In a new article in Science, Professor Pamela Samuelson examines the copyright questions raised by the rapid development of generative AI, particularly through lawsuits brought by visual artists and writers.
These suits allege making copies of in-copyright works to train generative AI models is an infringement, as are outputs of these systems. Many in creative industries are worried that AI-generated images, writing, and music will undercut the need for human artists.
The current argument is a new version of a familiar story, Samuelson writes, from player pianos to VCRs and photocopiers. If generative AI seems poised to be more disruptive than previous technologies, one factor is the the rapid pace of its development. Evolution in the law and policy often lags behind, she explains and it can be difficult to calibrate how to balance competing copyright interests in the early stages of development — not to mention accounting for different laws across the globe.
In addition to the lawsuits, Congress is exploring the issue and the U.S. Copyright Office is planning to seek public comment ahead of releasing its own analysis. Scientists with an interest in generative AI would be wise to weight in, Samuelson writes.
“If the plaintiffs prevail, the only generative AI systems that may be lawful in the U.S. would be those trained on public domain works or under licenses, affecting everyone who deploys generative AI, integrates it into their products, and uses it for scientific research,” she writes.
Samuelson also gave the final talk of the UC Berkeley Distinguished Lectures on the Status and Future of AI in April addressing some of the same concerns.