Created on 8th September 2024
•
Inactive and not-driven communities are a huge problem in the crypto space. Most of the votes on DAO are made by the largest holders who do not engage in discussions. Attempts to solve this problem by rewording the most active users simply fail when bots flood the forums and Discord servers with spam. Another issue arises when a small number of people manually pick the rewarded contributors. This way of selection may be deceptive to bias and corruption.
Our solution comes in handy to solve all of these problems by using AI and specifically verifiable Large Language Models.
Doing it this way gives users a verifiable and fair judgment to pick and reward the most active contributors. LLMs are good at analyzing text so they can pick not only the most active users but also select the ones that have given the most valuable insights.
The largest challenge by far was the limitations implied by the current ecosystem. Available oracle contracts didn't have sufficient functionalities to make a custom web request. That could potentially save us a huge amount of gas, where instead of saving the whole data on-chain we could save just the hash and later reference it while making a request to the AI model. We build our project in with this architecture in mind as a proof-of-concept. Because of the current limits, our program execution was constrained to what is currently available.
Tracks Applied (5)
Ethereum Name Service
Ora
DBForest
Mantle
Mantle
Technologies used