âī¸How it works
Last updated
Last updated
This section describes, at a high-level, how the Fair AI Data platform addresses the current challenges of AI data capture practices. We're building a fair, trustless system where Data Contributors (such as artists or coders) get compensated and accredited for their work, while Data Buyers (AI companies or developers) have access to high-quality and ethically-sourced data to train their AI models.
Creators across various domains face the challenge of having their work used in AI training without any attribution or compensation. With Fair AI Data:
Creators that wish to monetize their work for AI training can contribute data (such as art or code) to the platform.
The platform validates ownership of, structures, and stores the data to form specialized datasets (or data pools), licenced specifically for AI training. The decentralized storage layer ensures the immutability and "ownership traceability" of the data.
Creators in turn are issued FAIR tokens equivalent to their data contribution, expressing their immutable share in the dataset.
When AI companies purchase a dataset from the platform, the revenue is distributed back to the creators in the form of FAIR tokens, split fairly based on their initial data contributions.
AI companies and developers on the other hand, have trouble finding high-quality, ethically-sourced and diverse datasets while making sure they don't violate creator's data rights. Through Fair AI Data:
AI companies can browse a rich marketplace of ethically-sourced and curated datasets to train their models
Depending on their needs, they can select between various licensing options such as complete buyout of the dataset with license to use for training, or even instrumental pay-as-you-go training.
Whenever a dataset is bought or used, the amount paid is automatically split fairly between the original data contributors whose work is powering these datasets.
We will release more in-depth information of the system and its various components soon, with the release of our whitepaper.
