AI DAO stuff

There’s a crippled thread where replies can’t be posted so I’m sharing here.

I don’t like this approach. I mean, it’s a clever “incentive” to those who’d vote in favor if promised a cut of this Agent’s revenue, but we could get the exact same outcome if this DAO burned xx tokens “earned” in the process.

There’s no need for such gimmicks. This (and future) proposers should remember to burn tokens if they can earn “more than they need”.

Applicants should apply for multiple grants if their proposal is fuzzy and cannot have precisely defined content and objectives.

That makes sense, as soon as the service is ready and its usefulness can be evaluated.

Musings

  • There have been some ideas/concepts to provide AI-over-cMix. The simplest use case is to proxy AI API calls via cMix, which can work fine for text, but - after file transfer support is implemented - blobs (e.g. images) as well.
    • Such use cases don’t have to use xx coin and - given that cMix is currently free - don’t need to compensate xx Network (as far as xx Network is concerned).
  • In terms of xx coin-related AI ideas, projects could assist in development of xxPostage RFP, and maybe implement parts of it related to their use case. That would be very valuable.
  • Another xx coin-related use case would be to build on xx chain, and run on cMix nodes. This is probably a “sensitive” topic because of security concerns, etc. but can be solved nicely and AI tasks could be scheduled through cMix scheduler (another “sensitive” topic, as the scheduler should be developed to work in a decentralized fashion) or separately using a cross-chain messaging protocol (also requires “touching” the chain, but if someone wanted to help integrate it into xx chain for an AI project that uses cMix, this would be great). With these approaches it would be possible to implement payments to validators who run AI jobs, and reward them similarly to the way they get rewarded for cMix points (which would save the grantee some time).
    • One obvious difference from the “10% cut” approach mentioned above is that payments could be split between those who perform the work (validators), the grantee and whoever else (cut be done in a smart contract) without any gimmicks
    • cMixx scheduler could use idle time between rounds to dispatch short-lived AI jobs “for free” (well, not really for free, it’d cost more in power and network traffic, but the right people could get paid and most validators have GPUs and/or can buy better ones since they already need them)

That’s what would work. Mere proxying of AI API calls through cMixx is a $3K job required to modify existing proxxy.xx.network code, to be brutally honest.