Doug Petkanics, Illia Polosukhin, Niraj Pant & Joyce Yang I Pragma Denver 2024

[Music] we are ready for our next panel and it's one of the the awesome topics that uh a lot of the audience people can actually mention this is what they want to see so I'm excited to effectively fulfill this which is we're going to talk about open source intelligence and kind of how do we actually not only just intelligently but also thoughtfully merge crypto and AI uh speaking of uh the the people we're going to be talking about these topics I want to bring on duck from Life Pier Ilia from near n from ritual and moderating this will be Joyce from Global coin research so please welcome all four of him on stage um so let's start off by uh with some introductions my name's Jo Yang I am the founder of global coin research where are research and investment Dow and next to me is deia everyone I'm co-founder of near uh but in my previous life I was a machine learning researcher worked at Google research on natural language understanding and I'm a co-author of a technology called Transformers which is a t in GPT and uh pretty much the technologist is powering a lot of the modern inov in AI hey guys uh my name is nage and uh I'm a co-founder at a company called ritual and our goal is to make it very easy for smart contract developers to integrate AI models into their products um before ritual I was a investor at Poly chain for six years and invested in a bunch of different stuff in a bunch of different categories and um crypto aai was was starting to take up all of my time while I was working and so I I just decided to to leave and and and go start something in the space all right my name is Doug Panic I'm the co-founder of a project called live Pier which is a decentralized open video infrastructure for the world we power video streaming across hundreds of applications um we in just three million minutes of of video per week and we have a large network of gpus that do video encoding which is recently being put to use to support gener AI video so we'll talk a little bit about that in the context of AI today awesome thank you guys so let's start off withad you kind of came from the AI space and gradually went into the crypto space and now I'm particularly curious about why do you think there needs to be a crypto layer in AI yeah I mean there's so for background when we uh started near it that was actually an AI company and uh back then we were trying to teach machines to code which in 2017 sounded like a science fiction and right now s like yeah I mean it works right now um and so we did not have you know infinite money from Microsoft to you know build infinitely large clusters and so we were trying to be smarter and get better training data and we actually had a lot of students around the world who were contributing data uh and we wanted to pay them and actually paying people around the world is hard uh even now and unless you use blockchain global payment system now 2018 even back then transaction fees on bitcoin and ethereum were already higher than what we paid people and so that's kind of how we started the journey with near protocol now kind of turning around now there are kind of I split it into three sections there is a data section infra section application section for kind of web 3i space um specifically for data section there's a lot of important pieces that because whatever you feed these models is what you get in the result right now most of this models actually trained on proprietary data that um somebody in some company decides what goes in and this biases the model in very you know particular ways it also means means that if somebody is like adversar injecting for example data to influence everybody's opinion later we will even not know if that engineers in that company will not AIT that for example right now somebody can be running you know bots on Reddit to post a bunch of stuff that you know a year later your chat GPT will be actually outputting and biasing you know who you should vote for or what kind of uh you know things to pick so data curation crowdsourcing orchestration of kind of data sets uh is extremely important there's also a very important piece around open source um what we see right now more and more of uh kind of reddits and New York Times Etc are saying that you cannot use their data for training and it's extremely kind of important to preserve that uh some of the website that I host open source data sets are being sent legal Tak down notices so we have an opportunity here in blockchain to actually host that data again orchestrated crowdsource it in such a way that is you know preventing that so that's a big stack on infra I'll let N rash to talk about it as they're working on that but obviously there's a lot of pieces on infra that are relevant otherwise I'll like take it for a whole hour um so yes I I think I think those are those are great um great kind of ideals and I I think taking the backdrop or the sort of meta of where AI is today I like the crypto AI area as a as a nice alternative um to to what exists out there today which is distribution is controlled by a few players and computes controlled by a few players and the model development is controlled by by a few players and I think we've seen as you know centralized companies with with kind of deferring incentives and and social incentives will start to do bad things and I think we're we've already seen it in a couple of cases over the past several months in in various different products um and so what we're building and what we're really excited about is building you know execution trustless execution around AI um and allowing you to build these new crypto enabled AI applications or AI enabled crypto applications whichever way you want to take it um and so there's a lot of really exciting pieces to be building in this space whether it's computational Integrity privacy building around model governance model development model training and so more we