eThAi

The AI space today is highly fragmented, with major players like OpenAI, NVIDIA, Google, and Microsoft competing to build better AI models or achieve AGI (Artificial General Intelligence). Each develops their own models, often in silos, which creates barriers to collaboration. Additionally, training AI models requires substantial GPU resources, and even NVIDIA's GPUs are limited in availability and affordability.

Now imagine a world where everyone like individuals, organizations, and companies—collaboratively trains AI models or LLMs (Large Language Models). Instead of isolated efforts, platforms like Google, Microsoft, and OpenAI could pool their efforts into one shared central model. This collaborative approach could dramatically accelerate progress in AI development.

This is the vision behind our platform. It enables users across the globe to contribute their computing power and efforts toward training AI models. The platform aggregates these contributions, creating a strong, centralized model managed via smart contracts.

The platform operates with two approaches:

1. Open-Source Models: A community-driven initiative where DAO (Decentralized Autonomous Organization) participants propose, vote, and decide which models to train next. DAO tokens are earned by contributing to the platform and can be used to influence decisions such as including or excluding specific models.

2. Closed-Source Models: Clients can list AI models they want to train, along with datasets they provide. They leverage the platform’s distributed training capabilities, allowing users to help train their models. Users are rewarded with DAO coins or monetary compensation for their contributions.

Our project leverages cutting-edge technologies and innovative integrations to build a collaborative AI training platform. Here’s a breakdown of the core components and how they come together:

1. DAO Framework - Scroll: We use the Scroll blockchain to manage the DAO (Decentralized Autonomous Organization), enabling community governance for the open-source AI models. This ensures transparent decision-making and voting processes using DAO tokens.

2. Dataset Integrity - Sign Protocol: To maintain trust, we implemented Sign Protocol to verify that datasets remain untampered throughout the training process. This ensures data integrity and security, which is crucial for reliable AI model training.

3. Collaborative Communication - Push Protocol Chat: We integrated Push Protocol to facilitate discussions and collaborations among contributors. This creates a seamless communication channel for users to share insights, provide feedback, and work together.

4. Transaction Transparency - Blockscout: All transactions on the platform are recorded on Blockscout, providing full transparency and traceability. This ensures accountability for contributions, rewards, and model governance actions.

5. Wallet Connectivity - Coinbase Developer Platform: For wallet integration, we utilize Coinbase's developer platform to handle connection, checkouts, and funding (OCK) processes. This enables a user-friendly experience for managing contributions and rewards.

6. Model Naming System - ENS (Ethereum Name Service): AI models are named using ENS, allowing easy identification and categorization. This naming system provides a user-friendly way to track and refer to models across the platform.

7. Dataset Management - Pyth Network: To handle dataset distribution, we use the Pyth Network to randomly split datasets among users. This ensures fair and efficient allocation while maintaining the confidentiality of sensitive data.

8. Secure Storage - Nillion: Nillion serves as our solution for decentralized storage of datasets. This ensures data is securely stored and readily accessible for training processes.

9. Cross-Chain Integration - LayerZero: To make the platform cross-chain compatible, we leverage LayerZero for interoperability. This allows trained AI models and related data to be accessible across different blockchain ecosystems.