Can an artist sue an AI over copyright infringement?
All We'd Ever Need is One Another is a project by Montreal-based artist Adam Basanta. It features colourful, abstract images, which are posted on Basanta's website, and his Twitter and Instagram Accounts.
But Basanta doesn't create the images himself. Scanners controlled by a computer create random images. "So basically what's happening," Basanta said, speaking on Spark, "is it's this ongoing art factory."
A machine learning program then checks those images against a database of art. If any of the images match any existing art, the system saves them and posts them online.
Basanta is now being sued over one of the pieces produced by the system. Artist, and owner of Montreal's Galerie NuEdge, Amel Chamandy says that the piece, called 85.81%_match: Amel Chamandy "Your World Without Paper", 2009, violates the copyright she holds for her painting, Your World Without Paper, as well as the trademark for her name.
In a statement to Spark, Chamandry wrote, in part, "if art is not protected on the digital market, then it wouldn't it be any different than someone working off brand names such as Apple or Nike. Creative art as well as the name of the creative artist must have the same protection as in the real world or else there would be no brand."
Jeremy de Beer is a law professor at U of Ottawa's Centre for Law, Technology, and Society. He points out that Basanta, the defendant in the lawsuit, has left himself in a tricky position simply in virtue of the way the art project was conceived.
"It doesn't actually move these [AI-created] works forward in the process unless it verifies that it's similar enough to an existing work to be valid," de Beer told Spark host, Nora Young. "And so they've put themselves in somewhat of a Catch-22 from a legal liability point of view, by essentially admitting the question of similarity."
For de Beer, though, issues about copyright and intellectual property law go far beyond this one case, as we move into an era of more advanced AI.
One of those issues is simply how to decide authorship as AI systems get better at producing creative works on their own.
"There are a couple of different options from an ownership point of view. One is to grant ownership to the machine itself. Another is to give copyright to the programmer, and yet another one is to give copyright to the program user. Or we could put copyright into the public domain," de Beer said.
Deciding how to assign copyright is shaped by the reasons society grants copyright at all. "The rationale for giving copyright protection is based either on the need for an incentive...or the reflection of an author's personality in a work. Neither of those justifications hold up when we're talking about machines," said de Beer.
Beyond the question of what kind of rights -- if any -- an AI program might be granted in creative works, we may well need to reform Canadian intellectual property law to allow Canadian innovation in AI.
Without law reform in a Canadian context, we're putting our innovators at a real competitive disadvantage - Jeremy de Beer
"To get involved in AI, you need to do massive amounts of text and data mining to develop the algorithms that make AI work," de Beer said. "In some countries, the copyright system is flexible enough to permit that...It's highly uncertain whether AI companies in Canada can do the text and data mining they need to do to compete in this space. Without law reform in a Canadian context, we're putting our innovators at a real competitive disadvantage."