Why User-Owned AI Matters More Than Ever with Sreeram Kannan

Transcript

if we can be on top of it in the earliest days of AI now with data synthetic data like Ledger doing there's teams doing models clearest reason for crypto to exist of course we have the Bitcoin and the ethereum and the decent L all that stuff that we love more UNL welcome to igen lay unlocked an interactive educational Journey Through the igen lay ecosystem created by technical Founders and builders for the entire crypto Community a special thank you to our partners allir polymer authentic skate and lrange for helping to make this happen our goal is to raise the collective knowledge of the Igan Lair AVS ecosystem and unpack the technical designs of the top teams

in space welcome to I lay unlocked s we at the be AI Summit hey you had a great talk oh thank you always have these te points and I think the beginning of this is the two parts of society that grows C it's coordination and Innovation so with this kind of framing of of of Mind how does decentralized AI come into this where does this happen yeah so if you look at like you know to expand on that the idea is that there is an underlying you know successful complex systems have an underlying layer of coordination on top of which free agents innovate right so

it's the coordination layer and the innovation a simple example is coordination layer government Constitution things like that Innovation layer free market where free agents can freely exchange goods and services and when you think about Ai and what what has happened is AI and in general technology massively accelerates the free agent so each individual has more power and so that means more Innovation happens in in the system but the coordination layer is still the old like hundreds of years old Technologies like Constitution and like the double entry accounting and money and so on and what happens in AI is there's a debate

between is AI going to kind of like break safety somewhere and the the the solution that people are going to is let's go to the working coordination systems like the government and get regulation yeah instead I think a different way of thinking a crypto native way of thinking would be to say Can these agents bind themselves into coordination conditions using crypto and blockchain as a substrate and one extreme example of this would be like all AI are controlled by a human da right a human da basically unless the human da approves an AI cannot do some actions or you can terminate an AI and things like that that'd be like an extremal version of a coordination layer which is cryptonator which can control

like an uh an AI or an innovation layer there so but there are all kinds of inter intermediate things that we could do right can you incentivize more decentralized AI can you incentivize the person with the data to work with the person with the model can you work can you incentivize people with different models to come together to create meta models can you incentivize uh you know U all kinds of different things can you make sure how do you bind a model to uh blockchain yeah right like how does a blockchain control a model right and you know sentient has this thing called loyalty which is that you know a model will not do something unless instructed by his master it should be like a blockchain

yeah and there are all these like and these are new computational problems actually like these are more fundamental theoretical problems that we need to solve but I think the direction this is much more exciting in my view where instead of saying oh technology is going B let's bring the Regulators in who move at like you know hundreds of years old technology instead let's upgrade our coordination layer and build a Pang Global automated self enforcing coordination system yeah I think um not necessarily just for the the sake of decentralization but for the sake of of who gets to kind of have the the end say in the systems and and where

these models and this data is originating from is also very important because I it's kind of fly wheels and can create this Butterfly Effect I think where if we can be on top of it in the earliest days of AI now with data uh synthetic data like like open Ledger doing there's teams doing models inferences if we can be on top of this now then the the ramifications in the future are kind of built on a strong Foundation absolutely I think this is the clearest reason for crypto to exist yeah even exist right like of course we have the Bitcoin and the ethereum and the decent L all that stuff that we love go and love but today if you look at like what is one of the most urgent problems of like Society say how do we

grow and how do we make AI more decentralized how do we make it bind to commitments that are not you knowable yes they're verifiable and like self enforcing that that is really a superpower that we can bring and there a short window of opportunity for us as a crypto space to actually actualize it right because when we think about all kinds of other things imagine was there before Facebook then the right Facebook would have been built on crypto rails and we are kind of in that phase right now yes there is a chance for the best AI in 3 years or in five years to be on crypto rails and crypto has the advantage especially in like payment rails and you know like intent based

designs for I want to do XYZ on chain multi-step processes there's things that on with credit cards with trfi with bank accounts that you just can't do transactionally in the current setup of like web 2 stratify that blockchains enable absolutely peer-to-peer interaction is the blockchains were made for AI and for AI agents but it's just about how we capitalize it and how we build it in the right way now yeah and so just kind of closing thoughts on how you think as a community we can kind of capitalize on making sure that we are are building the strongest Foundation possible yeah I think like you know you mentioned you know crypto as statement rails that's one direction that AI absolutely will and you know youth the

other one is crypto as the incentive layer and tracking layer for oh I contributed data you contributed the model you take 3% I take 10% like all these kinds of coordination mechanisms Bound by contracts and going beyond that I think uh there is a lot of really powerful use cases where verifiable AI right like I I I do an AI inference and I need to know that the answer is correct because I can use it for various things Downstream like intelligent D5 is one simple example that we talk about there but in general I think there is a short window of opportunity right now for us to build all these substance Pro of personhood is another one how do you make sure that I can distinguish between a human and a AI right especially with

the on agents and things with it's going to get really weird going to go crazy really fast the other one I wanted to say is intelligence is by nature decentralized right people are decentralized and crypto has a mechanism you know one of the things that you know all the AI people are talking about today is running out of data yeah what better global data collection mechanism than crypto yeah right like you can incentivize people really easily not just for like pay for use but also as owners in the system so I think the scope for what sets of things can be done on crypto is actually quite large yeah and on the other side I think AI also has huge things to add to crypto

like which we are not really doing right now yes one of my like favorite things that nobody's talking about is I you know people say oh you know these legal contracts are so difficult to read smart contracts are much better but my experience being a crypto protocol for is not not bad okay I can read and understand legal contracts I barly understand like smart contracts because you know one underflow bug and like your money is all gone correct right like that's just extreme yes nonability and so I think the my thesis is in the longterm contracts will be return in English and will be synthesized in quote underneath the human layer contract is

WR in English and then you come in and trade with it underneath some llm already getting there we we're already getting there like it's going so crazy so fast yes so in in ethereum one of the things we think about a lot is client diversity correct right like oh there are many clients and they all come to consensus what about contract divers yeah can I have a contract in English and then synthesize it into 100 different languages with different y with implementations like they exactly and then they all have to come to consensus otherwise your state doesn't progress and if they disagree then you Fork yeah so all of these things I think there are superpowers that we're badly seeing as a community y but I think the crypto xai is the most in lot of people think it's like some new meme coin thing

no I don't but I think there is just a lot my first my first tweet about key takeaways from this conference was literally crypto times AI is the most interesting and and we're just stretching the surface and we you know we pay some attention to it but we're so caught up on like scale ethereum and interoperability and these things and it's like a lot of these problems that we're like thinking about constantly on like in like this kind of like modular ecosystem are can be solved by AI or at least AI can have play a role in that yeah ABS so and and conversely for example you know one of the questions was in the panel how do you uh have like these models which run at such insane scales and blockchain throughputs are so small like how are you going to kind of like how where you put all these data

and I think like a lot of the thinking from the modular Paradigm can actually play a big big role here because one note can do the competition everybody can verify it and so on so all those things I think there's a lot of things to be done at the interface but I think this the time sensitivity of it is lost on people yeah like that that's the Keen takeaway here is that we have this like transitory period Transit period we need to do something capitalize on yeah thank you J thank you so much this awesome