The explosion of generative AI has raised critically important questions about ownership and fair compensation for art and other content on the Internet. Who retains the rights of art generated by an AI that was trained on millions of images belonging to artists that never gave permission for their work to be used? Today, multi-billion dollar AI companies are thrifting the creators with whose work they make enormous profit, by for example, refusing to reveal their training data.
In 2023, thousands of artists went on strike and removed their art from the popular platforms like DeviantArt and ArtStation because the platforms failed to protect their copyrighted material. There are ongoing high-impact court-cases on copyright disputes, including one involving the New York Times. While the Biden Administration drafted an AI Bill of Rights, legislation is struggling to come to an agreement in this fast changing landscape. Backpack will provide the first affordable, human-centric, and ethically based marketplace for original 3D and generative AI models – a(n ironically) disruptive yet ethical solution. Today, no digital asset marketplace has a business model like the one we propose because we are willing to share the profit in favor of the artists.
So what exactly is the solution to copyright infringement lawsuits in the age of AI?
The solution is simple: you compensate the artists for contributing to the training data. When a generative model produces work that produces a piece that is in the style of an artist or was heavily influenced by a particular artist, you compensate the artist for that generation once it gets downloaded.
Some companies have tried to do this, but they haven’t done it ethically. Take for example Adobe Stock, an Adobe platform where artists can sell their art as stock images and a platform that uses those artists to train their Firefly generative AI. In this platform, an artist can’t Opt Out from having their work used to train AI models. There was no transparency from the get-go. At Backpack we will ensure that artists are compensated for contributing to the training data. Furthermore, at platforms like Adobe Stock, a consumer can use a reference image to generate a new one (e.g., image-to-image algorithms) – but this means that the produced artwork could in theory infringe the copyright of the original reference image. This can quickly become a legal liability issue for all involved parties.
Why do you think OpenAI, Facebook/Meta, and Google are striking multi-million dollar licensing deals with platforms like Shutterstock ever since they started getting sued? Because they weren’t creating ethical AI. Even the OpenAI CTO struggled at answering questions of how they sourced their training data (see video below).
Join Backpack and join the movement for ethical generative AI.
