Patent Applications Naming Artificial Intelligence System as Inventor Raise Intriguing Questions
In August 2019, a team of patent attorneys led by Ryan Abbott, a law professor at the University of Surrey in the United Kingdom, filed patent applications in various jurisdictions, including the United States, naming a sole inventor: DABUS, an artificial intelligence system developed by the physicist and computer scientist, Dr. Stephen Thaler. Some of the patent applications relate to a new type of beverage container design based on fractal geometry while others to a device for attracting search and rescue teams, which flashes a light in a rhythm that mimics neural activity. According to the University, if patents are granted, they would be the first patents ever issued with an artificial intelligence system as the sole inventor. These applications raise novel questions of U.S. patent law: Can a U.S. patent issue designating an artificial intelligence system as the inventor? If so, what would that mean for the rights and obligations that normally apply to inventors in the United States? If not, how should the law handle inventions made by artificial intelligence?
These patent applications are the latest in a series relating to Dr. Thaler and artificial intelligence. In 1998, he received his first U.S. patent for an artificial intelligence system he called the “Creativity Machine.” According to Dr. Thaler and Professor Abbott, the Creativity Machine already has invented the claimed subject matter of another U.S. patent, U.S. Pat. No. 5,852,815, directed to certain neural networks. Dr. Thaler has indicated that, although he is listed as the sole inventor on that patent, the real inventor is the Creativity Machine. See Abbott, Ryan, “I Think, Therefore I Invent: Creative Computers and the Future of Patent Law,” Boston College Law Review, Vol. 57:1079 (2016). In addition, Dr. Thaler has said that his artificial intelligence systems have created new innovations, such as the cross-bristle design of the Oral-B CrossAction toothbrush.
According to the press release by the University of Surrey, DABUS, the purported inventor of the latest applications, employs “a system of many neural networks generating new ideas by altering their interconnections” and a “second system of neural networks [that] detects critical consequences of these potential ideas and reinforces them based upon predicted novelty and salience.” In other words, one set of networks generates new ideas while the other set attempts to discriminate the more valuable ideas from the less valuable ones.
U.S. Law Relating to Whether Artificial Intelligence Systems Can Be Inventors
A patent application in the U.S. naming an artificial intelligence system as an inventor raises the question of whether a patent can have a non-human inventor. It appears that the Patent Office will weigh in on the issue initially, but eventually the matter will likely be decided by the courts.
The patent statutes do not expressly address whether an artificial intelligence can be an inventor. That is not surprising, as many of those statutes were drafted in 1952. The law defines “inventor” to mean “the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.” 35 U.S.C. § 100(f) (1952). Another provision establishing standards for patentability, 35 U.S.C. § 102, states that “[a] person shall be entitled to a patent unless…” before listing categories of prior art that can make a claim unpatentable. (emphasis added). The statutory provision on joint inventorship, 35 U.S.C. § 116, refers to those who make inventions as “persons,” referring to the situation “[w]hen an invention is made by two or more persons jointly.” (emphasis added).
Based on those provisions, courts may rule that an inventor must be a “person” and, therefore, a human being. Notably, courts have held that the copyright statutes do not permit animals to sue for copyright infringement. Naruto v. Slater, 888 F.3d 418, 425-426 (9th Cir. 2018). Courts, however, may find that Congress did not intend for “person” in the patent laws to exclude artificial intelligence and only used that language to refer generically to individual inventors. Professor Abbott and his team may argue that Congress was not thinking of artificial intelligence, as we now understand it, when it drafted Sections 102 and 116, and that Congress did not intend to rule out the possibility of artificial intelligence inventors in all cases.
The legal test for whether an individual is or is not an inventor, which has been developed by the courts, focuses on identifying who first “conceived of the invention.” Ohio Willow Wood Co. v. Alps S., LLC, 735 F.3d 1333, 1350 (Fed. Cir. 2013). The explanations of “conception” in the case law frequently use terminology referring to the “mind.” For example, conception has been defined as “the formation in the mind of the inventor, of a definite and permanent idea of the complete and operative invention, as it is hereafter to be applied in practice.” Hybritech Inc. v. Monoclonal Antibodies, Inc., 802 F.2d 1367, 1376 (Fed. Cir. 1986) (emphasis added). Courts have noted that conception is complete when “the idea is so clearly defined in the inventor's mind that only ordinary skill would be necessary to reduce the invention to practice, without extensive research or experimentation.” Burroughs Wellcome, 40 F.3d at 1228 (emphasis added). On that basis, the Federal Circuit has ruled that inventors must be “natural persons,” as opposed to corporations. See Univ. of Utah v. Max-Planck-Gesellschaft Zur Forderung Der Wissenschaften E.V., 734 F.3d 1315, 1323 (Fed. Cir. 2013) (“To perform this mental act, inventors must be natural persons and cannot be corporations or sovereigns.”).
Yet those cases do not address whether the “mind” in question can be an artificial intelligence. The references to the “mind” serve, at least in part, to distinguish conception from what is termed “reduction to practice,” which involves the ability to create a physical manifestation of an invention. It may be that some artificial intelligence system can form a definite idea of the complete invention, as required for conception.
If artificial intelligence systems can be inventors, it will be interesting to see whether courts will allow them to be sole inventors, as is purportedly the case for DABUS, or whether humans who worked with or developed the artificial intelligence should be joint inventors. Individuals should be joint inventors if they made contributions to the conception of the invention. The operator of an artificial intelligence system may provide input to direct its development of new ideas. For example, one of the inventions purportedly created by DABUS involves controlling a light to flash in a rhythm that makes it more noticeable to human observers. Depending on what information and guidance was provided to DABUS in looking for solutions to this problem, a court may find that the people who provided that input made a contribution sufficient to qualify them as joint inventors.
