This article was originally published by The Lawyer’s Daily, part of LexisNexis Canada Inc.
How courts deal with the concept of authorship is an area to watch as artificial intelligence (AI) becomes more creative and humanlike, noted panellists at Bracing for Impact: The Artificial Intelligence Challenge Part II conference series hosted by IP Osgoode.
Lawyers and academics pointed to examples of creative works made by humans, machines and even a monkey during the conference’s panel, Intellectual Property at a Crossroads, as they discussed the impact AI would have on patent and copyright laws.
Shlomit Yanisky-Ravid, a faculty member at Ono Academic College in Israel, and Fordham Law School in New York City, kicked off the March 21 discussion by playing clips of music, asking which score was created by a human or AI. It was anyone’s guess as the music sounded eerily similar. She deployed the same test while displaying paintings, asking the conference to tell her which works of art were made by human hands.
“It’s very difficult to distinguish between an AI system and human,” she said, pointing to the humanlike capabilities of AI.
“AI systems can be creative, autonomous, unpredictable, rational,” she explained. “That’s why I think it has free choice. It can communicate with the Internet and get data even without the engineer knowing that this happened.”
Yanisky-Ravid questioned whether an AI system should own its creations, adding that giving ownership of the work to the AI’s programmer would be like saying whoever invented the camera should be the owner of a photo.
“My suggestion is taking the ‘AI for hire’ doctrine and seeing the AI system as an agent,” she added, noting that this way of thinking better reflects the understanding that AI is creative.
Dave Green, assistant general counsel of IP law and policy at Microsoft, said AI, in some instances, can create useful and expressive work equivalent to that produced by humans, which is why, from a policy perspective, he asked what behaviours copyright law is trying to incentivize.
“The challenge with existing copyright law, obviously now, is it’s really developed around this concept of personhood. In the United States Copyright Act and the (U.S.) Constitution, it’s very clear that there’s a requirement that a work be produced by humans. And I guess the question that I’d ask is, does it matter? And why does it matter?” he said.
From a policy perspective, Green suggested, that the fundamental question should be, “do we want to insist upon the requirement of humans and personhood?” If the answer is no, then how will that impact data stream issues typically associated with copyright, infringement, reproduction and liability?
Green noted producing a creative work involves a lot of steps, not all of which are protected by copyright. However, when they are protected, the law requires a human to be associated with that contribution to the work.
“When you’re dealing with artificial intelligence it’s not simply about a creative contributions and to the extent that copyright law is incapable of contributing protection for those expressive elements; there’s always patent law that can potentially step in,” he said, noting that a lack of novelty may be a challenge in obtaining patent protection.
“The challenge from an ownership perspective is a lot of these processes take place, and are increasingly taking place, in cloud environments. And so, there’s a huge amount of difficulty in detecting that infringement and being able to then understand, and apply, and do the analysis necessary to determine whether or not there’s been patent infringement,” he added.
Jurisdictions around the world will have different perspectives in applying patent law, Green said, so there will be “a set of fundamental questions about whether the law today is fully equipped to protect, not just the aspects, but the output that AI delivers.”
AI is “exploding,” Green stressed, noting that it remains to be seen “whether and how” courts can grapple with a concept of authorship.
“There’s a number of cases out there, currently existing in copyright law that go back 15 years, that look at factual components and recognize copyright ability because of the selection, and arrangement, and the judgment that was applied in determining that particular output and have given it a minimal level of copyright ability. Certainly, from a patent protection [perspective], there [are] certain limitations. That doesn’t appear to have slowed down the patent activity,” he said, adding that in the future we will have to consider how to shape the expansion of intellectual property to protect AI’s output.
“I think from our perspective; intellectual property is doing quite well. It’s furthering the policy initiatives it set out to do. It’s certainly not slowing down, and I think the amount of activity, and the growth of activity, suggests that at least for the time being we’re in good hands with our current intellectual property statutes and provisions,” he concluded.
Catherine Lacavera, director of intellectual property, litigation and employment at Google Inc., said that while she thinks AI is at a crossroads, IP is not. She noted the more interesting challenges with AI lie on the regulatory and social impact side of the discussion.
“I think that everything that applies to software patents, equally applies to AI. So to the extent that there are things that need to be worked through on tightening up the patent system for software patents, I equally think that applies to the AI space,” she explained, adding that the need for data combined with privacy protection concerns is a challenge in innovation.
“There’s this sort of grappling struggle between enabling innovation versus all the privacy challenges and potential abuses of the technology. I think along those lines, and in a regulatory mindset, [are] where the bigger challenges are,” she said, noting that Google has released a set of AI principles to help protect against abusive technologies.
Lacavera said that according to the Google principles, AI needs to be socially beneficial.
“You don’t want to be overly restrictive in how people use open source AI technology, but on the other hand, we have committed to not allowing use in certain areas, like weapons development or surveillance, these kinds of things where you can see where really powerful AI can become concerning,” she explained.
Maya Medeiros, a partner at Norton Rose Fulbright LLP, stressed the importance of collaboration in AI development and how IP can help facilitate that.
“Collaboration in development can help maintain leadership positions for AI innovation,” she said, noting that this helps avoid people having to start from scratch as it connects silos in technology, data and knowledge.
To remain competitive globally, Medeiros explained, collaboration is key.
“How can IP rights facilitate these multiparty collaborations? To protect AI innovation, how can the law incentive collaborative behaviour?” she asked, noting that protecting the freedom to operate transformative technology can provide the answer.
“I think the importance of IP rights and the freedom to operate come together in order to provide these more defensive assets for the law to prevent others from claiming exclusive rights,” she said, suggesting a “freedom to operate licensing” to encourage collaboration.