Comment by @ruthless • Hey
You don't need to collect anything onchain in order to take someone's data and train AI on it. In fact I could already take all your posts and make train A
Comments
- Yes, blockchains are not required and I stress that actually. Star Wars doesn't need to be onchain. It's easy to prove who owns Star Wars and what that IP is, legally. When someone posts a Yoda GIF there's no confusion over who owns that IP. Star Wars GIFs and parodies and fan art do not siphon or steal value from Star Wars, they add value. The original IP gains value with its cultural relevance. The more people who see Star Wars content, in any form, the more valuable the original films become.
A blockchain enables everyone on Earth to do what Star Wars does. It's not required but it's more efficient, in the same way posting on Facebook was faster than creating a whole web page or printing a physical book.
So let's say you scrape all of my Twitter and Reddit posts, feed them into AI, and now you're creating content in my style. Where does the value go? Well, partially to you if you're creating new valuable content with it. Partially to me, depending on if I can prove I actually said those things and created the original works. It may be hard for me to even quantify that value, though, let alone capture it if I don't own the database. I can't track anything. For this reason the value goes mostly to the actual owners, Twitter and Reddit, as evidenced by the recent API fiascos. They own the database, they own the data, they capture the most value.
Now let's say you scrape all of my Lens posts, feed them into AI, and now you're creating content in my style. Where does the value go? Well, partially to you if you're creating new valuable content with it. And now that I can prove I actually said those things in an incredibly fast, public manner the value goes to the actual owner...me. I own the data, I capture the most value because it's provable what the original work is.
So the first point there is to own your data.
If you don't have a publicly provable method of doing this you're likely going to struggle with capturing and quantifying the value of your output. People and corporations are going to take your data no matter what, wherever it lives. If my data lives in Reddit's database, Reddit captures the value. If my data lives in a public database in an account I own, I capture the value.
Ok, so if you own your data onchain...how can you let people use it?
How can you let people add value to your data? Just like Star Wars doesn't pull down Yoda GIFs, if people want to turn your art into memes let them. If you want to scrape my data and put it into Generative AI go for it. More relevance is more value for me to capture.
In fact, what if there were tools that made this really easy and actually incentivized people to add value to my work? I argue artists should be making their own AI products available. Create tools which let others add value.
If you publish music on Spotify you can get listeners. Listeners add value to your music. If someone remixes your song that will add more listeners and thus more value to your original music. So how can you encourage people to remix your song? By releasing the stems. Make it easy for people to use all the parts of your song to create brand new songs. Gen AI is the same thing. Train AI on your music and let people create new music with your catalog of song data. If you, the artist, don't create this product then Spotify will.
Okay so how does this apply to social media? We're here on Lens where everyone can collect posts from each other. Why would you collect a post? It does (virtually) nothing right now. None of the above requires collecting posts. But just like releasing audio stems makes it easier for people to create, we need apps here that utilize Collects.
Give people tools which use Collects and they will be incentivized to collect.
- Create a LoFi music maker which uses your music NFTs to create new songs.
- Unlock new Besties to talk to in FriesGPT by Collecting posts.
- Combine your Collects to create new brand new Besties
So finally, the term "Followers" is a metric based on the news feed. If social media apps had no feeds you'd have no Followers. Myspace didn't have a feed, they called everyone Friends. The term "Follower" is so engrained right now only because every app has feeds.
Metrics are measurements. Instagram and Twitter internally have way more metrics than "Followers" and "View Counts" -- they choose to show us only a tiny amount of what's actually tracked.
"Collectors" and "Supporters" are new metrics starting to be used here on Lens because we own the database and we can see who is paying who, and therefore if we want to incentivize people supporting each other we can display those metrics. Orb and Buttrfly and Lensport are doing this, and I'm sure there's others as well.
So if we can Collect and use that data in apps, the metric to add would be User. If I Collect your songs and put them into an app which uses my Collects for remixing music, it's now something in the database we can track. We can clearly see I'm using your data in an app, and we can see the output. If I collect posts from @jessyfries.lens to use in FriesGPT then I'm a @jessyfries.lens user. It's a metric which can be displayed because it's onchain. It was hard to know this before, but now it's very knowable.
Hope that helps.
- interesting take 👍, there are def two ends to this spectrum. Guess it boils down to who uses a users DB first and perhaps more importantly by when.