Indeed, some may argue that an artificial intelligence inventor must always have a human joint inventor when the artificial intelligence only sought to create an invention in the first place because it was requested to do so by a human being. Courts have held in some cases that “[o]ne who merely suggests an idea of a result to be accomplished, rather than means of accomplishing it, is not a joint inventor.” Garrett Corp. v. United States, 422 F.2d 874, 881 (Ct. Cl. 1970). But one could argue that a person who made such a suggestion to an artificial intelligence system should be a joint inventor, if the invention never would have been created without that contribution.
Courts determining whether artificial intelligence can be a sole inventor will also take into account the requirement that a patent claim not be obvious in order to be patentable. Obviousness is determined from the point of view of a hypothetical person, the “person of ordinary skill in the art,” (“POSA”). A claim is unpatentable if it would have been obvious to a POSA at the time the invention was made in view of the prior art.
In the case of an invention purportedly created by artificial intelligence, the artificial intelligence itself is not what is new and inventive. Thus, one could argue that the hypothetical POSA would be another similar artificial intelligence system or researchers using such a system. One could imagine a hypothetical artificial intelligence system built to be the same as the one that purportedly made the invention. Would it be obvious for that artificial intelligence POSA to come up with the alleged invention? After all, in patent law, the POSA is presumed to know all the disclosures in the prior art. Given that, could the artificial intelligence that is the named inventor have come up with an invention that the hypothetical artificial intelligence POSA would not also have created? If the answer is that the artificial intelligence inventor was programmed differently or given different input by its users, does that mean that the humans who used or programmed it must be at least joint inventors?
Permitting Artificial Intelligence Inventors Would Raise Issues Regarding the Obligations and Rights of Inventors
In the United States, a key duty of an inventor is to “execute” an oath or declaration. See 35 U.S.C. § 115. Under 35 U.S.C. § 115, the oath or declaration must confirm that the inventor “believes himself or herself to be the original inventor or an original joint inventor of a claimed invention in the application.”
It will be necessary to determine in particular cases whether the artificial intelligence can form the belief that it is the original inventor. That may be possible, as some artificial intelligence systems may be able to confirm that they generated the idea for the invention. Notably, it is no longer necessary for inventors to state that they are the first inventor of an invention or that they submitted the application without deceptive intent. It will be interesting to see how artificial intelligence systems purport to sign a declaration or oath in order to execute it. The law provides exceptions to this requirement allowing someone other than the inventor to sign the oath and declaration if the inventor is dead, insane or legally incapacitated, refuses to sign, or cannot be found or reached after diligent effort. But none of those exceptions apply to an artificial intelligence inventor.
The possibility of an artificial intelligence inventor also raises questions
relating to ownership of inventions. Under U.S. law, every inventor has an ownership stake in the invention. An inventor can enter into an agreement assigning its ownership to another, but until it does, the inventor is the owner.
In order for artificial intelligence systems to be inventors, it will be necessary for ownership of the inventions to pass to human beings or organizations, such as businesses or universities. The artificial intelligence software could be programmed to transfer its rights to its developers or owners, and some may argue that there are artificial intelligence systems that can autonomously decide whether to enter into an agreement. Artificial intelligence systems, however, owned by other persons or organizations will not be able to voluntarily consent to transfer their inventorship rights the way human inventors can. It is not clear how courts would handle attempts by an artificial intelligence inventor to transfer ownership of an invention.
The Implications of a Ruling That Only Humans Can Be Inventors
If artificial intelligence systems cannot be inventors, that would raise the possibility that technological developments that normally would be considered new and non-obvious inventions would not be patentable because they had no human inventor.
One solution to that problem would be to designate the humans who developed or worked with the artificial intelligence as inventors. Although no human generated the new idea, at some point, the idea was shared with a human being. That person would then be the first person who conceived of the complete invention. If only human persons could be inventors, that person would be the inventor. As a practical matter, that person will often be one of the people who was working with the artificial intelligence system, such as its owner, or someone who works in the lab that employs the artificial intelligence system. The artificial intelligence system could be viewed as simply one of the tools that the inventor used to generate the invention. If other individuals provided input to the artificial intelligence system that contributed to obtaining the result it reached, then the law could consider those individuals to be joint inventors. After all, they would have made contributions to the conception of the invention.
One possible objection to such an approach is that the human inventors merely waited for the output of the artificial intelligence system and did not generate the idea through an exercise of innovative thinking. Although courts in the 1930s and 1940s required a patentee to demonstrate a “flash of genius,” see, e.g., Cuno Engineering Corp. v. Automatic Devices Corp., 314 U.S. 84, 91 (1941), that is no longer the law. Congress rejected the Flash of Genius Doctrine in 1952 when it passed the statutes that still form the backbone of patent law to this day. Now, it is “immaterial whether [an invention] resulted from long toil and experimentation or from a flash of genius.” Graham v. John Deere Co. of Kan. City, 383 U.S. 1, 15 n.7, 16 n. 8 (1966). Accordingly, operators of artificial intelligence could be inventors even if they do not make a contribution that demonstrates a flash of genius.
It will be important to monitor how the Patent Office and courts deal with questions raised by the possibility of inventions made by artificial intelligence. Their rulings may have implications on patentability issues, such as inventorship and the requirements for inventors to transfer ownership.