How AI Might Reboot Patent Law
Editor’s note: This article is the fourth in a series of posts about AI and inventorship, developed as part of Professor Colleen V. Chien’s AI and the Law class at Santa Clara University School of Law. The first blog describes the Thaler v. Hirshfeld appeal; the second, the technology behind DABUS; and the third, how patent practitioners view the issue of AI inventorship.
Author Jungyeon Kim (SCU Law ’22) is a technology specialist and an incoming associate at WilmerHale. He was previously an engineer in the server industry.
To be patent eligible, an invention must be useful, novel, and nonobvious. The “nonobviousness” of a claimed invention has been judged through the eyes of a “person of ordinary skill in the art,” or POSITA for short. As a matter of course, this is a human-centric standard. But what happens when machines — not humans — start pumping out inventions? Do we judge their obviousness from the framework of a human mind or an artificial system?
As AI becomes increasingly advanced, such a question is no longer so easily dismissed as a speculative mental exercise to be consigned to the realm of sci-fi. In fact, it has already come up in courts of law around the world. In his case to give AI inventorship, Dr. Thaler argued that his machine should be the sole inventor, yet also argued that the modification of prior arts would have been nonobvious to a POSITA — a person — during the prosecution of “Food Container.”
How Does Patent Law Determine “Nonobviousness”?
While a claimed invention must be novel, useful, and nonobvious to be patentable, the last prong of the inquiry is often the most important.
The novelty requirement asks if the claimed invention was disclosed in the “prior art,” i.e., if there is evidence that your invention was already known, such as prior publication in a scientific journal. The nonobviousness requirement asks whether the claim as a whole was obvious in view of prior arts available at the time of invention.
But obvious to whom? What a Ben Franklin or Leonardo da Vinci sees as gobsmackingly obvious may not be so comprehensible to the rest of us. Here, U.S. patent law (and that of many other countries) relies on the fictitious “person of ordinary skill in the art.” POSITA is defined as “a hypothetical person who is presumed to have known the relevant art at the time of the invention” — not any person, but also not a superstar inventor. It is presumed that this hypothetical person has access to all publicly available prior arts at the time of the invention.
Even after a patent has been granted, the validity of its novelty and nonobviousness can still be challenged — and the case is usually decided on the latter grounds. Practically speaking, patent examiners (or patent challengers in litigation) rarely find a single prior art reference that discloses the whole claim, and thus they spend most of their time examining whether the invention is “obvious.” Litigants will also argue over who exactly should be considered a POSITA for the specific patent in question, e.g., a person with a bachelor’s degree in electrical engineering versus someone with 4 years of work experience and a PhD.
No Country for Old POSITAs
The advent of human-level intelligence in machines threatens to overturn many of the assumptions in this elegantly designed system for issuing patents.
In an October 2020 report on AI and IP policy, the USPTO reported that a majority of respondents found that the current standard of POSITA was sufficient for judging the nonobviousness of AI-generated innovations. In the words of one commentator: “[J]ust as the existence of general purpose computers impacts the level of a person of ordinary skill in the software arts. . ., so [too] would AI affect the level of skill in the arts where it can be made useful.”
IBM, a leading AI company, shared similar views, stating that “the level of skill of the [POSITA] will naturally increase with access to AI as a tool, just as the level of skill of the [POSITA] has improved over the years with access to other new and improved tools.” Automation company Siemens elaborated on IBM’s comment, stating that if a POSITA is presumed to have access to AI-created inventions, such as AI searching tools, then “the hurdle of non-obviousness will increase but this might be a wanted balancing effect to limit the number of (automated) AI created inventions.”
Thinking through the issue to its logical conclusions, Ryan Abbott has raised some important questions about the viability of sticking with the POSITA standard, even as it adapts to technical progress in the field. Abbott, Dr. Thaler’s lawyer who is also on the frontline in advocating for AI inventorship, generally agreed with the other commenters. He argued that as the importance of AI to innovations in specific fields grows, the POSITA standard would also encompass the routine use of inventive AI by average workers. Taking it a step further, however, he argued that conceptualizing POSITA “as a skilled person using AI might be administratively simpler but replacing the skilled person with the inventive AI would be preferable because it emphasizes that the AI is engaging in inventive activity, rather than the human worker.”
Yet he also pointed out that simply substituting an inventive AI for a skilled person could exacerbate existing problems. With the current POSITA standard, patent examiners must reason, in hindsight, about what another person would have found obvious, which Abbott says leads inevitably to inconsistent and unpredictable determinations, much like Justice Stewart’s highly subjective definition of obscene material: “I know it when I see it.”
“This may be even more problematic in the case of inventive AI,” Abbott wrote, “as it is likely to be difficult for human decisionmakers to theoretically reason about what an AI would find obvious.”
