Summary

Artificial Intelligence (AI) generated images are in need of provenance, especially when they are intended for commercial use. In some cases, these images should include attribution to the prompt used to generate the art and any influences from other artists that had a significant impact on the algorithm's response.

This white paper aims to demonstrate an approach that promotes the use of blockchain technology as an immutable source of provenance, attribution, and royalty payment for AI generated creative works.

The goal is to demonstrate that the best approach for artists whose copyright protected works are being used to train AI models is to volunteer their catalog for training in exchange for compensation and attribution. We will examine how this could first be deployed in the context of generative artwork created specifically for digital sales and NFTs. Then, we will explore how the blockchain-based system could be expanded to all AI generated creative works through the use of a digital image identification system such as the widely industry-backed Content Authenticity Initiative (CAI).

This approach assumes the creation of a new AI image diffusion model with the capability for creators to opt-in and add their artwork to training. This model could report the weight of influence from each artist's work, and write its creations to a blockchain along with metadata about the prompt used, artists' influence weight, and, in the event of any generated revenue from a sale, execute a smart contract to pay a royalty to the appropriate artists.

Blockchain-AI-Model-Paper

Rewarding Artists For Their Contribution in Training AI: A Blockchain-based Solution for Secure and Automated Provenance, Attribution, and Payment for Generative Works

Introduction

• Overview of the need for provenance, attribution, and artist compensation in AI generated images
• Introduction to the proposed solution: using blockchain technology and smart contracts to automate and secure payment for AI-trained images

Background

• Explanation of the challenges of attributing and compensating artists whose works are used in the training of AI models
• Review of newly launched initiatives from companies like Shutterstock to help compensate artists
• A look at pending legal action that could threaten the adoption of AI

Provenance and Attribution via the Blockchain

• Explanation of how the blockchain can be used as an immutable source of provenance and attribution for AI generated images
• Discussion of how artists can volunteer their catalog for training in exchange for compensation and attribution

Use Case: Generative Art for Digital Sales

• Explanation of how the proposed solution could be deployed in the context of generative artwork built for digital sales and NFTs

Expansion to all AI Generated Creative Works

• Discussion of how the blockchain based system could be expanded to all AI generated creative works through the use of a digital image identification system like CAI

Proposed AI Image Diffusion Model

• Explanation of the new AI image diffusion model that would need to be created to implement the proposed solution
• Discussion of the features and capabilities of the model, including the ability for creators to opt-in and add their artwork to training, reporting of influence weight from each artist's work, and execution of smart contracts for royalty payment.

Conclusion

• Summary of the key points and a call to action for further research and development in this field
• Discussion of how the proposed solution has the potential to revolutionize the way that artists are attributed and compensated for their contributions in the AI-training process.