can we can certainly dive into and talk about but I think that's kind of a high level on the types of things that we're excited about that's really helpful and a really good context for the audience U maybe Doug if you could tell us a little bit more about how I mean life Pier has been around for a long time right um um and you guys started out using gpus for video transcoding how has that involved yeah that's right um lier has been around for seven years the Network's been live for six years we have this you know Global Network of node operators running thousands of gpus to transcode video or convert it from you know the HD formats where they're captured into whatever format and bit rate um you need to consume on your device and it's it's really you know practical infrastructure Plumbing component of video streaming which is super popular but like every industry um AI is changing the game and having a tremendous impact and uh that you know certainly extends to the world of video with one of the heaviest and most popular forms of media that we we all consume that ranges from you know how video gets processed or delivered in terms of these optimized trans Gooding upscaling jobs and whatnot which the gpus perform but also to how videos create and I think many of you in this room probably saw this really exciting and interesting moment that occurred um you know a week and a half ago when open AI released a preview of their generative video model called sora and you saw how all of a sudden people can type prompts and generate 30 to 60c Beautiful video clips and it's really amazing it's really inspiring but like Naraj was alluding to that's a closed proprietary model which is you know created by open Ai and access is restricted or biased and then the amount of compute required to generate that 60 second video might take you know 30 minutes of GPU compute which is very expensive and it's run on you know the gpus that Nvidia makes that only Microsoft has access to buy and we know from looking at llms from looking at diffusion models how quickly open versions of these models are going to catch up um and are going to become available um but you also need open infrastructure to you know run that heavy compute that heavy inference to produce these AI video models and you need a surrounding video stack for video ingestion storage serving analytics um optimization and coding Etc so that's how we think about why it makes sense to put um you know additional tasks Beyond just trans goting uh onto the live p network is that uh you need this full video stack to keep up with the way the video is changing yeah thank you for that I think we addressed and touched on kind of multiple layers of the stack so let's talk about um where you're thinking about how crypto and the AI stack within our um kind of decentralized um space is comparing to what we're seeing in web 2 and kind of the close-ended models and software that's available so meaning um what kind of resources do we need to have in a decentralized world to catch up to the centralized entities and centralized competitors there's probably a lot and maybe you could talk about a few that think really yeah I think there there's like a whole kind of at every level of the of the stack there's something that we need to improve on I think maybe to just touch on one moment is when we say crypto and AI right I think like people kind of jump a lot conclusion of like okay there's a crypto currency involved that doesn't always need to be the case the decentralization can come in in the form of running this models on your device on you know for a specific use case and um a lot of it is powered by the open source uh and kind of AI development that's happening now open source AI development has pretty kind of uh broken incentives because you either working for a big company and you open source and because it's your compliments so like Facebook opma amazing work but they're doing it because it's not their Core Business and they're using it to attract a lot more people to work with them or uh you're trying to push actually envelope but then you have no resources you need to incorporate a company raise money and then you're spending ton of time figuring out a business model when you just giv an app all your models for free right and so one of the kind of aspects that I think is important here is creating the incentive structure that can actually facilitate development of Open Source software and still have kind of value capture that is kind of open and transparent to everyone and like Fair uh and allows to fund continuous development and I think again across every pieces of Stack we see that as the main tool and principle and I think that's like a really important property that we using here and so it doesn't need to be like okay well you know we're going to be running models on blockchain which is like some people are asking it's more about like how do we use this Primitives to really benefit the open source AI development as well as to provide proofs to provide kind of better tooling faster more effective as well as not manipulable not accessible by um you know other Pary yeah I think there's like a good good example here which is I mentioned like the the GPU shortage and just like the restricted access to the compute you need to build um AI platforms you have to go to one of the big clouds um but there's so much more compute that exists in the world um in gaming laptops in GPU mining op mining rigs um that can be coordinated and put to use to create sort of Open Access AI compute infrastructure um you know to run open models and if you look at like two of the superpowers of blockchain Technology two things it's really really good at one is like massive instant Global coordination through incentives so you if you use cryptoeconomic incentives you can solve the Chicken and the Egg problem you can bootstrap a the supply side of these compute networks very quickly on a global scale um and that's one