group and trying EKG dedicated channels so you can actually come and and uh you know share this and learn together with the people who are building this also um there's also going to be like I said a grant 1 million Trace tokens uh which basically you can apply for if you if your solution falls under the auspices of that previous slide and uh yeah the idea is to basically in a way again build components that have help tackle misinformation and it's been it's going to be a journey we cannot build this in a day so ideally we all build this somehow together um there's also going to be a steering committee so this is not it's not going to be me who picks maybe I'm going to be part of the steering committee but essentially the idea is that we're looking for people who are also knowledgeable in these areas web3 knowledge graphs semantic search AI things like that who would be ideally supporting this effort so it can be you guys were also looking for folks from academics we literally just started this uh last week so it's Super Fresh um and the steering committee will be responsible for these things as you can see so basically guiding and development supporting Builders and facilitating these discussions But ultimately also picking who gets the grunts so we're um we're looking for applications if you're interested uh they're going to open up soon but before that you can also approach us and chat about it um so yeah that's that's the first step and then later we we can we build and we we spend this cool budget together um so you can join the channel by scanning the clear code or just coming towards Intel Discord there's a community that's already there and interested in building all kinds of things um be happy to see you guys there and uh I didn't manage to share too many things about origin Trail in depth I didn't want to but if you're looking to know how the technology Works what you can do with it what type of tools we have it's not just me it's a bunch of people here around you'll see them in origin 12 shirts or maybe this dkg shirt feel free to approach us we're super happy to hear your feedback or questions or whatever also questions can be said now I suppose because it's an open discussion time thank you for the attention folks [Applause] hey thanks for presentation really cool I have two questions um uh what's the payment model for your clients so uh how would I pay and the second one uh you mentioned native tokens uh so what what will be the use cases uh for for it okay so question number one we as Trace Labs as a company we we have these clients we basically like the Swiss Railway company for example they would come in and they would ask for a certain solution to be built um and we either find some partner to help them build it or we help them integrate origin Trail directly but it the prerequisite is they have to build it on with origin Trail otherwise we are not like it doesn't make sense for us to do it the way they pay for it obviously if we build something custom we we charge it for charge them for it um but however what we also help them do is on board to the the technology so one of the big hurdles of Enterprises so far has been the day actually at least since 2018 they've always had problems uh putting the the crypto tokens on there balance sheets so it's really a problem of how do they go and buy Trace tokens and so they can publish knowledge too so what we actually built is something called Network operating system I think I have a slide on that um somewhere just a second which is basically a tool uh these are some of the solutions I don't have a slide I'm sorry um it's basically kind of a web to to web3 onboarding platform for Enterprises they pay a certain amount of credits in Euros which they can pay and then we basically provision tokens for them we go and buy it on an open market on on an exchange and then with these tokens they publish knowledge assets as they create them so the model is we built something maybe custom for them but essentially we are selling these subscriptions to to this network operating system as we call it and we actually see them as Network operators for example Swiss Railway company has over 20 companies who are publishing knowledge assets but through their account you know so it's it was the easiest way to onboard them same goals for BSI so so that's that's I guess question number one um question number two was about Native tokens yeah basically uh maybe one thing I didn't mention um there's two tokens actually always in place so there's the blockchain token uh and then there's a trace token and uh essentially double the problem right for balance sheets so we help them buy both but essentially the blockchain token is always you know basically encapsulating this utility of the blockchain right so using the blockchain while the trace token is the utility of the dkg so publishing knowledge assets essentially you have to use a certain amount of Trace tokens to satisfy the the fee Market of these uh nodes who are running the origin 12 dkg by the way right now we're in version six we just launched that in December the network has around 115 nodes if I'm not mistaken run by the original Community I see Nikita nodding the head because our infrastructure responsible so it's correct number but the previous version version five was actually having two thousand nodes being ran on the network around the world so yeah we see it as a good achievement but um it's also a challenging one because it's quite hard to manage such a network it takes a lot of a lot of energy but it's it's worth it for for the principles uh and the benefits from from the technology so I guess I don't know I guess did I answer your questions we have some other ones as well awesome uh hello so hello I saw here in your website that you have solution for identity Solutions so are you do you have a playing to integrate x domains inside your application or project and also logging with ethereum absolutely so the short answer is yes uh the longer answer is we're looking for ens like standardized let's say solutions that can and anybody who's by the way in the indentity space I would really love to talk to you because we want to essentially enable all of those emerging identity ideas to be used and the way we thought about it was the best way we think we can do that is to follow standards such as decentralized identifiers verifiable credentials things like that from w3c but actually that is the missing piece that we're looking to integrate with so absolutely all for ens and going deeper into all of the components that DNS can bring but let's chat more about it let me ask a question can you a bit describe how you actually struggle with incorrect dating your network say cyber attacks or like intentionally or non-intentional like you mentioned that you cannot struggle uh Byzantine type of behavior of users right so it's it's not planned to do that or can you expose it a bit yeah um Okay so there's a couple of different ways you can build let's say a decentralized knowledge graph has anyone ever well you guys all used Wikipedia for example one of the big sort of hurdles about Wikipedia is a lot of people say you should not trust it because you know it's not great there's an academic source and stuff like that um but the Wikipedia Wikipedia has a bunch of people a community behind who are actually taking care that this knowledge is somehow true this information that's there actually there's Wiki data which is the counterpart to Wikipedia it's pretty much it's the same data but in Knowledge Graph form and you can query it which is also really cool the problem there is you have some Authority some group of people who decide okay what this should be inside or not we go from the perspective of trusted knowledge infrastructure so we provide tools to verify that certain knowledge asset has a certain issuer audit tray like a set of information identity connections but we don't necessarily we don't want to build into the infrastructure some let's say truth protocol rather the idea is that anybody can publish any knowledge inside so we don't say this is okay this is not but that actually by leveraging this information who the the knowledge issuer is who the source is whoever is on the querying side of things on that top player knowledge asset application layer actually gets to filter out based with these Primitives some misinformation out of it it's not an easy task but we we cannot essentially encode this directly into the protocol now some cool ideas for example there's ideas in knowledge graphs where you have a certain community like a Dao that goes ahead and and votes in a way uh giving certain reputation to certain knowledge assets or people um and that makes sense but again it also brings us to the question who are these people who is this Dao so the the approach we took was we we built the actual infrastructure under the hood and then whoever this entity is be it the Dao a company a person um that's querying this knowledge actually gets to decide so in a way the similar Principle as you do today for example I don't know I don't know where you're from but I can say for example that we all know that I don't know Fox News and CNN are very different information sources and depending on your preference and who am I to say that you should not listen to Fox News but uh like you can pick like you can say I don't like what these guys are saying or the way they are saying it or maybe I'm gonna somehow combine both so it's a very very broad question and absolutely on point but it's not something that the technology aims to solve directly however supports the ability of somebody to understand if something is misinformation or not because of these key Primitives knowing the origin knowing the audit Trail of the information and knowing the ownership of of knowledge assets or nfts for knowledge as I mentioned thank you for all the questions however we are going to have to proceed with the next lecture thanks everybody let's let's give a round of applause thank you