Vector3: natural language interface for AI x Web3

Decentralized, Open-sourced LLMs Pipeline for All of Web3

Abstract

Vector3 is a decentralized AI infrastructure designed to build and maintain open-source language models, specialized in querying and answering RAG workloads (Retrieval Augmented Generations) for all Web3-related questions in natural language. By leveraging a network of competitive nodes and a continuous feedback loop of real human and AI agent queries, Vector3 aims to become the unified natural language interface for Web3. The platform integrates diverse data retrieval providers and incentivizes participants to contribute to the growth and accuracy of the models. Through decentralized governance and a lifecycle of data generation, validation, and model training, Vector3 seeks to provide the best AI models, best public datasets, and the best RAG pipelines for the continuously evolving landscape of Web3.

Problem

The evolving landscape of Web3 presents unique challenges for data accessibility and interaction:

• Lack of Open-Source RAG Models: The absence of Retrieval-Augmented Generation AI models tailored to Web3 makes it difficult to query Web3 information comprehensively and accurately.
• Integration Challenges: Diverse data sources and interaction points lead to fragmented AI solutions that cannot offer a unified querying experience.
• Rapid Evolution of Web3: The continuous growth of new technologies, diverse data sources, and evolving user demands within Web3 requires agile AI models that can adapt swiftly to new data and constantly shifting user query needs.
• Limited Incentives for Dataset Creation: High-quality, large-scale datasets with real human interactions are scarce, primarily because of insufficient incentives for companies to share or build their data openly.
• Scalability Issues in Data Labeling: Effective AI models require extensive supervised datasets, but creating and scaling labeled data is complex and costly.

Solution

Vector3 is designed to solve these issues through decentralized collaboration and continuous improvement:

• Decentralized LLMs for Web3: Develop and maintain Large Language Models (LLMs) that can answer any Web3-related natural language query.
• Unified Natural Language Interface: Acts as a backend for both AI agents and human users, providing a consistent and comprehensive layer for handling all Web3 queries.