Comment by @ruthless • Hey

Ryan, I appreciate the time you took to articulate your thoughts in such detail. However, I see several flaws in your hypothesis that I think are worth dis

Stats

Comments

There's also the fact that Ryan says that if someone trains AI on his data and the AI produces new work then the only way he can benefit from that is to prove that the AI's work was trained on his data, but you can't spend all day searching the internet for AI work and trying to guess if they trained their AI on your data or not, and then, if they did, proving it. There's not enough time for that, and if you're popular it's just impossible. The entire idea falls about at this point because you can't manually prove that AI work was trained on your data, especially if you're popular. It needs to be automated, but ofc, it's impossible to do that.
Definitely appreciate the discussion, writing is thinking. 🙏 First of all, that's exactly right with Collectors and Supporters being an old idea. It goes back to the art world and the Renaissance and even before that. My grand thematic message is that what we're doing with tech is almost always a reshuffling of old patterns--hence my banner and bio. What is different this time is that the database we're all using here is public, and we can choose to display whichever metrics we wish. Instagram forces us to focus on Followers and Likes. Orb and Lensport gave us Top Supporters. A new app could get rid of the feed completely and focus on Supporters, Users, Friends, Stalkers, or any other metrics imaginable. Whatever is possible to calculate between the database and the app can become a metric which is displayed to users. Second, I think perhaps you and @larryscruff.lens are thinking I'm being definitive about this being the only path to do things. This is very much **NOT** the case. I'm laying out one possible future that's more efficient than what we're doing right now. By no means is this the only method. When talking about building web3 apps you'll usually see me on here telling devs they should build just a small part of their app on Ethereum, use a private database for the rest, and eventually migrate as the tech evolves. I advocate for getting the end user to the result as fast as possible, regardless of the tech. Currently I see a public database staring at us, devs who can build anything at all, and a bunch of artists worried about companies stealing their data for AI training. Seems clear to me what can be done about this, today. **Intellectual Property Ownership:** A blockchain is just a tool, like a trademark or a copyright. It's not required. There are many paths to ownership. Let's take the blockchain out of it and stay old school for a minute. Star Wars GIFs and merchandise are derivative works. By definition they're separate. A Valentine's Day GIF with a cartoon of a little green man saying "Yoda one that I want" is new material. A Darth Vader mouse pad is new material. These things haven't changed the source material. Generative AI does the same. I'm not sure why you think it's different? It does it faster and at greater scale, which is why people are freaked out, but it is the same thing. It takes existing work and generates something new. MidJourney photos of Storm Troopers haven't changed the original films. You could create immersive Star Wars experiences in VR with brand new stories, or even reimagine the original films to be set in New Jersey using the cast of the Sopranos. None of that changes the original work. Hollywood is full of remakes. Pyscho, Scarface, The Karate Kid even. Better or worse, the original work remains. Value can be added, subtracted, or completely orthogonal. Same is true for your onchain data. **Public Data Ownership & Value Attribution:** I never said a blockchain offers more control. By some metrics that's true, but by others it's not. There's a lot of nuance here. A thumb drive which never touches the internet gives its owner a great deal of control in terms of restricting read/write access. It's lousy at granting global read access, controlling data monetization policies, decentralized backups, immutability, etc. Everything has tradeoffs. You do not need to put everything online or onchain. You can post anything online or onchain in open or encrypted formats. Putting anything online comes with assumptions, three of which include: - If you post something online you probably want someone to see it. - When you post anything online it will eventually become publicly visible forever. - Big Data is going to scrape everything you've ever posted. If you're not posting something online then this whole conversation is probably moot. So let's assume you're posting online and that right now we're all just giving away everything we've ever posted. Given those assumptions, what should an artist do? I argue this is another Napster moment. In 1999 consumers were stealing music. How did the industry respond? Well by suing at first, which was wrong. But then by creating digital stores and streaming services. They created an even better product than what the consumers were stealing. It was easier and priced correctly. It worked. Now imagine you find an artist on Instagram whose work looks amazing. Imagine there was a button on every photo that said "Add to AI" or imagine there's a "Generative Reels" button on the top of