Using AI and ZK for DAOs

This article is in response to a Smart Contract Research Forum podcast on Culture and Incentivization. Link is here: https://www.smartcontractresearch.org/t/scrf-interviews-culture-and-incentivization-ellie-rennie-and-zach-anderson-ep-3/1249?u=lndavis6

Thank you, Ellie and Zach, for such an insightful episode. I've interacted with SourceCred and Coordinape and think both platforms point decentralized organizations in the right direction. It was interesting to hear Zach's first-hand experiences with compensation and equitable pay in DAOs and the questions that arise from them. I'm excited about DAOs addressing gender and racial pay gaps; merit-based pay hasn't been done well yet.

Another point that I agreed with was Ellie's comment about what you take away from jobs other than LinkedIn connections. If you worked well with someone, you can get a recommendation from them on your profile. I feel this is LinkedIn's effort to create more value from moving on from a workplace. Though, in my opinion, the best thing you can take away from a project is an addition to your portfolio. This is optimal for open-source projects where you can show your code/contributions to anyone. Suboptimal for projects with NDAs. Portfolio displays are possible on LinkedIn, but far more tangible on a platform like GitHub or Medium/Mirror. Recommendations and portfolios are strong incentives to boost collaboration. Another one is reputation.

One of my favorite web3 projects is Unirep. Unirep uses zero-knowledge cryptography to enable a private reputation-based social network. Users can privately give others a reputation score and verify that their reputation is above a certain level. For example, if I see a post on Unirep that is very helpful, I may use my high reputation score to give reputation points to another user. The important part is that they won't know who did it. This can be extremely helpful in settings where privacy is respected. I think privacy is a great way to encourage participation because it gives people a sense of trust and autonomy.

To answer the question that ended the post, I don't believe complete pay transparency will help DAOs or encourage participation. However, I do believe that giving people free will over their information will help tremendously. By "free will," I mean giving people the opportunity to be completely anonymous, doxxed and open (with pay transparency and everything else), or anything in between. If I am a junior smart contract developer and want to know how much a smart contract developer makes in the DAO, I should be able to ask to view someone's salary/pay history. If they approve, then I can view it. Public salaries are good for verifying pay equity, but it is still data that many people would prefer to be private to the greater population.

The majority of the great work in DAOs will not come from extrinsic, but intrinsic motivation. This means a DAOs' greatest responsibility is clearly defining what needs to be done and how much someone can earn from doing said task. This is easier said than done, which is why I propose an idea to get help from artificial intelligence. Trent McConaghy has already written about AI DAOs, but his papers were more about AI-only DAOs. I am proposing an AI that acts as a universal project manager and doesn't have a consensus vote.

We've seen the capabilities of models like ChatGPT and stable diffusion, but what if an AI could learn to be a project manager? Users can define goals with a roadmap and definite executables, propose them for the DAO's approval, and the AI can handle the assignments, reassignments, bounty payouts, and verification of work. There would need to be lots of training data (perhaps extracted from Scrum / Kanban workflows), but over time an AI like this could end up managing the DAO to a level of complete privacy and autonomy. An employee of the DAO only knows what they do and the people they work directly with. Other parts of the work are abstracted until one sees the final product of whatever the DAO is doing. Here's an example:

The same junior smart contract developer above gets assigned a role to build an ERC-1155 token contract for an upcoming DeFi feature with NFT positions. The smart contract is finished within the designated time, and after the AI reviews the code, it is automatically put through Slither and sent to auditors & reviewers. Any feedback goes back to the junior developer. A couple of months later, a social media post (by the marketing team), write-up (by the writing team), tutorial (by the design team), and bug bounty (sponsored by the developer team) are all posted on the DAO's social media and blog pages. The junior developer had no idea what these tasks were or who did them, but when looking at the full project on Github, they saw their code with the appropriate @author tag giving them credit for their work. The developer only knew the other people working on the smart contracts. The different aspects of the project were handled separately because everyone could follow the project specifications with the AI.

Just a moonshot idea, but I get the feeling something like this would insanely boost productivity. The key to getting people to do more stuff is to allow them to do less and focus on one thing at a time. AI managers and zk worker identities can help with this; it's just a matter of time until they exist.