A 'Paradigm Shift' Is Happening in AI, Fetch.AI CEO Says

Coin desk, Michael Casey notes that one of the most talked about topics among crypto folks at the World Economic Forum in Davos last week was the intersection of A I and Blockchain. Joining us now to discuss this is the CEO of fetch dot A I Humay, Sheikh Humay. You welcome, great to be here. Great to see you. Great to have you. It's great to have you and I'm going to jump into a very tough question. Uh It's almost the kind of question when, when you're looking to get hired, they ask you, you know, where do you see yourself in five years? This one is where, where do you see the intersection or the junction of crypto and artificial intelligence headed in the next five years? Yeah. So we, we've been working with this intersection for the last four or five years anyway, so we have a very clear picture on where it needs to be. And uh I'm so glad that everybody else is now starting to see that. So Blockchain as a technology is, is really great for record keeping and actually inclusiveness. And I think the current status of A I the models the machine learning model training is, is one where a record keeping is pretty poor. Uh inclusion is pretty poor. You're actually not. Um people are not making sure that the training data where it's coming from, what the copyrights are. And ultimately, you know, once you've trained these models, you need to actually monetize them somehow. And that's another very interesting thing which currently people are not focusing on, but that, that focus will come because how do you actually keep a record of monetizing these models? Um Unless you give those models to bigger corporations um If, if individuals or a collection of individuals are making these models and sharing uh these models for training uh and also um actually getting some um output from it. How do you monetize it? How do you, you put it and Blockchain is um is a very good technology for both those things. So we, we've um we've always been a proponent of using Blockchain with A I with machine learning training, the models, keeping a record of the data and then actually also monetizing and actually distribution of wealth or the economic value, which comes out of these. So are you saying that that will actually happen in the next five years? It's already starting to happen? So I'm sure it will be less than five years. But yes, definitely. OK. And, and why do you think that the A I sector has just gained so much more interest? Uh within the crypto space, like within the crypto space. I mean, on your own, you can talk about A I and it's, it's in the hype cycle but just within crypto, everyone is like a Iaia. I, why is that been the case in the last year? Right? So something controversial a little bit. I mean, crypto generally looks for something in every cycle and every upcycle. Um A I seems to be there but there, there is, there is some very interesting but genuine reasons why we should be talking about um all of this. Um When I say crypto, I, I'm more talking about decentralization because crypto as a token or market trading is not really where I, you know, I would like to comment as such. But if you think about OK, so we, we got this open A I, we got this generative A I which is delivering some value, right? We, we see the value coming through. There's no argument that people don't see the value but that value that has to be converted into actual um actions, you know. So at the moment, you know, creating a generative A I um model using it for just textual purposes is one thing, but then you actually extend that output and then you convert that into actions. That means somebody, um somebody will have to monetize it somehow. And if you are looking at this new wave of A I, uh you need to one have inclusion from uh whoever is going to use it. And then also you have to deploy it in a way that when people use it, they actually can generate and create and actually distribute that economic value. And it's not just in big corporate hands. And, and currently, if you look at it, you, you can see um the whole infrastructure at the moment is mostly in four or five big companies. Um Yes, there is open source on the rise, but that's a model. But you know, if you think about what open A I did, which was not meant to be open, that's the whole point. It's not open because it was meant to be open. You look at it and you think, well, OK, who is paying for it? And you see Microsoft or bigger corporations are paying for it so it will never be open. So there are multiple aspects of it which need to be taken care of and crypto as a digital currency, as a decentralized ledger technology can play a very vital role. The question is, is it actually going to be allowed to play a vital role? And is it actually playing a vital role that those are two questions which I wanted to come to, that? I wanted to come to just that like what vital role it will play, what use cases it will have. And I want to cite a report from the Boston Consulting Group. Um you know, it noted that A I autonomous agents could enter the mainstream in the next few years and fetch dot A I is your website effectively. It also notes that A I agents will re invent the way we live and work. What are the use cases? I come back to that? And, and by when do you see it happen, the start is happening right now. But, but it's, it's a paradigm shift. So if you, if you look at how, how things change in the past, every time there is a method of communication change of communication method. Every time there is a method of search and discovery which changes, then you see a paradigm shift and it generally takes some time to get to that paradigm shift just to