Provenance and Intellectual Property (IP) Rights in the Blockchain Ecosystem l KBW2024

founder and executive chairman of Anoka Brands guys please take it away thank you um yeah so welcome to the panel I think as uh we heard I think it's going to be a very interesting conversation especially given all the interest and excitement around AI that has uh happened over the last few months and I think to start it would be good for you guys to give a brief intro about what you do and uh what you're work on hey everyone I'm shanan I'm a CEO and co-founder of Sahara AI also a associate professor at USC I've been working on AI for over 15 years most concerned about the ethical um issues that are creating uh imbalanced economics and losing of user control over data and intransparency of all centralized AI provider so s AI would really trying to prioritize um users sort of sovereignty and control over their AI uh we're building a blockchain platform that prioritize the Providence and sovereignty of AI and trying to build a sort of a copyright system that help us to really bookep how AI been developed who are the stakeholders what are the key transactions that's that's been involved that we can use this whole provin structures to back Trace attribute and do a fair compensation to all of the stakeholders uh I'm Richard Moss founder of Quant stamp uh started a company more than seven years ago and my original background is that I was a Quant so I was using um machine learning and AI to trade the financial markets uh and over the last years I've you know angoy invested in a lot of AI startups and it's one of the most important uh issues of our time so really happy to be part of the panel my name is Yan I'm chairman and co-founder of animal Brands we're probably more known for the fact that we've done today over 540 investments in all sorts of web3 related projects a lot in gaming also Ai and this a great place pleas to be here to have a conversation around Ai and the future of Ip rights thank you so um we've already been having some interesting conversations uh as we started and I think uh to start with yet you know we were discussing is AI generated art or AI generated uh you know technology or or assets can they really be considered intellectual property so I think it's an interesting question uh it's somewhat philosophical and also ethical uh so I'll let the panel fellow panelists talk about some of that too but generally speaking I would say that the difference is that I think AI today doesn't really have free will in the way that we think about it right when you think about free will it's not just can I make a decision by myself it's also accountability right I'm I responsible for the action that I took and if you you know it's like in the real world if you don't have free will which is what some people say then if you accidentally kill someone or you kill someone are you who's responsible right it's the same thing so same thing goes for AI if if an AI starts to take data and use it to create something is it it responsible in itself has it the ability to excise that Free Will does it have Consciousness at this point I would say no so the creator of the EI I think is the one who has to be accountable or the person that is actually sort of behind it so maybe even the data providers are responsible because they are the ones who have the the decision you know if I post something publicly then I am exercising my free will to prove or to demonstrate something that I have an opinion on and therefore I stand behind it so I think what an AI creates is not really intellectual property protectable for that reason but I think the root stuff ought to be and I think there's some interesting conversations maybe later on we talk about you know the meaning of public domain um you know Providence and of course um uh property rights perfect Richard or John um I think this is really like um questions depending on what granularity that we talk about this attribution problem um in a very like technical way if you ask for every single image or video that's been generated out of a model what are the um data points on the trending set of this AI model been contributing to them this is really hard problem uh for two reasons one is the computation cost that you actually need to derive this attribution going to be x times we're looking at five to 10 times as expensive as the cost that you actually spend on deriving this uh image or or or the video so it's not really real realistic to really like do this calculation every single time but then um also because all these like neural models Behind These AI models are not determinist they're actually stochastic so every single time you ask the same question to give you different uh combinations of the data points that really attribute to the to the to the prediction so um in a technical sense I I think this is not a super well defied problem to be solved but then uh I think at the higher granularity like more course gr level if you ask um who are the owners of the data set and who are the owners of providers of the GPU compute and who are the developers of the AI model recipe these are much more well defined so we are looking at a better way that we can do a forward pass we understand from the Upstream to Downstream who has been involved into the development of the AI model and application and once the application start monetizing itself through API calls and licensing we can use the bookkeep result recourse to back trades and then to give back the revenue in a sh way to every single uping stakeholder I think that sort of prominence is feasible and could be implemented by the sort of blockchain technology so I think this is kind of mixed problem that I this is my current thoughts yeah yeah I