Where do people go for trusted Bitcoin educational resources? In the ever-changing and evolving world of Bitcoin, false information, distrust, and confusion about complex topics is common. This makes navigating the Bitcoin landscape challenging for many, especially for newcomers.

Enter Bitcoin-PAL, a bitcoin focused AI chatbot coupled with an incentivized crowd-sourcing platform to train LLMs with bitcoin. Using Bitcoin-PAL, bitcoiners can rely on and contribute to trusted and vetted bitcoin documentation rather than query the general internet and GPT models.

📺 Video Pitch and Demo

Below is our pitch and demo for Bitcoin-PAL. You can use our replit to look at our code alongside our presentation. Here are our slides as well.

<iframe class="remirror-iframe remirror-iframe-youtube" src="https://www.youtube-nocookie.com/embed/i-LUgcZiPEA?" data-embed-type="youtube" allowfullscreen="true" frameborder="0"></iframe>

🧑‍💻 Tech Stack

1. Bitcoin

   1. bitcoind + lnd

2. LLM

   1. PrivateGPT with a GPT-J model

   2. Python = 3.10

3. Payouts

   1. Github actions and Lightning address

4. Front end

   1. React with Node 16 or higher

🔭 Future Enhancements

1. Add Filtered Criteria - AI scoring of submitted documents on relevance, source, and copyrighted information. In the beginning, the AI will reject often as it is trained on merged PR’s.

2. Remove General LLM - Due to the limitations of our LLM techstack, we had to filter out the general LLM knowledge. Future versions of PrivateGPT will hopefully make this task easier so only trained source documents are used exclusively. We may also look into llama2.

3. Add More Documents - The more accepted documentation, the better the tool. We plan to add SLIPs, more RFCs, developer bios, and additional BIPs as more bitcoin features are released.

4. Run a Hosted Version - The community can run a hosted version for community consumption and document submission. While users can always run and train their own model locally, the power of Bitcoin-PAL lies in the community and crowd-sourcing information together.

5. Enable Multithreading - Currently the backend locks execution to a single user session. We can enable multithreading to have multiple queries at once (with some resource  and computational limits).

6. Add New Frontend and Chat History- Implement the frontend that Niku designed and a persistent chat history as we were unable to complete integration by the hackathon deadline. The chat history will allow the AI to learn from historical chat sessions, questions, and answers and will help guide future conversations.

🐶 Dog-fooding AI

We used AI to help us with several facets of marketing and branding. Below are some screenshots of using ai tools like ChatGPT4 to help us with product names and logos.

📚 Resources

1. Project Charter and User Stories

2. Github

3. GitHub Actions

4. Backend Documentation

5. Project Board

6. Replit

7. Figma designs

🌎 Team

• ecurrencyhodler (Product Manager)

• Ron Stoner (Backend and AI hacker)

• Anthony Albertina (Bitcoin and backend developer)

• Niku Singh (Frontend developer)