Inspiration

AI is reshaping industries, but its training process is often centralized, limiting diversity and fairness. We envisioned a decentralized platform where anyone can contribute to AI training, ensuring inclusivity, transparency, and fair rewards for participants. By combining blockchain for trust and AI for innovation, we aim to democratize AI development while compensating contributors for their efforts.

What it does

AI Forge is a decentralized application (dApp) that allows users to:

1. Train large language models (LLMs) and domain-specific Selective Language Models (SLMs) by contributing data, providing feedback, and refining responses.
   
2. Earn tokens for their contributions, which can be traded, staked, or redeemed for premium features.
   
3. Participate in community governance to guide model improvements and priorities.
   

How we built it

1. Frontend: Built with React.js and integrated with Web3 libraries for wallet connectivity.
   
2. Backend:
   
   • AI: Hugging Face Transformers for LLMs and fine-tuning SLMs.
     
   • Blockchain: Smart contracts on Ethereum manage rewards and governance.
     
   • Storage: IPFS for decentralized data storage and zk-SNARKs for privacy.
     
3. Data Sources:
   
   • Open datasets such as Common Crawl, Wikipedia, and OpenSubtitles for initial model training.
     
   • User-contributed datasets, validated via consensus algorithms and privacy-preserving techniques.
     
   • Domain-specific datasets sourced through collaborations with industry partners.
     
4. Architecture: Combines federated learning for privacy-preserving AI training with a robust token economy powered by blockchain.
   

Challenges we ran into

1. Ensuring data quality from diverse contributors while preventing malicious inputs.
   
2. Balancing privacy with transparency using technologies like differential privacy and zk-SNARKs.
   
3. Creating a seamless user experience that integrates AI interaction with blockchain mechanics.
   

Accomplishments that we're proud of

• Successfully implemented a reward system for decentralized AI contributions.
  
• Fine-tuned an SLM to excel in a specific domain based on user feedback.
  
• Built a secure, user-friendly platform that democratizes AI training.
  

What we learned

• The importance of incentivizing quality over quantity in user contributions.
  
• Strategies to integrate blockchain and AI without compromising performance or security.
  
• The value of community feedback in shaping AI models and governance.
  

What's next for AI Forge

1. Expanding domain-specific SLMs to cater to industries like healthcare, legal, and finance.
   
2. Enhancing reward mechanisms to include staking and dynamic tokenomics.
   
3. Building partnerships with organizations for real-world adoption.
   
4. Scaling the platform using decentralized compute networks to handle larger datasets and contributors.
   
5. Introducing multilingual support to broaden accessibility and inclusivity.
   

AI Forge aims to redefine AI training by putting power back into the hands of the people while ensuring everyone benefits from the AI revolution.

PPT :- https://drive.google.com/file/d/1uZu3w0vyqAwt08FR1ifi_JQkBKrbyzvV/view?usp=sharing

Built With

• ethereum
• federated-learning
• hugging-face-transformers
• ipfs
• react.js
• selective-language-models-(slms)
• smart-contracts
• web3.js
• zk-snarks