think like in the last couple years you know the cost of creation has gone down a lot uh due to Ai and so that's a really positive thing um the issue is that this soup of of attribution that it's creating um I mean the thing that I've seen that uh probably still is the best attempt at blockchain at solving that is still nfts like um at least it's kind of a tangible thing that you can send to people and um it has a lot of issues but I think that's currently the best attempt you know still looking for sort of the next generation of that to uh you know make it at least tangible right instead of this like you know just a a million pictures of you know uh K-pop bands that are generated and then put on the internet so yeah yeah I think it's an interesting problem and I think yat mentioned it in his dead talk last year right so you know the whole um issue with uh gy images suing because the AI models they were being used you know uh and I think what you had said yet was uh it's because they got caught because you know you could see Getty kind of all over the images so I don't think it's a simple solution as you said said but do you uh like how do you foresee that battle and how does it change kind of Ip law considering that it's very hard to attribute um you know you know stochastic models to what they're creating and the actual data sets that we used to create it so I guess the question is what do you think needs to change from a framework perspective so that you can actually start um you know creating intellectual property that you can then basically transfer all of that Revenue Downstream to the creators so maybe if I take a first crack at this uh you know one of the things that I really like about blockchain is I think of blockchain almost like because of its ability to have permanent Providence and the distributed Ledger I think of it essentially as the world's largest trademark office in some ways right one of the problems we have in the physical world is that to protect something intellectually you need to go to the trademark office or the patent officer you going to do that type of stuff and so there's a cost of actually creating that that becomes reachable for certain type of IP Assets Now imagine what the music industry would be like if we didn't have intellectual property rights protection from music if you know Tor Swift no BTS right no you know girl generation all that type of stuff wouldn't be possible um unless you had intellectual popular protection but the problem is that you need a certain amount of money that it can make to make it worthwhile for the legal costs and everything around it so with blockchain though you can actually immortalize your action or your words you said uh could be dispute re resolution in terms of I said it but then you can at least say who said it first and you can use it from a timing standpoint and I think this to me is one way in which I think intellectual property rights could be expanded broadly because everyone who shares a photo even if it's meaningless if he get you know if his onchain if the AI starts using some of that value of that photo could be 100s 1, one millionth there's a way in which that sharing of that value actually can be distributed blockchain can do that at scale in a transparent way that you can actually do in any other system auditable in that fashion right so I think that's a solution and the accountability still is the one who created the content at the end like if someone said hey you know what you took content that I created and you made it your own because you minted it on blockchain well the thing is that dispute resolution happens at the content level not at the so it's more like okay maybe there's a $1,000 at stake then the dispute is where does a $1,000 go to but there's no dispute that this content was created by someone um to for what value should be perceived and I think of the world as content creators every person who posts photos on Instagram every person who basically plays games even and makes moves and does stuff actually are creating content I think blockchain has shown that it can share that value in the form of tokenization like when we talk about play to earn it's not just about earning it's about rewarding your contribution to the network your contribution to the intellectual property as a form of reward that's the stage we're in but the next stage when AI is starting to do all that stuff is actually for you to basically receive value for that stuff that you generated maybe a year ago six months ago maybe just last week I think uh I think that's the future of Ip protection on chain do you guys have anything yeah and just to that point um I think like the attribution of blockchain is really valuable to y's Point uh because uh you know normally like if you wanted to get paid for something or let's say you wanted to rent an IP for a day to I don't know hold a conference and maybe use Pokemon for one day that is very very hard like you have to go talk to the Pokemon company like meet a bunch of people they might not agree and um you know it's very hard to get in contact with IP owners um and so like the ability to do that on chain or to be able to like easily pay people for that based on a certain set of rules that's very valuable and that thing basically doesn't exist today so there's like a lot of economic activity that involve smaller transactions uh or like you know renting IP for a short amount of time that's just not done today because it's too expensive and takes too much you know resources to get in contact with these people I just wanted to add one more point just just triggered me on something which is that look you talked about nfts before right actually what I find really interesting about nfts like board