The Future of AI is Collaborative Agents - TOKEN2049 Singapore 2024

good morning it's great to be here thanks for joining me bright and early I know everyone had a lot of fun last night kicking off token 2049 um very much uh excited to share with you uh our vision of how the future of AI as collaborative agents um maybe to start I'll give you a little bit of my background uh for me I've been Anon entrepreneur most of my career and working in AI for over a decade um I was working in an adtech Pioneer using AI to build lookalikes called quantcast then I started an Enterprise AI company and sold it to terod dat a large database company um I was at Canada's largest AI research lab Vector Institute working with 500 researchers really scaling the future of AI research and I went to Google where I was responsible for apped AI in the CTO office and in my experiences working with AI in my career I became increasingly convinced that AI was going to be the most transformational Technologies of our lifetime but I also believed that it was incredibly important that it land in a positive place in society so specifically um even working inside Google where I thought people were were well-intentioned for the most part and very capable I still felt and saw that um the incentives were all around behaving like a bad monopolist not really doing more than paying lip service to the way AI impacted the future and I felt it was really critical to do something about it so we started theoric me and my co-founders are AI veterans who' worked with me at other places like at Vector Institute and teradata and and Google right and we really felt that it was important to embrace web three principles to make sure that AI benefits all of us and is controlled by all of us the community not a few billion s who make it something that benefits them and increasingly tell us what we can do now very specifically I'm going to talk to you about AI agents agents I'll tell you more about what agents are but at the highest level it's about making it so AI can do the things we want we hear a lot about the models and chat Bots right but we really think that the future is a gentic that more and more you're going to have capable AI systems and we passionately believe collaborating AI agents working together to accomplish great things so which Future do We want do we want a few billionaires to create agents that tell us what we can and can't do that they control that they extract more and more of our data tell us take more and more of the economic value and increasingly become the dominant power in the world or do we want it so that we all have a voice in that that we all have a chance to influence that everyone has something to contribute we passionately believe in that latter future and that's why we We Believe web 3 principles are so important so at theor we're building an open AI agent Innovation ecosystem a base layer where it's possible for agents to work together they can be composed permissionless anyone can teach agents how to do amazing things they interoperate and there's communication and payment all of that built in we think that's critical to making sure we have the right future for AI I'll share more about that as a high level as a company we've raised over $10 million our last round was led by hack VC and we're a team of veterans with AI experience as well as web 3 experience with leadership roles at places like Ripple and Tron so to give you a little more detail about AI agents they are software programs based on the power of large language models the chat GPT type engines um made so that they have access to all kinds of useful data and tools a lot of people you probably heard about all the data that goes into training the models but really what's important is the um the way that agents can access data in real time right what can it look up right now and understand to take action right so for example we see a lot of excitement around agents helping people analyze tokens and trading whether it's for meme coins or more stable tokens right so what kind of social data can you access in the moment what kind of index blockchain data can you use right so we're uh we see that is really important important the data that agents have access to and the tools right the ability to run code to do analysis to call apis to do searches right those are the kind of the core capabilities of an agent and Incredibly importantly agents are able to do planning and reasoning just in the last couple weeks we've seen big breakthrough um in Frontier labs open AI in this case showing how agents are getting better and better at planning and reasoning coming up with smarter and smarter ways of doing things we think that's important but we also think that it isn't going to be a future where there's a one-size fits-all agent some Universal agent that does everything instead it's about specialized agents working together in collectives right so in the case of like analyzing uh defi right you want to say hey are there major fluctuations in token prices maybe I should go in and make some modifications to my Holdings you can have at the one level a group of Agents a collective that one knows a lot about socials and can look up sentiment and commentary other that looks into the news with vetted Publications and then also we can have a collective of data analyst agents that can go in and crunch the numbers right it can query data from a a an index database like space and time it can generate analysis code and visualization code and it can interpret it and give results right so breaking it down into these substeps working together you have a much more powerful system just like a company a department is not one person with universal skills but Specialists working together so these Collective we think are already working really well and driving meaningful value in the way that our uh system works and we see this moving forward very rapidly as the underlying AI keeps advancing its capabilities its knowledge keeps advancing so every month we're seeing big improvements right like we've built a powerful data analyst Collective and it's gone from uh about 70% accuracy to 95% accuracy just in the last 6 months rapid progress indeed now collectives are exciting because of three key things one is specialization works well right everyone has something to share so developers can build specialized agents and we really believe in a future of no code where anyone can just teach an agent from what they know what they've written what they've talked about with some feedback make an agent better at the things that they know and be a force multiplier for their specialization bring those those together and that specialization is powerful Dynamic assembly is another important