of the superpowers and advantages that dpen networks have and the second is just like near instant settlement of Global permissionless Payments which is necessary in order to create these marketplaces where anyone can access open AI infrastructure pay you know infrastructure providers or model creators that are running them and contributing to them you know permissionless without needing to be on the same currency or the same payment rails or be approved by some gatekeeper and I think like those two Primitives um are what you know crypto plus AI coming together will enable yeah I would just Echo what what um you know dog and ilas said just now there are things that when thinking about building you know how to build these applications in crypto AI like what are the best things to build and I think we should really lean into the strengths of of crypto um crypto because it's it's this tool for self- sovereignty it now allows you to um you know in effect kind of own your your identity own your data and and we've built some you know the the crypto Community as a whole has built some really amazing things off this you we have a permissionless payments mechanism we have a permissionless application protocol and um leaning into those things and figuring out you know what are the most enticing interesting things to build in the in the space and whether it's um doing um you know sort of autonomously updated defi or recommendation systems or generative uh generative art um there's a lot of exciting stuff to build that I think will draw um you know the Curious open source AI folks uh into the space even further so we have a lot of Builders here and I want to address some of the questions kind of coming from the builders who are also building in the intersection of AI and crypto I think the number one often come up question may be Talent right how do I find researchers how do I get AI uh talent to join my company either whether it's from a you know large comit from web 2 or someone within crypto how have you guys been hiring in that respect and what advice would you give to the builders um the way that we've been thinking about hiring we're we're building at the intersection of these two spaces so we we obviously have a ton of crypto people um the hiring AI people has has not been an easy thing and and also hiring people at the intersection has not been easy as well I think part of it is because there's kind of a lack General lack of ml education amongst crypto people whether you're a developer or not and I think a lot of people also ask about you know what are the most interesting use cases but I think once you do the prerequisite ml 101 education you start to see how these things can can kind of intersect in interesting ways you know if crypto is sort of like a verifiable computer you think you can think of it as um AI sort of like a statistical computer it's able to take these like vast swaths of data and you're you're generating this like compressed intelligence that's um that that's highly effective and so I think in thinking through these different angles I think it's really finding in you finding what the most interesting pieces um you know of of what you're building is to an AI person some people might be really attracted by the open source ethos and how crypto can help accelerate that some people may be attracted by the the ability to build a whole new swath of applications some people might just be really interested in technical problems privacy is really difficult and you know people are working on it Integrity is difficult governance there's a lot to do for you know AI people whether they're you know developers researchers policy people um so that's you know generally how we've thought about hiring but you know it's still early and I think as this space grows I think that'll only help the the cause in bringing more AI people into the space and getting more crypto people to to kind of learn Ai and and kind of build at that intersection can you just share a bit about how many people are in your company so that we kind of have context from each of you also sure yeah uh 12 people for us thank you Doug or Ilia yeah we're approaching like 30 people within the you know the core team and maybe another 20 that's sort of orbit around and contractor contributor um Landscaping out to the AI point I would um I would Echo that like a mission orientation around open AI or decentralized AI resonates a lot stronger than like figuring out what crypto plus AI means um which can be a little scarier people will approach with skepticism but there's so many AI researchers and developers that don't want to see you know open AI plus Microsoft or Gemini be you know the only AI platform of of the future and really want open alternatives um and that's attractive to people and you show you show them and build you know earn their trust about how crypto can help enable that rather than need them to take like a big leap of faith yeah I mean for near itself right is more of a platform that you can build applications like this you can build uh incentive systems and so for us it's more working with great Founders in the ecosystem and as well as going out and working with AI Founders and seeing how we we can actually bridge the gap so like I Eno inves in in some thei companies as well who for example building a new economy you see economy that I'm like well there's an opportunity there to actually integrate some of the crypto Primitives uh to actually Bridge Gap and so I think a lot of it is kind of educating about open source and find finding places where uh you know individuals Builders Founders actually resonate with the uses of blockchain without actually like driving in into the kind of crypto specifics but more like hey how can we solve this problem right you know lots of people creating new generative assets how do we you know how do we create a Marketplace for them well actually we have tooling to do that that's way more effective than you