Decentralized Intelligence Network (DIN) Abraham Nash · 23 pages
Written while a PhD student in the Department of Computer Science, University of Oxford.
Decentralized Intelligence Network (DIN) is a theoretical framework tackling AI challenges like data fragmentation and siloing. It enables efficient AI training on decentralized data stores, overcoming barriers to diverse data access by leveraging: 1) Decentralized data stores to maintain data ownership, ensuring data stays securely within Participants' chosen stores; 2) A scalable federated learning protocol implemented on a public blockchain for decentralized AI training, where only model parameter updates are shared, keeping data within these stores; 3) A trustless cryptographic rewards mechanism on a public blockchain to incentivize participation and ensure fair reward distribution via decentralized auditing. By eliminating centralized gatekeepers, DIN enables open access to and contribution toward AI training resources while maintaining data control. Participants benefit financially through a trustless peer-to-peer network and actively contribute to a decentralized, scalable ecosystem that harnesses their data to develop impactful AI algorithms.
This paper sets out the original design of the protocol, then called the Decentralized Intelligence Network (DIN). The project is now called InfiniteZero, and some details have changed as the protocol has been built and tested. For how the network works today, see the DevNet documentation.
- InfiniteZero on GitHub
- DevNet: the protocol's code
- InfiniteZero Protocol Proposals (IZPs): how changes are proposed and agreed
Questions about the paper: abrahamnash@protonmail.com