Autonomous Artisit

Autonomous artist generates one artwork per day with their capabilities of synthesizing knowledge graphs and entity extractions based on aesthetic value, symbolic importance, and popular opinion. They take the news feed and create the artwork through Stable Diffusion, and sell it on various decentralized NFT marketplaces.

The profits are distributed as follows after deducting expenses for compute, storage, and APIs: • 10% to people who submitted the prompts (the more prompts are accepted, the larger their share) • 10% to people who assessed the artworks (the more their assessments led to successful sales, the larger their share) • 20% will be allocated to Gitcoin grants to support autonomous artists • 20% will be used to update the artist's code through hiring developers • 20% will simply accumulate as its savings and will never be spent

Tokenization layer is built on Gnosis, and humans interact with the artist via the DAO. People can propose and vote on how the artist creates art and therefore decentralize AI model training in a collective intelligence way.

We used an api call to scrape news headlines. These are then routed through a Heroku app to query an LLM, which provides a well formed prompt. This prompt is then sent to a stable diffusion instance, which returns an image. The returned image is then used to create an ERC-721 token with the image in the Autonomous Artist's smart contract.

We store our the data in Gnosis and add a tokenization layer to empower people to collectively train the AI model as well as ensure the artist grows in a decentralized way.