building like a huge system around it we have you know open source projects that are not getting funded uh well we can have a system around that so like how do we create this um and systems and leverage blockchain for that and then working with those Founders Yeah you mentioned like creating new generative assets it's interesting we haven't we've talked a lot about like building the stack or building the infrastructure we haven't talked a lot about like the users or consumers of of generative Ai and in that world I'm interested in it going the other direction where you actually have crypto oriented Creator economy Builders people building nft projects like decentralized social projects thinking about what can I do in terms of creating media using generative AI so if you can you know create a 10,000 nft set using generative AI you can look at generative video and say how does this help me create interesting like music videos that um can be owned uh on chain and and sold and distributed I think there's you know we're at each Global event there's going to be a big hackathon this weekend I think this like intersection of crypto plus media plus plus generative AI is an area where you're going to see like just an explosion of new awesome things being hacked and created um and these Stacks should be key to enabling not not to mention the this that is true for so many different modalities text video audio um and so I think it both these creative tools can both help kind of assist existing creat but it also increases the T of like who can be a creative you know like you might be creative but not a creative um but that's still kind of a beautiful thing that you're able to do now with with this um this generative AI tooling that that's out there yeah I would say like a specific example um somebody like one of the founders I'm I'm an an invester and they build a AI game engine which allows anybody to build a game by just explaining how it should work there's actually like really cool like they have cool kind of people who built it and their comments are like hey you know I cannot code right I'm a designer or I'm like a Storyteller and I'm being able to now build this like interactive experience which you know can be a game can be more of a story tell like storytelling through interaction and then when something doesn't work we just like fix it it literally type that in and it's actually tries to itate and um address that issue so there's like a huge opportunity in like where we have ai startups now which are cre something new which exactly enables new modality that you would not be able to be that kind of Creator before without like ton of specialized training and and experience and now I mean Sim similar with Sora like you can you know be a Hollywood director by typing in you know a few lines of uh text so the opportunity is huge now would there will be a flat of those things happening there there needs to be a market on the other side to actually balance it out there needs to be some form of like other I it and scarcity and reputation attached to it to actually build kind of and like maintain that system and that's again where blockchain is technology to power that yeah I mean I haven't seen any projects that's PR pushing you know crypto to the Main Street narrative right you leveraging AI for example that's actually you know explaining and breaking down blockchange in simple formats or helping Bridge a 80-year-old person to get crypto instantly by leveraging AI have you come across any of that because I feel like that's we've been seeing a lot of projects building you know and on the infrastructure middle layer middleware layer but like what about the the the top layer where we could really push crypto to the main street I mean we have like couple projects that are uh allowing to interact with blockchain through natural language problems so literally in chat GPT there is a GPT from mbase which allows you to M nfts and kind of interact with some this from GP Cher like literally chat and you don't need to know about like accounts transactions Etc and so I think that interface is going to be interesting and then another project kind of working from a different direction which is you just describe the user interface you want and it generates that and then it can hook in into smart contract and other things on the other side so you can kind of generate yourself an interface for interacting with applications now I mean that's again a new type of creator that we'll see that like are not developers who are now building kind of custom ways to interact with some of the maybe Financial Primitives or Creator Primitives that are and marketplaces that exist on chain but that's kind of pushing that to this and user level right where I'm as a user now don't need to like understand apis you know RPC calls and kind of transaction formation you know rlp en coding but actually just like hey you know create me like a you know a Marketplace of videos and uh you know be able to upload new one and then you know be able to sell and price it with you know first price auction boom generated you know and you know potentially still relying on existing smart contract but maybe even helping develop that smart contract as well so I think there's like few different pieces coming from different direction yeah I also I I don't think the moment occurs when the mainstream 80-year-old knows that they're using crypto plus AI like I I think the moment occurs when there's crypto enabled AI where the crypto is hidden or transparent so you know in live here in the video world it's like oh yeah the other day there was like hundreds of thousands of people watching like Fred Fred again you know stream his concert um through the lot Radio website which is like powered by a crypto enabled Network Live pure under the hood but like none of the viewers know that they're using crypto right it's like crypto was used to coordinate and incentivize the infrastructure the power