Gensyn Launches Testnet, Integrates Blockchain for Decentralized AI

Ben Fielding, a young AI researcher, began his PhD track at Northumbria University in northern England in 2015. His workspace was a noisy desk housing a large machine stuffed with early GPUs, which he used to develop AI. The machine was so loud that it annoyed his lab-mates, leading him to cram it beneath his desk. Despite the inconvenience, Fielding had innovative ideas about how "swarms" of AI—clusters of many different models—could communicate and learn from each other, potentially improving the collective whole. However, he was limited by the computational constraints of his noisy machine and the superior resources of tech giants like Google.

Fielding realized that compute constraints would always be an issue, especially as AI moved towards mainstream adoption. This realization led him to the concept of decentralized AI. In 2020, he co-founded Gensyn with Harry Grieve, initially known for building decentralized compute. The project aimed to create "the network for machine intelligence," developing solutions across the tech stack. Gensyn recently released its "RL Swarms" protocol, a descendant of Fielding’s PhD work, and launched its Testnet, integrating blockchain technology.

In an interview leading up to the AI Summit at Consensus in Toronto, Fielding explained the RL Swarms protocol. This protocol allows pre-trained models to undergo reasoning training, where they learn to critique their own thinking and improve recursively. Fielding extended this concept by enabling models to communicate with each other, forming a swarm that collectively improves the outcome of a final meta-model. This collaborative training process allows any model to join and benefit from the swarm's knowledge, improving both the individual model and the swarm as a whole.

Fielding also discussed the integration of blockchain technology into Gensyn's infrastructure. Blockchain provides lower-level primitives, such as persistent identity, payments, and consensus mechanisms, which are essential for trust and verification in a decentralized AI network. The Testnet launch marks the beginning of this integration, where models joining a swarm have a persistent identity on a decentralized ledger. This ensures trust and consensus, allowing for the termination of disputes programmatically.

Looking ahead, Fielding envisions a future where all resources under machine learning are instantaneously and programmatically accessible to everyone. This would democratize AI development, breaking the moat created by centralized AI companies. Gensyn aims to build the low-level infrastructure necessary to achieve this vision, making machine learning technologies accessible to all innovators, not just tech giants. The ultimate goal is to create an open-source ecosystem where everyone has the right to build and improve machine learning technologies.