Inventions such as the wheel, the printing press, light bulb, telescope, microscope, transistor, microchip, and the Internet, are amazing in and of themselves. However, these, and thousands of other inventions have also provided an indispensable foundation, and a toolkit, for other, newer inventions, leading to a pace of innovative progress unlike anything seen before. For example, the microchip, leading to the computer, has helped humans conceive of and find new inventions by helping them process information more efficiently. But the computer, until recently, has only helped to solve inventive problems framed by humans and arrive at solutions that are, in some sense, only anticipated by humans. Until now, prior inventions have only provided assistance to the inventive activity of human beings; historically, the human mind has ultimately been the source of invention.
That paradigm, however, is changing. Recent advances in computer technology, as well as the exponential growth in available data, are leading to the advent of artificial intelligence and machine learning. Some have said that most of the data ever created has been created in the last several years. What we call “artificial intelligence” represents a massive increase in the power of computer problem solving that has been enabled by massive amounts of new data. Data is like fuel – the more data available to computer algorithms, the more powerful those algorithms become in operations that approach machine learning. And, with this new power, machines are becoming increasingly able to formulate problems and “imagine” (i.e., invent) solutions in ways that were previously reserved for human beings.
The possibility that a machine can be an inventor raises interesting questions for how we think about incentivizing inventorship and the kinds of monopolistic protection we afford to inventions in the future. Patent law is the body of law that deals with, and specifically, provides certain protections for, inventions. The concept of inventorship is core to patent law, and, with the change in the inventorship paradigm noted above, the question naturally arises who, or what, under the law can be an inventor? Can a machine be an inventor? More specifically, can artificial intelligence software be listed as an inventor on a patent application? This is the question that was recently addressed by the United States Court of Appeals for the Federal Circuit on Aug. 5.
In Thaler v. Vidal, the Appellate Court held that an inventor must be a naturalized person. Put another way, only human beings can be inventors. This case arose when Thaler tried to acquire patents for inventions developed by his “Creativity machine” known as DABUS. The United States Patent and Trademark Office (USPTO) denied Thaler’s applications, claiming that there must be a human inventor. Similarly, patent courts in the European Union, the UK, and Australia, all ruled against Thaler. Only South Africa allowed for an artificial intelligence inventor and granted Thaler a patent.
Here, in the United States, Thaler appealed the USPTO decision to the US District Court before appealing to the Appellate court. Both the District Court and the Appellate Court made the same conclusion that non-human entities cannot be inventors. No other American courts have addressed this issue, and unless the United States Supreme Court has an opportunity to consider the issue (in Thaler’s case or in a future case), the Federal Circuit Court of Appeals is the final authority on patent matters.
In its analysis of the issue, the Court of Appeals declined to engage in an analysis of the nature of an invention or the rights that might be attributed to artificial intelligence. Instead, the court left these issues open in favor of the safer, if perhaps no less controversial, practice of statutory interpretation. The Patent Act states that an inventor is “the individual, or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.” Because the patent statute does not define “individual,” the appellate court instead relied upon a previous United States Supreme Court case, Mohamad v. Palestinian Authority, in which the Supreme Court held that the word ordinarily refers to a human entity. Thus, the Appellate Court ultimately held that the term “individual” in the Patent Act refers only to natural persons and that artificial intelligence does not count as an inventor on a patentable invention.
The Mohamad case dealt with the application of the word “individual” as it pertains to the Torture Victim Protection Act of 1991 (the VPA). It is also worth emphasizing that in Mohamad, the Supreme Court held only that the term “‘individual’ ordinarily means ‘[a] human being, a person,” and that its holding with regard to the VPA does not mean that the “word ‘individual’ invariably means “natural person.” Furthermore, the Supreme Court opinion dealt with whether a corporate or governmental agency could be considered an “individual,” and did not address the applicability of the word to a singular, individual, artificial intelligence.
The Appellate Court buttressed its decision denying the title of “inventor” to artificial intelligence by noting that nothing in the law shows or implies that the legislature intended the word “individual” to mean anything other than a natural person. The Court pointed to the fact that the Patent Act uses pronouns such as “himself” or “herself” when referring to inventors, indicating that congress did not intend to allow non-human inventors. The act does not use “itself,” which is the term that the court reasons Congress would have used if it “intended to permit non-human inventors.”
However, these are not the only ways in which the legislature could have illustrated an intent that the term “individual” be interpreted broadly. Indeed, as Thaler argued before the court, limiting innovation to natural persons is contrary to the general policy behind the Patent Act, namely to encourage innovation and public disclosure. As already stated, artificial intelligence could facilitate innovation at a rate and efficiency previously unseen. By limiting patent protections to inventions created purely by a human mind, the Appellate Court removes much of the incentive to utilize what promises to be the most powerful innovative tool in our toolbox. However, the court rejected this argument, and briefly categorized it as speculative, before referring again to its textualist approach.
Because the court relied on this textualist approach and did not consider the nature of inventorship, several questions remain to be answered. For example, because Thaler actually listed DARBUS as the inventor, Thaler presented no fact question regarding inventorship; he was simply asking the Court to determine that DARBUS could be an inventor. The Court expressly acknowledges this point: “We are not confronted today with the question of whether inventions made by human beings with the assistance of AI are eligible for patent protection.” So where, exactly, does the involvement of artificial intelligence in the inventorship process cross the line into an inventive activity that deprives the invention of patentability? How will companies navigate that line and structure their R&D to optimize the benefits of massive computing power and the potential for patent protection?
Additionally, Patents can only be granted if the invention is new and non-obvious. With the advent of powerful computers that can anticipate many inventions of which a human is capable, will those innovations, when eventually created by a human being, be determined to lack novelty and non-obviousness on the grounds that artificial intelligence as “already thought of it?” Will we reach a point in which artificial intelligence preempts the ability of a “natural person” to acquire a patent when that person eventually comes up with the invention on his own? And if that is the case, what effect will that have on the ultimate incentives for innovation generally?
Thaler plans to appeal to the US Supreme Court and argues that the Federal Circuit adopted a “narrow and textualist approach” that “ignores the purpose of the Patent Act” with “real negative social consequences.” Apart from any possible future action by the Supreme Court, further legislation is always possible after lawmakers, policymakers, think tanks, and academics have had the opportunity to re-evaluate existing law and its impact on innovation in light of growing experience with AI and emerging technologies. The Department of Commerce, which houses the USPTO, will no doubt continue to monitor this issue very closely and issue periodic reports. For further exploration of issues related to inventorship as related to artificial intelligence, see the USPTO’s report here; and see generally, the USPTO’s AI website.