this thing and so in AI Creator economy Social Network apps you're still going to you're going to see and get the benefits of what AI enables for the mainstream um because you know crypto coordination and incentives enabled an open version of the software to exist for developers to build as opposed to it having to be a you know Google project there's a chain abstraction Summit going on on this specific topic which you know trying to remove the barriers for entry and obstru the blockchains for the users but we not a BL at all but we're also here now listening to these great panelists here at pragma so thanks for being mobile here guys I think um I have two questions but I We R out of time right I think there's just so many questions feel free to find these guys and like pull them aside and ask them as many questions as you want while they're here but I think one question I always have you know is how do we make decentralized AI I think following up to what duck said maybe centralized AI will just be the underlying layer and no one actually ever know you're using it so people may not even care about it but then how do we actually make someone a consumer decide hey I want to use this decentralized um AI option instead of the centralized AI option so I think there's a couple different angles I I would go back to the kind of like the values that you're going for with with crypto AI so one might be you know let's just take a a couple of random examples one might be a user that is using a big box llm that doesn't like the results that are coming out and want to go use another open source alternative or doesn't even have to be open source um they will move to ideally something crypto enabled maybe there are incentives in place to help improve keep improving that model and I think we're starting to see this already um again given that these companies are kind of steering the models themselves um I think there's going to be a lot more need for more of a kind of user-driven approach and how you know what kind of models they run and what types of applications that they're using I also think there's a bunch of unique kind of web three enabled crypto AI use cases um I think things around imagine if there was a little Michaela type artist uh you know fully for those that aren't familiar Lil Michaela was a you know or is a 100,000 plus follower account on Instagram it's a fully virtual character but it like sells a ton of merch and has a bunch of you know Watchers and I think being people being able to create those types of things and it's just uniquely kind of innately enabled by crypto where there's maybe some sort of crypto token associated with that character or you can buy nfts um I think the like leaning into the consumer web 3 elements I'd love to see more companies building in the space I think is another interesting way to kind of think about all of this yeah I would me would say fundamentally it needs to be driven by either new products new kind of economic value new entertainment value right anything new yeah but like at the end it can it cannot be just based on kind of you know oh this is decentralized or this is open source you know people mostly don't care uh but the fact that it's open source means you can build on it way easier the fact that it's you know decentralized means you can plug it in you know into Financial system really easy so you have like this new benefits that you know as developers you can now build some new experiences that you couldn't on for example chat GPD and also even if you could CH GPD tomorrow can just close their API check up the prices you know you know neuter the model even more so it doesn't actually work as well as it was yesterday and you will not be able to have any control over that and so like you have these benefits for developers and so now kind of the models need to catch up to on the quality um at the same time you can have like specific you know customized models and again have a model Marketplace for example to leverage on that so there's like opportunity and kind of some of the you know future protection that you get from leveraging open source and decentralization and then that will lead to application that people want to use because they're uh yeah enable something new for them another question I had was you know we're in a growing I guess burgeoning or currently already a bull market right Bitcoin just hit 60k this today um how do you guys think about the price of crypto when it comes to adoption of your kind of the intersection of kind of decentralized models and how that could be potentially pushing that forward into the attention of um currently centralized players right what what I mean by that is you know when um Bitcoin went up to 50K or when the ETF passed it brought to it brought public attention right and that's where you started getting acknowledgement from Banks you started getting more um direct recognition and approval from regulatory maybe around the world um do you see that happening for AI and that kind of not specifically about the pricing action but the the fact that if the crypto industry does well it potentially push the projects within um the AI project within crypto further so what I think the the nice thing with with prices going up does is it removes a lot of the the skepticism that the mainstream media and and regular people have about crypto when when crypto prices are low I think even crypto people they start to they start to maybe lose sight of the the fundamentals and um they get a little bit too in their own heads about oh does this have any value what's going on here and then at the kind of opposite in the Spectrum you know when prices are up you know sort of anything is possible and I think you know that kind of helps in bringing in new people and you know ultimately you kind of need that cycle to to bring new users into the space I remember in uh 2021 the the nft pfps was like that was the way that regular people were starting to use crypto