Scarcity in the Era of AI Superabundance | S.Y. (​​Seung Yoon) Lee - Story Protocol

uh we want I want to really focus and deep dive on the state of AI uh the state of Ip in this age of AI and kind of talk about what we can offer as a solution so from the perspective of intellectual property the biggest thing about AI is the fact that it represents insane amount of content getting pumped out at a very short scale so look at this graph so the amount of images that actually AI produced in 1.5 years is equival of amount of images that's actually been produced by uh photography for 150 years so it's a basically it represents an insane amount of super abundance of content at an incredible speed that's unprecedented and really trying to abstract and see the having the we want to think about having a right mental model of to think about the super abundance I think Ben Thompson of Strat actually had a really good way to think about how to think of AI what AI is doing doing to IP and he basically says every single technology disruption unbundles a part of creative process writing unbundled consumption writing and literacy unbundled consumption printing unbundled duplication internet has unbundled distribution anyone with a social media profile can instantly distribute to everyone in their world with zero marginal cost in the same way actually generative AI is unbundling substantiation you can have an idea a bullet point of ideas about your next fiction next movie concept and it will actually turn you turn it into a fully-fledged fiction Pros movie anything and actually what's actually probably the most incredible thing that came out recently is what you've seen from Sora and anyone as a dog parent would know how hard it is to get your dog to look at your camera and actually you get this kind of incredible result with the sentence so the whole process of substantiation is gone which means there's going to be insane pace of content creation because you no longer have to have this lengthy process of substantiating your content to have a fully PR fleshed product and this is going to get worse and worser even you know it's it's it's going to it's the pace is going to get greater and greater right because the amount of compute that's going into uh AI is exponentially increasing actually that Bel um uh M's law almost like went down went from 20 month to 6 month after 2010 the amount of data getting into actually training this model is explanation increasing too which means the infra is accelerating and actually the pace of this AI pumping out content at a faster speed and actually having higher resolution substantiation will never decelerate it will only accelerate so we will see exponentially growing abundance of content uh com going forward and in that environment I would like to argue actually the gating problem of AI is not the speed or performance or getting to a higher resolution substantiation it's actually attribution Providence and compensation it's the intellectual property that is the problem right and don't we have actually an IP infrastructure already you may ask it's a copy law right but the copyright law is all based on this idea of mechanical reproduction right the copyright law only knows how to actually punish or protect or incentivize if someone's actually copying your content right but it doesn't but AI is not a copying machine it's not a printer it's un printer it's a creative generative machine that takes away takes all this data and gobbles up the data takes the essence and creates something totally new totally new cre creative generative and copyright law which thinks in terms of mechanical reproduction doesn't know how to wrap their head around right uh this generative creative machine they've only know how to deal with copying machine right and what we are seeing is that the copyright law only knows how to cover and deal with the previous generation of content but actually now the inter internet is getting carpet bumped by generative AI now you'll see most of the internet's content being generated by gen uh models in that world how do you assign Providence attribution to the content that's gone in the contents that's gone out of the AI models so we need a new way to new IP infrastructure that's going to cover this gen uh age right and and actually this is already playing out you see the New York Times open AI lawsuit right you've seen the Getty Images uh you know stable diffusion lawsuit and New York Times if you ask about New York Times article it gives a incredible synthetic output right and what New York Times is actually arguing is that this is actually worse than what the web 2 big Tech the the previous generation uh practices that big Tech used to do because at least back in the days big Tech was decent enough to drive some traffic to your websites right it gives your website at the top of the search result and you can actually go to that website and we will make some money from that traffic now there's no need to actually go to these websites right you get the synthetic output really incredible synthetic output on the AI interface so what it means is the AI interface will get all the upside of this incredible things that come coming out of this uh coming out of this incredible technology and there's going to be no traffic nothing getting back to the websites or entrepreneurs or creators and this is actually a classic tragedy of common situation think of internet as Commons actually all of us especially the creators are creating incredible outputs into this commment and actually you know these AI models are using these Works to create an amazing AI model but they're taking all the offside the creators are not getting benefited from this um AI Revolution and there are three actually visions of how to actually you know how actually now the current big Tech is trying to solve this or deal is dealing with this the first is the kind of the Adobe model which is a wal Gard model but wal Garden model which is we're only going to train on the imagees that we own right but this is incredibly bad for actually creative of the internet because this is an incredible technology and only the people with Incredible number of images can actually benefit from the upside of this revolution so it's not actually incentivizing entrepreneurship and creativity the second one is the open AI model let's not put fences around these IPS where we'll take away all the data where we're going to charge for it they make all the money but the creators not making money at all right the third is the open source model we we're going to get all the data Works we're not going to charge for it but still the creators are not benefiting right so if this internet goes into this direction constantly what we actually think is we it's going to actually get to the tragedy of commment the internet is going to become a Barron land you may see the flourishing of genetic content but the humans don't have the incentive mechanism continue to pump out the content that is going to these training models which is actually a very bad outcome for all of us right so what we need is actually a new IP infrastructure to address this massive challenge that's going to put us to the brink of tragedy of Commons and I just want to give a quick historical context if you think about basically IP infrastructure every breakthrough in technology requires a new way to incentivize creators basically that's what IP infrastructure IP layer means and the printing technology was the very first outbreak by unbundling duplication and the writer and author didn't have a way to actually own have a claim of ownership over these copied works and only when copyright law became an IP layer that's administrating administrating this kind of new paradigm of Technology actually you saw things like shakespare and Charles Dickens come out in the same way we are at a very big revolution like the printing technology generative AI