give you an example, yellow pages, phone phone books, we used to call people to find them and then we used to, to kind of transact came along, came the website, the mode changed. Now we find them on websites. So hence e-commerce was kind of made e-commerce resulted in a lot of big aggregators. If you look at the top tech companies at the moment, they're all aggregators. And when I say aggregators, they are aggregators, information, the aggregators of services products, whatever you look at it. Now, what is happening with this new wave of A I is that you don't need those aggregators and these models can actually help each other. Now, how do these models interact with each other. How do you actually carry out these actions? Is, is the autonomous agent because autonomous agents starting from nothing is a piece of software which connects, communicates, finds um services uh products, different machine learning models. Because now if you think about it, there's no, let's let's just think of a world where there is no aggregators, there's no one company controlling this big model. Then the agent's job is to find you the right model for the right purpose. Connect the models, connect the services, connect the people, connect the businesses and then deliver you your objective. That's, that's really the point because if you look at what's happening with um chat GP T with LL MS you enter your objective, you give it a prompt and it kind of tells you what to do, but it doesn't actually do it. So you need something or technology to do something that's when the agent come in and it breaks it into smaller chunks and puts it together to actually deliver the final service. So you can see if, if we believe that yes, um generative A I is going to be the thing and you think A G I and intelligence is going to be the thing, then you definitely need to have the new way of delivering the value out of those. And the new way of delivering the value out of those is the autonomous agents. You know, very interesting because I was coming to uh the A I agents because a recent post on X it showed that an A I agent meet up. Yeah, an A I agent meet up actually took place. It was held during E ID at India offices a lot happening in India. Do you view collaborating and educating with these types of big tech companies as part of your strategy to bring A I into the crypto mainstream or the mainstream? Uh is that the only way to go? Like, how do you see this, the new technology needs people to be educated for sure. I mean, as you said, we we did, did a whole hackathon with the agents meet up and everything we did that on in Microsoft India as well. Uh We were part of that. Um a lot of people very interested the, the the beauty of this technology is that you don't need to build these uh overly expensive complex systems, you can actually build components and then you can string them together and all actually automatically they can string with each other. So if you have the right uh framework where all of these agents can actually connect with each other, then you have something very interesting because then you have empowered everybody to actually create something and then be discovered and then be deliver the value out of whatever they're thinking now that that's a, that's a huge paradigm shift. So, but, but the problem you have is that we are used to one thing and we're now kind of converting to another thing which luckily everybody understands that that's what needs to happen. So we do need to, right. Ok. You know, I want to talk to you about some dangers because you might be able to throw some light on this fetch A I and singularity net. Recently teamed up to curb inaccurate or ir irrelevant outputs also known as A I hallucinations, which is fascinating. Uh what are some of the dangers associated with this? And, and how do you aim to fix it? Yeah. So in a very top level way, the the hallucinations are, it's the strength and the weakness all in one because you want it to be um innovative, you need it to be the, the A I to be more than just answering very straight questions from data. But what the bad bit is that when you actually look at doing something specific uh like an action, you know, you, you need it to be very deterministic, you can't have it probabilistic. Well, I want to travel from point A to point B and if somebody suggests you could take a, you know, a spaceship or you know, you could have a, you know, beat me up Scotty kind of arrangement, then you know, it's not useful, right? So you need it to be very specific. So that's what causes problem with hallucinations. So the the the kind of the collaboration we're coming together is to really take a very specific data sets, making sure that the actions that we are delivering are very deterministic and are are not impossible actions. So hence, no hallucinations and looking at different combinations like rags and neuros bolic A I uh putting it all together to actually make it much more useful because hallucinations are OK when you're building pictures or uh creating some videos and making a movie, but when you actually start looking at doing physical things, they have to be exact. So for example, you don't want to make a transaction which says, well, you know, I'm sending $100 or I could be sending 1000 or $1 million right? You know, it could be anything. So that's uh that's me a million dollars, but that's fine. It doesn't work. Ok? Uh uh Yeah, I can understand that. I understand the need for being uh deterministic. Uh uh But thank you. Uh I, I'm probably gonna have hallucinations about this conversation going forward. Uh Umay, thanks so much for joining us. That was, that was a lot of fun. Thanks so much.