The Satori Testnet, an early implementation of Vana, is now live! It is a distributed network that runs on computers all over the world—including yours if you want—to enable people to collectively own, govern, and earn from AI.
The recent progress in AI makes data ownership a critical problem. Data has been brought to life by LLMs and applications like ChatGPT. Trained on masses of public web data, these models and applications result in tremendous revenue generation for AI companies and for the platforms providing the underlying data, but to date, users have been excluded from that value creation (1, 2).
Data should act like equity in a model. It’s not about being paid $300 once in exchange for your data – it’s about contributing your data so you can own part of the AI revolution.
However, tools to make data user-owned and monetizable for AI models haven’t worked. While crypto promises that users own their data, existing crypto infrastructure is not designed to work well with private data, the key to data monetization.
Today, we announce the Satori Testnet, an early implementation of Vana. Like Bitcoin, the network runs on validators all over the world, creating a permissionless ecosystem controlled by its users. However, Vana has been designed from the ground up for private, user-owned data to enable collective ownership of AI.
Data, the core resource driving AI, is most valuable when aggregated. Currently, web2 platforms like Reddit, StackOverflow, and Twitter have a monopoly on aggregated user data, which allows them to keep it locked away in their walled gardens while monetizing it themselves for hundreds of millions of dollars (3).
Data liquidity pools, a novel cryptoeconomic primitive, flip this dynamic. Users can now mine and contribute to the network by exporting their data from platforms, leveraging data rights to sidestep big tech's walled gardens. Proof-of-contribution ensures data quality by rewarding users proportionally to their data's impact on AI model performance.
With Vana, this valuable data remains non-custodial, putting users in full control. Just as you can use your funds with your Metamask wallet or add funds to a liquidity pool on Uniswap, you can now do the same with your private data using any evm-compatible wallet.
Already, the Reddit data DAO that launched last month saw 140,000 users join, with a machine learning engineer now training a collectively-owned model on the data. The Reddit data DAO contributors share:
Proof-of-contribution ensured only users with good data could join: 995,000 users attempted to contribute data, but only 140,000 users successfully verified via proof-of-contribution.
Builders are now working on the ChatGPT Data DAO, the LinkedIn Data DAO, the Twitter Data DAO, the Github Data DAO, and more.
Phase 1 of Satori Testnet is focused on giving Data Liquidity Pool creators an opportunity to test proof-of-contribution for their dataset. It also gives node operators hands-on experience running a validator. To get involved, find an overview here.
Thank you for your optimism and for believing in and working towards user-owned AI. We can’t wait for you to dive in.
Anna, Art, and the Vana team