abundant substantiation there's so much new content pumping out but there's no way to actually get Providence attribution right and now with that actually we're getting to the tragedy of common situation so I think you may ask what is a better way to forward in the situation right is there a more sustainable model to actually fix the situation and I think we argue that there's a two main principles around actually having a more sustainable future of the internet the first is that creators actually have to own their content and creators should be able to set licensing terms around their content how other people can use their content they should be able to assert their rights and Licensing usage terms over to the internet and these terms have to be actually machine readable and machine writable including AI models so the AI models can actually go and parse through all these data and all these works and get very very actionable next steps in terms of what they can do if you want to actually make a commercial model you should par through all this data and actually exclude the ones that actually ban commercial users it's only for noncommercial usage right only when these principles are actually enshrined and upheld we think we're going to have a more sustainable future of the internet so there we think there are multiple path to this go to go to this future but we have a workable solution in the name of programmable IP layer and we want to explain what that means to all of you so the programmable IP layer is basically where the IP layer where IPS are captured on chain so the creators own their IPs but creators can also set self-enforcing autonomous Rules of Engagement around their IPs and these rules are actually machine readable writable including AI models and AI agents and to actually give you a very quick comparison because most of you are from crypto the best analogy is I think programmable money programmable money took away all the middleman in basically financial transaction financial intermediaries and the beauty of programmable money is that now money is code there's a built-in API for money and money can flow to from you know Unis swap to Ether uh you know to compound and anyone can build so many different type of applications on top of this programmable money right in the same way what we're doing with programmable IP is now actually this creative assets can interact with applications and software in a very frictionless permissionless Manner and many different people creators and software can build interesting periers on top of programmable IP and you may ask you know has this been already been enabled by nfts right but nfts really is what it is at the moment is a pointer to a media file right it gives you ownership and scarcity to this media file but nfts at the moment are very static it's kind of Trapped what I mean by that is a program like let's say you have an AI chat uh character AI Style application let's say you want to take all the board Apes you want to take all the pedy Penguins to create AI new AI chat Bots how do you do that you have to actually talk to there tens of thousands of owners talk to a bunch of lawyers and have tens of thousands of legal negotiations going on in parallel right the nft doesn't have the programmable layer and composable layer to be machine readable writable you know let alone any AI models right so what we actually do now is we have this programmable IP that we've created as a primitive now with these programmable IP have onchain rights enshrined in their IPs and now any software including AI models can actually read and write on top of uh this nft uh this IP asset frictionlessly and when we talk about programmable IP it's basically what we call IP asset and there are two elements to it first is what we call the noun noun is basically the essential metadata around the IP and not only just pointer to the media file it has the relevant onchain licensing onchain royalty information around the IP and how we enable this is actually through the token bound Acom model ERC 6551 and once you register the IP on our protocol you get a story protocol bound account and actually the smart contct wallet that you get through the token bound account model will store all the onchain licensing and onchain royalty information into this wallet and how we actually enable all these functionalities is through our verbs this is these are the functions these are the enablers of our PR Primitives and actually what we do is you have a an IP asset getting registered onto our our protocol now with the module like licensing module you can set permissionless licensing terms around house humans and also softwares can interact with your IPS right you can set non-commercial commercial usage you can set royalty terms you can say you can say it's you want outr licensing you know fee you can also even choose arbitration method you can even choose whether you want to uh uphold Ledger Authority or legal Authority right and once we have all these Primitives we have a fully endtoend programmable IP infrastructure and actually now all the key three principles of what makes a better future sustainable future is all there creators actually own the IP creators can set the rules around their IP and machine can read and write on top of these rules really really easily and permissionless and we're don't think we're going to do all these alone we have a lot of Partners building on top of us to actually make this more sustainable future possible right you know the first application is really registering training data or training data sets as IP Assets Now all these data or data sets will there have their onchain licensing terms commercial versus non-commercial how to split their revenue and actually you know we have some interesting hackathon projects joining us to actually innov innovate on actually the training side of things secondly what we're doing is also the AI modelss can get registered as IP assets all the outputs of these models are also IP assets on top of that and we have a Providence and attribution around what's actually getting sped out of this model and AI models themselves can be forcable you want to use someone else's I uh ai ai model and fine-tune it there's now onchain licensing terms around that not only that we have actually Partners like magma which with 2 million artists actually now they're registering their finished asset and actually saying they have a proof of original work or okay this artwork was mostly generated by AI all these things are actually getting in getting attacked to the IP as an essential metadata to basically uh describe whether this is human generated or AI generated right so now we have a very interesting primitive where everything from training data to the AI model and the outputs of the AI model having essential machine readable data around what what thises uh what kind of unchain license in terms of will have whether it's human generated or AI generated and some may ask I think think okay how are you going to get all these IPS to actually have these unchain licensing terms and I think when people ask me that I would actually like to point out the case of Napster versus Spotify and when Napster came out everyone was downloading the downloading music for free right for 10 plus years no one really solved this issue and really it's when a very small startup in Sweden called Spotify started working with very few recording labels in a very small country small market and really showed a more sustainable future things that things get started to be solved and had people saw a better future right so what we need to do in I think the world of crypto because crypto gives now scarcity attribution and Providence we should Ally all with each other and one another actually show to show how actually this IPS with the right onchain licensing terms can be a better future for creators and IPS overall thank you so [Applause] much