their profile. You tap it and suddenly you're making brand new Reels in that art style. Maybe you want a green screen background using their art, or maybe you need a new phone case and IG lets you generate new artwork based on all the posts you've collected over the years. You hit the button and create the art using the collected artists' data, you upload the file to the manufacturer, the artists get attribution, and you're probably going to share that art on IG, promoting those artists. This scenario is a case of users making use of/adding value to other users' data. IG isn't on a blockchain. Lens is. Either path could work. Lens devs don't have to wait. It's not a matter of "everything in public gets owned by the public," or "everything onchain gets owned by the user." It's about being set up to capture value. Piracy was quelled but not eliminated. The music labels, Apple, and Spotify captured value that was passing them by on a daily basis, and in capturing it they actually grew their opportunity. What often wins are platforms, and the thing most people haven't grasped yet is that every single NFT can become a platform. It is a profoundly important development. Let's imagine you spent 12 years on automotive subreddits. You talked about engines and car parts for 12 years. Who captured that value? Reddit did at first, and then OpenAI did too. Multiple winners. Could you have as well? ABSOLUTELY. You could have turned your data into a platform itself, where users come to ask for your expertise. You could've used that data to create a platform which teaches people and enable others to create automotive content as well. Reddit, OpenAI, yourself, and anyone else could've used your data to win. It was easiest for Reddit and OpenAI. It was hard for everyone else. Blockchains make it easy for everyone now. Platforms are sets of tools which can leverage the public/users/fans/etc into adding value to the original data. Every NFT can become a platform. In the Reddit/automotive example it's hard (not impossible) to gather your data and turn it into a platform. An NFT has an owner/user table and can be used immediately in any app. We are root users of the blockchain, we can gather all our data in seconds. Again, it's not the only method, but it's like creating web sites vs publishing on Facebook. The efficiency is there. You're absolutely right in saying the value attribution isn't just about the data source, and that it's also about the use of that data. That's actually entirely the point: Currently, social media users are posting tons of data and it is really difficult to make that data useful if you're not the platform. Like the record labels, we are just letting value pass us by until we build useful apps and platforms. In the previous comment I said my data would accrue the most value because it's first. Important to clarify! This was assuming I built the thing and had the most useful data. This was assuming my data was worth $100 Million and you were using it to create small-time work. You are absolutely right that a blockchain does not ensure that the majority of the value goes to the original data. Using a blockchain means it can, though, and rather efficiently. A blockchain is probably going to be more efficient for most users in terms of centralizing their data and allowing others to make use of it. But users still have to do the work! You still have to make the thing, the app, the platform, whatever it is. It's easier now, that's all. The magic of all of this is that I, a bad artist, can enable others to use my art and make it more useful/valuable. I might make some art, and three other people make some better art, and then the next Beeple comes along and uses our art to make the next $69 million art. Sometimes the Hollywood remake is better! If my art goes from $3 sales to $20 sales because the next Beeple made $69 million using my data, that's still a win. **Relevance:** Without dwelling since I mentioned it in a previous comment, I'll briefly say again another big factor in determining value is relevance. Staying relevant is incredibly important, it's why there's Brand Advertising. Everyone knows Coca-Cola exists, why do they advertise? To make sure everyone knows Coca-Cola exists. Star Wars has a licensing model for merch and Disney doesn't sue its fans for making GIFs or fan films because they know if the property loses relevance it loses value. I would bet heavily we'll see some generative AI Star Wars products in the next few years. We can see today the musicians who block their music on YouTube and other streaming services are losing relevance, listeners, and value, every day. The Mona Lisa is valuable because it's on tshirts and mouse pads, not despite it. **Key Takeaways:** AI models have already taken all of our data, and that knowledge will live on forever. There's no taking it back. I assume there will be more models coming which are trained directly on my data. It's not a matter of me locking down my data and controlling every single use. It's about being set up to capture value. It's about being able to incentivize value add from others. It's about turning my data into a platform. We used to rely on platforms like Facebook for making our data valuable. Blockchains let us do this directly, anyone can add value. AI lets us make infinite use of our data. Useful data is valuable.