FUGA

FUGA, the Federated User-Generated Avatar (AI), is a groundbreaking solution powered by Blockchain and Federated Learning that is set to redefine the realm of Artificial Intelligence (AI). By leveraging the vast amounts of data generated by everyday users, AI has grown increasingly intelligent. Nonetheless, these contributors often lack control over the AI models' policies and receive no recognition for their invaluable data contributions. FUGA addresses this disparity by incorporating Federated Learning and Blockchain technologies, fostering a sustainable AI training ecosystem that enables individuals to participate directly in the AI model training process, obtain rewards for their data contributions, and even influence the model's policies.

Federated Learning is a decentralized approach to AI model training that allows multiple clients to collaborate on model development while preserving data privacy. By keeping data on local devices, Federated Learning enables models to learn from shared insights without exposing the raw data, thereby enhancing privacy and security.

At its core, FUGA aims to create AI chatbots that capture and represent the unique attributes of specific communities. Through efficient filtering and processing of vast data, robust privacy protection, and assurance of rightful ownership and fair compensation, FUGA lays the foundation for a more equitable and thriving Decentralized AI-driven future.

We utilized a decentralized federated learning structure for AI training. Clients (nodes) exchange trained models with each other, aggregate and evaluate them. By repeating this process, overall performance improvement is achieved, and models optimized for individual data are produced. We also deployed a contract on the Polygon testnet, which allows a smart contract to act as the controller without the need for a centralized aggregator. Furthermore, trained models are converted into NFTs and rewarded based on the degree of contribution through the smart contract. Consequently, data providers can maintain their data privacy while receiving rightful ownership and fair compensation for their data.