[MERGED] Custom GPT for OP Collective Governance

Custom GPT for OP Collective Governance

Due to similarities, have decided to delete this proposal in favor of a merge with [DRAFT] AI Assistant for governance support Mission Request.

Click on the pencil in the top right to see the original proposal.

6 Likes

This would be awesome!!

1 Like

I feel like there should definitely be more resources for people who are new to OP governance but i don’t think GPT is the right solution. I think it might be money better spent to give grants to new delegates to document their journey.

I also think the process for training a custom GPT is relatively trivial - $100k budget seems excessive for what could be done in a few days. There is a thread on the OpenAI forum that shows how to do this in less than 10 minutes for $1 or less of compute costs

6 Likes

nice feedback. i generally like the idea of training a critical eye on a lot of these requests to see whether there are more efficient, minimally viable ways to do things

2 Likes

This is a great idea.

But I don’t believe it needs 28,000 OP to be arranged.

From personal experience, as someone who has built a Custom GPT for an organisation, and continues to add to it on a weekly basis, there is very little cost involved. Only time is involved. Even then, the skill level involved is not that high (thanks to the work of OpenAI), so the time dedicated wouldn’t have to come from a high cost individual.

I believe that a project like this is best suited to RetroPGF, and not a mission (if a Custom GPT can be classified as a public good?).

I believe that 28,000 OP is a figure way too high for this project and possibly represents poor stewardship of OP reserves. As a cheeky example: send me 2000 OP and I’ll spin up a Custom GPT for OP Governance in a week, and commit to training it for an hour every weekend throughout 2024.

But for the record, it’s a good idea.
And I encourage every community member here to take the initiative, build this, continue to refine it, and then apply for RetroPGF.

9 Likes

nice @Axel_T , now we’re talking

2 Likes

I agree with the other replies: great idea but the budget seems excessive.

Inspired by @Axel_T reply, I will take the initiative and try to build this for free.

4 Likes

I totally agree, I wonder if there is a way to enforce this going forward. A third party just to sense check certain budgets for more specific requests (in this case AI). Could help save a lot of tokens

1 Like

This is part of what the Dev advisory board was for but I also think it’s as simple as people just taking a hard look at these mission requests one-by-one. It doesn’t take long after grants are issued for people to flag them—it’s just that it’s then when they’re paying attention and not before.

3 Likes

Hi, I think this proposal is similar to this one?

6 Likes

Is it possible to accomplish this without requesting Foundation action or support like providing a forum API key?

I don’t want to be the bad guy but if not this could be violating a rule. I want this to happen just asking about the implementation process.

2 Likes

Definitely agree with all the feedback about the amount being too high.

I had considered most of the work being in the maintaining of up-to-date information (not the GPT). I wasn’t aware of the ‘request archive’ thing so I was thinking someone might have to build a custom scraper, because I thought I had seen conversations before about it the forums not having an API available.

Although it seems like the archive is of a users personal history, and not the entire forums, right?

Either way, adjusting the amount down to 8k OP for now and will continue evaluating.

Thanks for the feedback
@Axel_T @jackanorak @0xOnChainChaos @0xDonPepe

2 Likes

You can access the forums publicly, so always possible to build a scraper :man_shrugging:

2 Likes

Agree, @Michael have you considered collaborating or combining forces with the other AI Assistant Mission Requests?

3 Likes

For our side, happy to potentially work on merging both proposals

3 Likes

If participants/foundations are comfortable using their dialogs/discussions as training data for the closed-source OpenAI, I’m all for it; if not, I think an open-sourced LLMs could be a better alternative.

3 Likes

Amazing idea. Agree with feedback re: smaller scope than op had maybe anticipated.

Would love to use a tool like this!

2 Likes

Send you a DM! Also happy to merge.

Originally I had read yours but it had some other components, now I think they are much more similar.

1 Like

Will be watching this with interest. Have been playing with creating custom GPTs by uploading Gitbook governance docs as knowledge and then fine tuning the prompts. Mixed results for me so far but the potential is definitely there.

I was doing this to create something similar for Developer DAO but have limited time to execute it properly. If this works well would love to be able to deploy to our forums with a different knowlege set. Are you planning to make this open source?

1 Like