they were like going onto Moon pay and paying with a credit card and bunch of celebrities were starting to have their nft pfps so um I think it's what it does it helps enable more attention in the space and ultimately can help drive more attention towards crypto AI because people are more sympathetic open to trying an alternative and more open to those kind of new values whereas I think at like a you know low market price everyone's like open AI is the only thing that's going to work and nothing else is going to work so that's just my my general view on it yeah I would say like having built through you know three Cycles at live Pier we've seen almost zero correlation between like crypto prices and actual demand side usage or demand for the network right like people are building video apps regardless of whether crypto prices are high or not high and like the product that they're using is either good or not good to meet their meet their needs has nothing to do with price but um to Echo narage you do see um you know easier access to Capital to enable you to build and fulfill your vision and um you know much more interested and open-minded uh talented product people Engineers business people you know congregating around the space to learn about it and consider consider building there um there's a lot of noise that comes with that signal but a lot of that you know signal does stick around and and uh do the real building and so that's where the benefit is another awkward or question that may not have been brought up as often is regulatory capture and I think I'm curious whether you guys have been approached about and thinking about regulatory capture you know coming from the the large Tech conglomerates and how that's affecting or what they're doing lobing DC affecting decentralized models and decentralized AI um any dots there yeah I mean I think that is one of the kind of uh both big risks for AI space right now that there is a lot of really strange kind of executive orders uh data laws that are coming out that are like trying to address what people p perceive as an issue with AI which really not and at the same time kind of creates yeah like creates an opportunity for big companies to have um kind of a more solidified position that others cannot displace and I think there are few fundamental things that are happening um one is this kind of like attachment between the capabilities of models and you know number of parameters which is really kind of you know somewhat arbitrary based on like I maybe maybe like two years worth of research of like hey it looks like models are improving if we put more parameters like that's not really proven that that's a real thing um and then on the other side kind of the requirement to have like all kinds of departments that going to be oversee the model like deployment Etc which is again like you if you're a startup you cannot do that if you're individual you cannot do that so so those are like really kind of big issues there's also Al I would say like one fundamental thing which is the copyright laws existent in the pregenerative world they existed to when the transition happened from kind of telling the stories to you know printing books and distribution to actually reward the original author like this was you know prec crypto way of rewarding people and so now we have generative models which I can go and like generate every single potential tune of a song and copyright it and then everybody will be viol like every new song will be violating what something I just generated right it'll be like every possible thing that sounds plausible to human you just generate it like you know in 10 hours there a lot of 00s and so the reality is now the copyrights is kind of in like needs to change it needs to be adapted to Jared of AI and and so those are all like all the pieces that right now kind of creating a pretty big mass in in the AI space and so open source is like one way to contract it web 3 and some form of anonymity is actually the other one and so I think these pieces need to come together they need to show that you can have kind of a model that for example if we have a model that is fully you know decentralized running from uh different um Hardware you cannot enforce any of those laws on it you cannot require it to you know follow something and so I think those examples will also help to frame better some of this Regulatory and actually reshape it to address the real problem the real problem our society is not ready for J by our society systems operating system the content the how we interact with government none of this is ready and again cryptography is the tool to use to fix a lot of the stuff so I would say like the regulatory right now tries to pretty much address very surface form problems they're not trying to address the actual problems and I think we have an opportunity here to Showcase that you can actually do things differently and at the same time showcase that you cannot enforce those things and to draw a parallel when there was in beginning of ' 90s there was a um potential that cryptography will be considered a dual used technology meaning it's considered military grade and you cannot export it and so there was a person who walked across to printed out the algorithm on the t-shirt and walked across the border to Canada to be arrested for violating that and then they went to court to pretty much prove that this is freedom of speech the algorithm is freedom of speech it cannot be considered a military technology military uh expert and so we kind of need something like that right now in AI where to show that decentralized you can actually decentralize Ai and run it and you cannot enforce any of the rules that are right now being considered and you know actually go to court with that and prove that we need to actually have different rules that are way more sane and addressing real problems yeah for um for those who aren't aware here in the United States there's already an