thing we believe in which is making it so you can discover the best in the latest agents and have them really come together as a team and work well together that's incredibly powerful things are moving so fast in AI you constantly want to be updating and finding the latest and greatest capabilities so we're building for a world where millions of agents are available to do the work that we all need and by having these collectives of multiple agents it gives you a lot of resilience right you're not dependent on any one agent any one model any one particular provider you can have Alternatives you can explore you can you know for high stakes task you can perhaps divide it up and have two competing efforts working together so resilience comes out of that approach as well now in order to power this future web 3 is important the first thing is around having standards right standard ways of communicating smart contracts that ensure the way agents work together that ensure payment right so that you have Micro payments agents can pay one another and they you can pay for use of AI models and data you know this this is important token incentives is incredibly important right I think there's never been a bigger opportunity to build a giant Network than AI agents and that network will Thrive with token incentives right so giving incentives for people to test agents to give feedback to create agents by having us all have a stake in it it will grow faster and get faster feedback back than the slow progress of large monopolists trying to tell us what to do this community ownership also comes in is incredibly important to have the road map and and the needs driven by what we the community want and it's really important to have the accountability so you can see on chain what's going on how the agents are interacting and what's working so you can have independence of how to assess what agents are good or not some of the use cases we see are that are important here in web 3 for agents include around defi right so analyzing portfolios helping with trading both in terms of analysis making it easy to express intense in natural language um building more automated trading liquidity management but also in predictions right so prediction markets are hot agents are are emerging as a really exciting way of participating in prediction markets so is for forcasting and general analysis in business developer tools like auditing smart contracts generating uh generating um whether it be meme coins or um daps marketing all kinds of like analyzing for sibles and fraud as well as generating helpful content to engage community in gamey whether it be you know NPCs I think increasingly gaming Studios we're working with are seeing the the future is more and more how do you have teams of agents that you as a player can harness to to help you right and for modding AI is really exciting and more generally in the whole community space agents can help Drive Community can help people understand governance flag important issues sort of be a force multiplier and drive up engagement in community to drive all of this it's incredibly important to have a knowledge of reputation what agents are good which ones work well together and so there's really two key things that we think um that innovations that we've brought forward right one is called proof of collaboration which is using non repudiated the non non-modifiable storage anchored on chain you can see how both individually agents contribute to results proof of contribution so you get independent feedback of how well did this agent work was it providing useful results a lot of times people may talk to you about things like verifying the Integrity of AI computation but that's just a subset of the problem of was the result useful was it helpful did it achieve the goal right so number one in a collective the collective as a whole work well together proof of collaboration did did I achieve the the intended goal was it effective was it helpful so that we can learn and help with Discovery and optimization of how agents work together and then the individual participants proof of contribution what did they do was it helpful can you Benchmark it can you give get user feedback so this combination of AI and user feedback is important to drive reputation and help Discovery because we think it's incredibly important that unlike a web 2 approach to uh marketplaces you need to have uh economic incentives to to give good high quality signals and have better ability to update as new and better offerings come on the agent Marketplace I'm also very excited to announce that our first test net for theoric is live today right now so you can go and check it out interact with about a dozen agents that we and partners like bondex and filecoin Foundation have put put forward you know whether it be analyzing Investments understanding filecoin uh understanding the igen layer and reaking ecosystem you know job preparation um with with bondex so go come check it out we'd love your feedback um our SDK and documentation are live so anyone who wants to build agents and register them in our protocol can do so and we really excited to have more Builders building great agents for the community and then we'll have an incentivized dashboard coming out early next week so that you can all um have a stake in the success of the protocol as you participate um from a road map standpoint we're very excited having entered this phase of test Nets to have some test Nets coming up following this with a no code Builder so anyone can create great agents you don't have to be a programmer and uh allow for pay payments and onchain audit leading into a test net or sorry a main net late this year and into next year we've got a whole series series of exciting updates coming up um around our uh Community governance and staking and continuing to iterate on on community feedback and Discovery but everything in the road map is something that we want the community to give feedback so we're excited at the launch of testnet to start this process of learning and you know we're excited we're collaborating with a number of great AI projects around uh dpin and inference and data um because we believe that truly the power of web 3 is the community and us all working together we can achieve a better future for AI so do come check us out check out the website at the. feel free to give us a follow on X at theoric Ai and Ron Bodkin and feel free to connect with me and uh again we really do believe the future of AI is a gentic and we as the web three Community have a chance to make AI serve all of us and be truly beneficial for the future of humanity rather than controlled by a few billionaires so come join the agentic revolution thank you for your time today