Apes when you own the board ape you actually own intellectual property around the design of that board ape the powerful thing isn't just that I transferred this IP for someone else to use stuff with I have created additional Network effects and distributed essentially my distribution power to a community that doesn't happen today right you know for inst from Disney I have to hire thousands of people to go out and basically do IP work for me at my payroll cost but by giving out the IP and making him a stakeholder in it basically you create an army of thousands of Asians who do the work for you for a share in that and that actually is an example of why You know despite everyone saying nfts are down whatever we're still talking about industry that's doing 10 to $20 billion a year when it was doing 0 5 years ago so I don't think that's too bad uh and on companies who are now competing with some of the biggest brands say what you will about Bor apes or pudgy Penguins or mocha whatever um these are brands that didn't exist two years ago and are up there in terms of Mind share that's something you know I think um is a good example of the power of web 3 IP just want to quickly add I me ground this to more on the AI part because I think the biggest challenge to really create a sort of copyright and IP system in AI is that you need the AI developers to really uh be comfortable and happy to use it and that creates the biggest problem because the user experiences on pushing everything on the blockchain is still really invisible especially pushing data sets and model onto there because lots of times the business model require the data sets and model to be actually proprietary rather than be open sourcing right so I think the key questions here is like you want to migrate the life cycle of AI development ments in onto a blockchain based platform but then you also need to guarantee that they feel the same as working on AWS as working on Google cloud and many out of the similar ecosystem platform and then it require sort of a really good mechanisms that you can combine the best state of yourr stuff onto this blockchain platform but I think we are really getting much closer to it today because so much things are open source say the Llama whole ecosystem from meta so I think we are in a pretty good timing for considering buing that I think that's actually doveet Tales very well into the next question which is really around you know the whole promise of the blockchain being decentralized and the ability to create kind of that Downstream effect of passing on Revenue to creators and and kind of everybody who's involved versus almost the centralization of compute and you know the power required for AI um how do you see that playing out you know over the next little while um given like I as you said they are going to need to work together to actually uh for us to actually have a system that works yeah I can take this first um I feel like we're in a pretty good um opportunity window to really convince the AI the web 2 AI Builders to come into this kind of a revenue sharing um economic models to work mainly because in the in the Cent I set up um building a prototype of AI is much more expensive than 5 years ago because so much you need to spend on GPU compute the St models out there even the Open Source One like llama is much much bigger than the model people can handle that means like you are really working on a very high bar that will start with consuming hundreds of thousands dollars on just GPU Compu that means like these Builders have to just do fund raise from traditional web to VC and those VCS now is looking at everything Revenue users profits net of the GPU cost that you spending really hard uh really high bar to start but I think um if you provide a sort of provid infrastructure where the different stakeholders the GPU compute giving you 100 hours of GPU compute taking in in its TR of 5% of the share of the AI model that you develop similarly data providers taking another 10% everyone can sort of a pull together their own resource in strange of the share which are sort of uh uh permission list and sort of trust list recorded using the blockchain system that way you create a new way of putting Capital to build AI in a much more efficient way and on top of that more even exciting things you have the whole D5 ecosystem that can be built on top to really find a secondary market value of the AI access assuming their high quality in terms of the performance of the model in terms of the track records of the developers all the usage of it so I feel like this is really exciting timing that we can find ways to combine the web3 community give them the Define sort of nature of the AI access but then you pulling the C Capital to really boost trap the AI developments in a much simpler way yeah um and just you know to that point I think like um there's a balance between the centralized AI companies and this like decentralized approach the the the issues with the decentralized approach is um there's issues with bandwidth and like cost right you know maybe like uh handling all these AI transactions it's not really something the ethereum blockchain can do nor is it something that the current ethereum l2s can do you know uh they just cap out at too few transactions per second and there's all these like data uh you know kind of uh problems that is not available so um the the benefit I think is like uh you know some of the biggest AI companies right now it's not open source right apart from llama so you don't really know like what's in the model that you're using you don't know what the original data that went into the model is right and so I think like for um you know like new technologies like zikl or opml or you know what Sahara