Challenges
Best Use of Polygon
Polygon chain and it's availability on opensea helped in development of the tooling. It is a fast blockchain which gave the proect real-world transaction scenarios.
Build on Base 🔵
3 smart contracts have been deployed on Base Goerli testnet. While development, we have interacted with coinbase wallet, basescan, faucets. We chose Base to be able to reduce on gas fees and develop on an Ethereum L2 chain. Apart from choosing the chain, we have integrated the best UX possible for a user. With passwordless login and gasless transactions the friction for user to use the app is close to negligible. For doing that we have explored how many integrations Base has with matured tooling.
Integrate Web3Auth Core-Kit SDK in your Ethereum Application
The login is based on Web3Auth, this was done to ensure that user faces negligible friction while using the app. In face, we took it forward and added Account abstraction and gasless wallet as well to improve upon the UX 10x.
#BUIDLathon 2023 NFTs + Metaverse + Gaming Track
In the project the results of the ML models and the models themselves become tradeale NFTs. There is an incentive for model developer to develop better models as the demand to run that will be high, which is good for the ecosystem.
Project details
Find it here https://docs.google.com/presentation/d/1XsIvxb-WAKP8LWtNok9UpNpCEubQQbNlCGja456dMrs/edit?usp=sharing
Abstract: Users earn coins by rating images produced by AI models. These coins are then used by user to generate more inferences. Nodes give inferences in exchange of coins and the ratings recieved is on-chain.
Motivation - Access, availability and incentive alignment
Anonymity: AI Inferences is hard to get for layman anonymously - gatekeepers of models require you login to their systems
Difficulty: AI Inferences is hard to get for layman easily - gatekeepers models are closed source/ open source models require you to interact with code
Open dashboards at capacity: Open AI Inference dashboards run out of capacity - eg. ChatGPT being at capacity most of the time. Paid options require intensive data sharing.
Developer incentive alignment: Independent AI Inference model developers have no incentive to host their nodes and pay for compute cost
Our solution
Best UX, built on Base Goerli testnet, Account Abstraction, Passwordless, Gasless Wallet, Open source
User (wants an image from a prompt)
Logs in with Gelato Network Wallet (powered by Web3Auth)
Passwordless and gasless experience
Earns DecentAI Coin (DAIC) by rating past images
Enters prompt and selects node to ask for inference at the cost of DAIC
Node (earns DAIC by fulfilling requests)
Clones the Github DecentAI Node Repo
Simple CLI commands to set up the node
Produces output for seed images for user to evaluate
Earns DAIC by fulfilling AI requests
Gets rated by users building up to a reputation score
Contracts deployed (Base Goerli Testnet)
DecentAI (ERC721) - 0x9d5CD448332A857F6BfDb7541CFc33C61789BB41
registerResponder - For registering node
requestInference
recieveInference
Output becomes an NFT minted to requestor
DecentAI Coin (ERC20) - 0x26b29286A6D8cE5d0BCD40121ACB8BdF120F08f4
Incentivizes useful participation for users and nodes
Users earn by rating images
Nodes earn by creating inferences
DecentAI Node (ERC721)- 0x2adC5C0bFc6C3084f5dEaF8BD7C14132EC444c6c
Each registerResponder call creates a NFT which can be traded
Let’s model developers trade model code, which sits in IPFS
Repositories
Decent AI - Smart contracts and User facing UI
Repo: https://github.com/shashank42/DecentAI
UI: https://shashank42.github.io/DecentAI/
Decent AI Node
Repo: https://github.com/shashank42/DecentAINode
Colab notebook to use the code - https://colab.research.google.com/drive/10k5QfP0eG8ivHL2XFmBxs3BKcKsqNfTl?usp=sharing