A recent news story reported on whether an artificial intelligence (AI) system should be recognized as the inventor of two new ideas (for food containers and a lamp) and able to have those ideas patented.
Intrigued by the dilemma, Science Node asked Rachel Herder, director of the Intellectual Property Clinic and assistant professor of Clinical Law at Penn State Law, to walk us through the arguments for and against.
What’s at stake?
In the US, a patent for an invention is the grant of a property right to an inventor. At stake here is how do we want to reward innovation—do we want to grant property rights to non-humans?
Traditionally, patent systems around the globe have tried to strike a balance between public disclosure and rewarding an inventor for their hard work. On one hand, the patent system fosters innovation by requiring the public to share new discoveries. To paraphrase Dr. Seuss, “the more you know, the more you can grow.”
On the other hand, the patent system aims to recognize the need to reward inventors who often shed blood, sweat, and tears while creating. Thomas Edison famously tested thousands of materials to find the best filament for his light bulb. Once he received his patent, he then had the ability to exclude others from making, using, or selling the invention for a limited period. In exchange, society received an early public disclosure of Edison’s light bulb.
When you think about whether AI should be an “inventor” under the patent statutes, you are really asking about how you want to reward investment in innovation. How do we want to compensate the companies that develop and AI systems that invent? Is the grant of a patent the correct mechanism for rewarding AI innovation?
It’s important to recognize that each country deals with inventorship in different ways. In the US, the Intellectual Property Clause of the Constitution (Article I, Section 8) empowers Congress “To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.”
With this power, Congress enacted the modern patent statutes, including 35 USC §101 which states, “Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.”
In the US, an inventor is a person who both (1) conceives and (2) reduces to practice the claimed invention. In the US, we have had a long history of interpreting an inventor as a “person.” (To read more about the US’s long history of interpreting inventors as persons and how that concept intersected with our country’s slave-owning past, see Brian L. Frye’s Invention of a Slave.)
Why should AI be able to be an inventor? Why not?
Let’s take an example: A pharmaceutical company spends several years (and a large research and development budget) developing an AI system that does, by itself, conceive and reduces to practice a composition of a new drug for treating cancer. If the AI system can be an “inventor,” the drug can be patented and that patent will be owned by the pharmaceutical company.
A similar outcome would likely be seen if the same pharmaceutical company hired an employee that invented a drug composition and had a contract with the employee that all inventions will be owned by the company (a standard contract term in modern US employment contracts).
Compare the situation if AI could not be an inventor: the pharmaceutical company would be unable to obtain a patent. To protect their invention, the pharmaceutical company might try to keep the invention a trade secret; however that would be nearly impossible in a highly regulated industry (e.g., the Food and Drug Administration would require certain disclosures during the regulatory approval process).
The pharmaceutical company, recognizing trade secret protection would not allow it to see returns on its large investment, might never begin the expensive drug discovery process in the first place. Or even if it did, it may not be able to invest in the clinical trials necessary to bring the drug to market. Here, the end result would be that a life-saving drug might never get to patients.
So, an argument for letting AI be an inventor is that the result will be business-as-usual for a company investing in an AI system. The inventions made will belong to the company and the company can see a return on investment.
However, I could also see a situation in which AI could be used by non-practicing entities (i.e., “patent trolls”) to stifle innovation. In other words, these patent trolls could use AI to invent a large number of patents they could use to block companies from actually making the useful product.
AI will be creating more and more things—this is a reality that is happening now and will only increase over time. I am not convinced that treating AI differently than humans makes practical sense under inventorship law. If we want to continue to reward investment in innovation, then AI needs to be recognized for its role as an inventor. If we want to stop patent trolls, that is another conversation— I do not think blocking AI as an inventor is the best way to prevent patent trolls.
If a computer can’t be the inventor, why can’t the human designer of the AI system just file for the patent on their own behalf?
In the US, each inventor must sign a declaration (or take an oath) stating that they are the original inventor of a claimed invention. The human inventor of an AI system could not say they are the “original inventor” of a food container that was totally discovered by AI.
In the United States, a false statement on a declaration can result in a fine and up to 5 years of imprisonment, so I would not recommend a person sign a declaration if they were not an original inventor. Not all countries have the same original inventor concept as the United States.
Is this the first time this issue has come up?
This might be the first time this exact set of facts has happened but it is certainly not the first time computers have been used to help invent. For example, US Patent No. 4,908,773 filed April 6, 1987 entitled “Computer Designed Stabilized Proteins & Method for Producing Same” details how computers can be used to produce stabilized proteins. In 2009 a company called Cloem started a project that uses algorithms and computers to help draft patents. (See Ben Hattenbach and Joshua Glucoft’s article, Patents in an Era of Infinite Monkeys and Artificial Intelligence).
What other legal issues surrounding AI are likely to come up in the future?
We have already seen some news related to non-human authorship in the copyright space. In 2014, the US Copyright office announced that it will not register works produced by a machine without any creative input or intervention from a human author.
In 2015 there was a lawsuit related to Naruto, the macaque primate, who took a selfie. The court ruled that the United States copyright statutes do not apply to animal creators. While Naruto was an animal, not AI, future courts might apply a similar analysis to non-human creators, such as AI.
I am also looking to see how AI may be used to improve patent drafting—not necessarily where AI is the inventor but how legal technology will be used to improve patents and take away some of the tasks traditionally done by lawyers.