executive order that's been signed by the president is already in effect and there's a series of Milestones each month that determine how anyone working on AI must you know comply with regulators and um it's very you know arbitrary if you're training a model over a certain size you have to you know comply with and report to the what is it the Secretary of like Commerce or agriculture or something there's you know different departments that you have to comply with to prove that you've tested your models sufficiently but it's not defined what it means to test those models so um who can comply with these and hire the lawyers to document everything they do and make the case to defend if there's an action later like only the big tech companies not a startup that's trying to innovate really quickly and be disruptive to those tech tech companies and by the way those big tech companies are the the ones who gave the input and lobbied um the politicians for the creation of these actions so as Ilia you know alluded to um de centralized and open AI is actually you know an alternative that can guarantee access to the rest of us to accelerate the same way that you see big Tech accelerating towards this yeah the example would be like if I have a flash drive right now is a mistal model like llama 60b and I give it to you this potentially can be like a violation of you know that executive order it's a mun they declared ammunition yeah that sounds super scary and so how how do you guys think about or revising projects building AI now and where they should be jurisdiction and building and their teams launch adap launch adap is that he said da Dow Dow oh I see I I mean I think there's like few different parameters what one is like if you're building like on top of some of the fundament foundational models right if you're using things like ritual uh it's actually less of a challenge because you're kind of using that as infrastructure it's pretty easily swappable unless you started to go in um now if you're building foundational models that's where things is becoming super complicated there's also like data center kyc requirements which is like if you're running on in a data center you need to kyc and data center needs to report you uh so there's like all kinds of strange things that's you know that like literally rolling out right now after April uh in US and so there you may need to consider you know having like domicile in some other countries I would just mention Switzerland as it's being a very pop positive for uh crypto it's also like it's outside of EU even though it's kind of in the middle of it um and so it so far has not introduced any of this um challenging rules and so it's a good opportunity to capture both good crypto and uh AI uh jurisdiction for your setup that's helpful and last question um what advice would you have for Builders now building an intersection of kind of AI and crypto you know I think my question primarily also is regarding you know where is the stack are we really lag lagging the most that we can really be uh making up for with um direct evolvement and where is the opportunity right so I'd say the first thing is um is the ml 101 is is important so go do that and then the next thing I would do is um I would go and just again start thinking about um you know again what did like chap GPT uh staple diffusion what did those things bring us that really valuable as consumers and clearly it's really really valuable if hundreds of millions of users in the matter of a few years and thinking about like what Contours of that can kind of align with the Contours of crypto and crypto product building and and so there's a lot of interesting areas and I think a lot of the people in this room have maybe a subject matter um area that that they're sort of an expert in maybe it's defi maybe it's nfts and thinking through like again what are these properties that are kind of uniquely um enabled at this intersection um that you can that that you can build applications on top of and I'd love to see more consumer apps I mean I think to me one of the biggest opportunities of this space is that I think it can be one of the greatest drivers of new users into crypto because the barrier to entry is so low like I don't think my mom is ever going to use defi um but she very well and nity does use chat GPT and and diffusion models and things like that so again if you can find the unique value of crypto whether it's incentives it's the privacy and computational integrity whatever um or it's you know the Open Access new way of distribution there's a lot of really interesting tools to to play around with and I think I think I would just you know just tell people to experiment and kind of see what comes out yeah hack experiment innovate like find that you know crypto Kitty moment that just onboards you know 100 thousand a million people into the space and inspires a whole new wave of applications because I don't think we have enough product Visionaries thinking at that layer right now but now's now's a really good time to do it yeah I would mention specifically on the data layer there's so much opportunity right now between like actually just making sure open source data sets don't get taken out and like leveraging you know decentralized storage blockchains for Registries crowdsourcing you know data curation kind of getting and incentivizing people you know bringing data that you know data marketplaces um and then from there going to like content marketplaces and types of those types of things as well which I think we see a lot more people want it to be creative with Chad GPT and being able to and so like how do we actually bring that economy because don't think open AI will be that interested in that and so I think those are kind of you know interesting aspects to start experimenting with this has been super helpful let's give it a round of applause for our panelists again thank you so much for your time everyone um really great panel you