Building the Future with HeLa Chain: Where AI meets Web 3

Transcript

hi everybody and Andy think from Hil Labs um so I'm I'm a founder of Hil labs and then I'm here to share with you the vision of Hil labs and um and also uh talk about where web tree meets uh AI so uh I I I used to work for the uh military and uh I was developing AI models uh over there and then um the AI that we we are focusing on is actually expert AI that means uh you need a domain expert to be trained the AI models so that the AI can behave like the experts so um I subsequently did a PhD on this subject and then um uh and then I I did further research on

this and publish a few more papers then during my research right I actually uh can see that eventually AI will be dominated by a few centralized party and that will actually lead to a path of towards a singularity which is um not idea so the the future that uh in haa we imagin will be to have like uh personalized AI which means that a set of AI Bots that is uh trained by the expert or the individual and also uh owned by them um so I I when I was running a trading firm I I was working also with a group of uh very successful Traders uh

however we could see that they were limited by the H there are human limitations which means that they cannot trade 24 hours and they can there's only so many uh markets that they can monitor at any point in time so I actually put forward the idea of uh getting the traders to train their AI bots so so that the the bot can behave like them and and actually help them to trade more and also do better in the markets so the the the thing right um the the thing about this uh experts right these Traders is that very often when the when you ask an expert or you ask a Trader to to be training and uh AI bought the main worry is the lack of trust because the

trader May worry that uh what if the developer uh you know cron the models resell the models or even use the model to predict uh like uh something to trade against them so uh as I when I speak to my my friends who are in other Industries uh who are similarly getting domain experts to train their their AI it's they face they face the same problems um and so there's actually a big gap between um you know what the AI practitioner wants to build and what the domain uh experts expectation are because uh as domain experts they they may worry that uh they

could get replaced if the uh if the model is uh or the AI is so so good that eventually uh they could get replaced or some parts of their job function could be get replaced so in uh in 2018 I actually I wrote a paper on like um how blockchain can actually solve this problem whereby BR uh breaching the gap between the AI practitioner and the domain experts so basically it's uh it's proposing to use blockchain as an infrastructure so to enable secure ownership of the AI models um so that the the the experts or the Creator who who train the model can

actually be the one holding the keys or the the access to the the models that they developed um so that was 2018 and then in around two years ago we we had the opportunity to uh work with AAR Singapore um to to work on this project uh so maybe for our foreign friends uh uh I can share that ASA Singapore is actually one of the top uh research agency they have like thousands of scientist and um most of them focus on AI research um so so with with with AAR we basically we are working to build uh a broad chain infrastructure that allows

you know coll collaboration between AI developers uh domain experts and data contributors so um yeah that was two years ago and uh well now I'm here so um we we have already developed the the infrastructure and then it's um it's basic it's a evm based uh event compatible modular layer one uh and then we we we have built different uh layers on on the Chain so that we factor in for compute layer storage layer uh as well as the encryption layer for our AI models um where the domain expert can train the modelers and then uh hold the keys to those model and they are the

only one that can have the access to that model so this whole uh infrastructure has already uh been realized and we and uh it has gone to mainnet in uh around 3 months ago uh currently we are at a stage whereby we are onboarding like uh um different AI use cases so we've been talking to uh different partners and uh you know anyone uh interested to put their AI on chain and actually uh enjoy this um infrastructure that we developed some of the you use case that we have done uh on our chain right that includes like um so some of the use case on our chain uh includes uh a digital

twin project uh done by autobus so autobus they map uh digital twin of Aer rook and then uh they use various AI techniques like um computer vision and uh other techniques to make a realistic uh and accurate digital Twi uh so we are working with them to put to put all you know their AI models on chain another use case that we have done is Al is uh with a company called Upsala security uh opsa security they are Consultants to uh interpo and um Korean police as well as um uh as as well as samung so uh they they develop an AI investigator on

chain AI investigator that allows you to use uh the AI to investigate like any hex or fraud that happened on chain um you know via the AI so that actually brings down the reaction time when a hat happen from a few days or a couple of weeks to merely a few minutes so all to build such a model the the the expert right he needs to actually contribute his years of experience and data points into the model and then uh to ensure that he doesn't lose the ownership over his models and whatever he built uh we uh we help him to depry that onto our broad chain infrastructure we also work on other uh

we also working on other things such as uh onchain data data management with aaps uh on chain uh with aaps and then the data management is actually focused on air traffic uh data management so LS is actually a joint venture between uh Cas um and Talus so I so we we are quite Keen uh that uh they on working on this project with them because I think it's a uh rare opportunity that uh you know local Aviation industry is going to Broad chain as well uh at at the same time we are also uh incubating some uh games project so um uh why games right is Al is because

we see that uh in games is there's a there's actually a huge potential for AI application so a lot a lot of games they they a lot of gains to create more real estate behavior um the AI actually plays a big part of it to make the Behavior more humanlike more engaging and then if if uh if someone develops a model like for example AI uh MPC or AI corrector and then if they can secure that model uh they could actually reuse the model or even uh resell the model so there will be some interoperative within the AI so uh that's one thing that uh we're working on that's why we we are

talking to also various games project and actually uh to be honest AI is still quite a new space so uh to use AI for like m mission critical task I think it's um sometimes could be uh daunting and challenging so uh we feel that uh games can be a very good test Pate to test different AI models and make sure this whole technology um gets more mature and trusted as it goes along currently um for our company I think what we are doing is that we are trying to get more and more um games or uh simply any AI applications or any projects to talk to us we have uh our

grant programs and uh we do some uh investment into some uh uh projects that we like as well so for I think if anyone is interested uh please uh reach out to us and we are happy to discuss any point in time and then uh you know to see how we can help you and you know I would say make the make your application better yeah so uh that that uh that's all I have yeah thank you