Spawning wants to build more ethical AI training data sets
How does Source.Plus aim to address the issues of rights management and fair use in AI model training?
Source.Plus is a new project by Spawning AI that aims to address the issues of rights management and fair use in AI model training by providing a platform for artists and rights holders to offer their media for use in generative AI training on their own terms. The platform's first initiative is a dataset seeded with nearly 40 million public domain images and images under the Creative Commons' CC0 license, which allows creators to waive nearly all legal interest in their works. The platform builds on Spawning's existing tools for art provenance and usage rights management, such as HaveIBeenTrained, a website that allows creators to opt out of the training data sets used by vendors who've partnered with Spawning. With Source.Plus, Spawning aims to set a higher bar for what it means to fairly source data, by filtering images for "opt-outs" and other artist training preferences, and excluding images that aren't licensed under CC0. The platform also employs moderators to verify data ownership and has remediation features built in, where users can flag offending or possible infringing works, and the trail of how that data was consumed can be audited.
What is the primary goal of Spawning AI's project, Source.Plus, as founded by Jordan Meyer and Mathew Dryhurst?
The primary goal of Spawning AI's project, Source.Plus, as founded by Jordan Meyer and Mathew Dryhurst, is to create a platform that curates "non-infringing" media for AI model training. It aims to provide a universal "opt-in" platform where rights holders can easily offer their media for use in generative AI training on their own terms, making it frictionless for developers to incorporate that media into their training workflows. The initial focus is on images, with plans to expand to other types of media, such as audio and video.
Can you explain the role of the tools and features that Source.Plus offers to support the AI training pipeline?
Source.Plus, a project by Spawning AI, offers several tools and features to support the AI training pipeline while ensuring ethical and responsible use of artists' works in AI model training.
-
High-quality, non-infringing data set: Source.Plus provides a data set of nearly 40 million public domain images and images under the Creative Commons’ CC0 license, ensuring a high-quality and non-infringing foundation for AI model training.
-
Opt-in platform: The platform allows rights holders to offer their media for use in generative AI training on their own terms, making it easy for them to participate in the AI economy and receive fair compensation.
-
Rights management: Source.Plus respects data rights by filtering images for "opt-outs" and other artist training preferences, showing provenance information about how and from where images were sourced, and excluding images not licensed under CC0.
-
Image moderation: Spawning employs classifier models trained to detect nudity, gore, personally identifiable information, and other undesirable content in images, ensuring a clean data set for AI model training.
-
Flexible filtering: Users can adjust the classifiers' detection thresholds to suit their needs, providing a flexible way to filter the Source.Plus data set according to specific requirements.
-
Artist compensation: Once Source.Plus exits beta, it will allow artists and rights holders to set their own prices per download, giving them control over the terms of their participation and ensuring fair compensation.
-
Revenue transparency: Spawning will charge a flat rate fee for using the platform and let artists keep the majority of the revenue, providing greater transparency and more favorable revenue splits compared to other platforms.
By offering these tools and features, Source.Plus aims to create a responsible and ethical ecosystem for AI model training, giving artists control over their works and ensuring fair compensation for their contributions.