is building the ability to see like what is in the model and what was put into it I think a lot of people will appreciate that and uh that's not really something that I think the web 2 AI companies are focusing on at all like the web uh just to the earlier Point like the web two companies there's very few AI companies that make a lot of money right now it's like three companies make most of the money so they're trying to figure out how to make money they don't you know they don't have time to figure out this sort of like attribution uh side of things so anything yet yeah and so quickly first I would say that we've seen some of that road map already when you think about sort of Ip Rights Management I think there's both the consumer and the business angle on the consumer side really the permissionless nature of Ip so I see the future as permissionless IP Rights Management uh in the same way that we see permissionless blockchains how they oper operate and we've seen this in the early days really with ugc I mean ugc wasn't official but it was basically a form of permissionless Ip Rights Management I take an anime a content and I just make something out of it I create satire I create parody I create sort of amvs and these were actually ugc and remember back in the day all the rights holders were like oh my goodness this is terrible right let's stop it stop it stop it and then actually well hold on we can give you a little bit of Revenue share from that and YouTube basically created sort of you know um Content ID and actually that made them more popular More Money More value as a result right now imagine when that kind of Ip Rights Management is not one that is only done by a corporation that holds music rights but by everyone's content and you have a global ID management system where you can track around that I think that's one major sort of area where attribution is useful but on the more corporate side talking about the language models who do we want these AI to be owned with and that to me is where tokenization is interesting right because at the end of the day we're owning a network or network effect right AI builds from the network effects it constructs these Network effects can now be owned through tokenization by the community and I don't believe a future there's going to be one or two or three big AIS I actually think AI is biased because of the content it's create right whether it's a Chinese AI Korean AI in fact a US AI right I mean for instance you can access open AI uh anything in the greater China region except maybe Taiwan for you know other re national security reasons they would say say either way you're creating different types so I think in the future we're going to have dozens of different AIS that will be like our private advisers with their biases and backgrounds and information and they will feed information to us these will all have their own different Footprints we will all need to have ownership in them as a community I think that's actually where we ensure a future where we don't have overlords of AI as a result I think as we can tell this topic is very uh complicated and detailed and and we're running out of time but I'd like to give each of you you know an opportun Unity uh to frame kind of the last thing you want to say and the way I usually like to look at it is in this type of complex industry what are the challenges or opportunities or the one thing you know if we came back next year you would say look we actually made real progress and this is what would matter yeah I think two things really the web3 would help AI web3 for AI one is the IP copyrights based on Providence infrastructure with the trust list and permission is um sort of a system number two is the really the collaborative and open economy that can possibly be pulled with the D5 sort of uh infrastructure back to the AI Capital really is sort of like a mutual bridging each other I think that will be a Healthy Growth yeah I would say like uh the web 3 Community has a lot of experience with trying to decentralize everything and the most of the time it doesn't work uh but having that experience is really helpful and it's like a you know that experience is not something The Wider world has right uh and then the other part is I just think um you you don't want a future where like a few companies control almost all like the kind of human knowledge right so that's not really a future I want to live in so then you have to do something to spread it out amongst more people and get network effects going so that it's actually the kind of economically best solution you know so yeah so the way I look at it is that we will have succcess in IP Rights Management with AI and blockchain when billions of dollars of more value goes to creators around the world than where they're going today uh T sort of interesting fact Point Spotify basically pays out 7 to eight billion dollars of royalties a year for the entire music industry which is insane in the sense that it is worth absolutely nothing nfts essentially paid to creators and and royalty holders the equivalent of almost $20 billion in that same year and that's just in the beginning so imagine what it what it could mean for the future when we use AI to create value and we create revenue streams from that actually it it'll dwarf the web 2 royalty system and I think we will see success and artists and creators will come and flock to web 3 when they realize that that's the better future as the people who are in the nft world already see so that's I think hopefully we can talk about that in the next one or two years thank you guys so much it was a privilege um you know to listen and learn from you guys and I hope everybody got got something out of it so thank you thank you [Applause]