Hi, I got AI LLM to run as an EVM optimistic rollup using Cartesi Discord. I wanna build a prototype on Arbitrum for an ai delegate, which votes for you based on how you aligned it. I’m looking to get some ideas from an active participant (AP) in voting to narrow down an initial use case. (I’m new to the governance)
a) I’d like the AP to describe their vision or guidance on how they vote.
b) What topics or which examples of proposal indicate a simple thought process, so that LLMs can skip multi-step chains of thoughts (they are bad at that)
c) 10 strong historical examples that I can use for few-shot learning
d) how to make it practical and user-friendly (i haven’t done much dao voting, so i’m clueless about what actually happens) I would also love some feedback.
the AI delegate works like this:
- get align data by showing examples of the proposals you like and didn’t, and also writing a description of preferences
- When a proposal comes up, the proposal gets sent to the LLM
- LLM decides if it is simple enough and low-risk enough to vote on
- LLM makes a decision and describes the decision process, notifies the delegators and a day later submits the vote.
- the delegator can cancel the vote
Kirill (@kirill_igum on x)
(I have a good ai and ok crypto background and won a bunch of bounties at hackathons, i code)
would love to know your thoughts