Reconciling AI-Created Innovations with the Patent System
A Kobayashi Maru
In the 1982 sci-fi film Star Trek II: The Wrath of Khan, Starfleet Academy cadets participated on a training exercise to rescue the civilian vessel Kobayashi Maru in a simulated battle with the Klingons. The trainees are left with two choices: rescue the Kobayashi Maru crew but face certain defeat; or leave them to die. The simulation was designed to be unbeatable, but the infamous James T. Kirk managed to do the impossible during his time as a cadet. The movie later reveals (spoiler alert) that he only passed the test because he reprogrammed the simulation to have a third scenario where the Kobayashi Maru can be rescued with zero casualty. Simply put, Kirk cheated.
The current patent system presents a similar no-win situation with the advent of AI. We are experiencing a rapid growth in technology that some machines have sufficient computing power to be considered creative. In certain instances, the machine creates a product that can be a patentable breakthrough. While in reality the computer itself is the inventor, the law refuses to acknowledge this. The present system is failing to keep up with the times and some developers are advocating the recognition of AI-inventors to incentivize the building of creative machines properly.
Patent experts discuss three types of computer creations. The first is when a human uses a computer merely as a tool, albeit a sophisticated one, to create new technology. In this case, the law regards the human user as the inventor. The same logic may be used for a second scenario in which an AI system outputs a large number of random “ideas” based upon information inputted in the system, which a human then assesses and evaluates. Here, the AI system merely generated a multitude of random outputs without considering the implications or assessing the advantages or utility. If anyone at all, it was the human who did the inventive process and should be attributed inventorship over the creation.[1]
The subject of this article is the third situation. The AI system generates a new concept, without any specific human input, teaching or guidance, evaluating that concept, and concluding whether the idea has technical value or utility. Here, no human should claim involvement in the generation of that concept or its evaluation.[2] It’s a tricky situation. On one end of the spectrum, if the application indicates the human to be the inventor, there is misrepresentation leading to the rejection of the patent. On the other, if the machine is said to be the inventor, the patent office will deny the application because the inventor is not human. Like the Kobayashi Maru, re-writing the rules is the only solution.
The Patent and Copyright Clause of the Constitution gives Congress the power to “[t]o 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.” This clause embodies the reason behind the patent and copyright system, which is to encourage innovation. People will be more inclined to invent things if they receive government-sanctioned monopolies for the commercial fruits of their inventions.[3]
The statute, however, limits inventorship to natural persons. Under 35 U.S.C. § 100(a), “inventor” refers to “the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.” Thus, the U.S. Patent and Trademark Office (USPTO) opined that only natural persons could claim inventorship.
On April 27, 2020, the USPTO published a petition decision stating that only natural persons may be named as inventors in a patent application under current law. The agency’s decision stems from its rejection of a U.S. patent application that listed an autonomous entity, named DABUS, as the inventor. DABUS stands for “Device Autonomously Bootstrapping Uniform Sensibility” and is a “Creativity Machine” that uses neural networks to generate new inventions.[4] Back in August 2019, a group of patent lawyers called the Artificial Inventor Project (AIP) filed patents for two inventions — a warning light and a food container — on behalf of Stephen Thaler, CEO of a company called Imagination Engines. AIP listed DABUS as the inventor instead of a human author. According to AIP, the DABUS AI came up with innovations after being fed general data about many subjects. Thaler may have built DABUS, but he has no expertise in creating lights or food containers, and wouldn’t have been able to generate the ideas on his own. Hence, AIP argued DABUS itself is the inventor.[5]
The USPTO ruled that the patent statute precludes a broad interpretation to cover machines under the term “inventor.” By way of example, the office cited several sections of Title 35: “Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter… may obtain a patent therefore, subject to the conditions and requirements of this title”[6]; “each individual who is an inventor… shall execute an oath or declaration in connection with the application”[7]; “[a]n oath or declaration under subsection (a) shall contain statements that… such individual believes himself or herself to be the original inventor”[8]; and “[a]ny person making a statement required under this section may withdraw, replace, or otherwise correct the statement at any time.”[9] The USPTO argues that the use of words such as “whoever”, “individual”, “himself”, “herself”, and “person”, implies that only a natural person can be an inventor.[10]
The USPTO likewise cited some case law, such as Univ. of Utah v. Max-Planck-Gessellschaft zur Forderung der Wissenschaften e.V.,[11] explaining that a state could not be an inventor because it is incapable of conception,[12] and Beech Aircraft Corp. v. EDO Corp., ruling that “only natural persons can be ‘inventors.’”[13] While the USPTO admits that these decisions were rendered in the context of states and corporations, it believes that the discussion of conception as being a “formation in the mind of the inventor” and a “mental act” is equally applicable to exclude machines from the definition of “inventor.”
Notably, it is not only in the U.S. that these patents were denied for the same reason. Applications were also filed with the U.K. Intellectual Property Office (UKIPO) and the European Patent Office (EPO). While accepting that DABUS is responsible for the invention, the UKIPO ordered the application’s withdrawal because DABUS is not a person as envisioned by the law to be an inventor. The UKIPO however noted that AI inventions are likely to become more prevalent and that it is correct to debate these issues since the current patent system does not provide a means for handling such creations. On the other hand, the EPO ruled that the name of a machine does not meet the requirements of stating a family name, given names, and a full address of the inventor. The EPO further explained that legislation would be required to create a legal personality for AI systems or machines because “legislative history shows that the legislators of the EPC (European Patent Convention) agreed that the term ‘inventor’ refers to a natural person only.” The EPO also noted that this is the “internationally applicable standard” followed by the majority of the EPC contracting states, as well as China, Japan, Korea and the U.S.A.[14]
In the field of copyrights, the U.S. follows the same conservative approach. From as early as 1973, the U.S. Copyright Office has applied the “human authorship policy” that prohibits copyright protection of works that are not generated by a human author. This is the same policy brought into light by the famous “monkey selfies” case.[15] The People for the Ethical Treatment of Animals (PETA) filed a claim on behalf of an Indonesian crested macaque, arguing that the monkey should own the copyright to the photographs.[16] The court dismissed the case and basically ruled that animals can’t sue for copyright infringement.
But whether animals are capable of creating art is not exactly a pressing question for our time. The more critical issue is if computer-generated works can be copyrighted. In 1978, the National Commission on New Technological Uses of Copyrighted Works (CONTU) reported to Congress that a computer could not be an author. “On the basis of its investigations and society’s experience with the computer, the Commission believes that there is no reasonable basis for considering that a computer in any way contributes authorship to a work produced through its use,” the report stated.[17] However, the CONTU compared the computer to a camera or a typewriter, both being instrumentalities that in themselves were utterly lacking in creative capabilities while requiring human direction to bring about an original result.[18]
However, today’s technology is very different. With the emergence of AI software, computers have created and will continue to create original works that could be protected if not for the human authorship policy. Ryan Abbott, a patent attorney, and professor of health sciences at the University of Surrey in the U.K., is among the lawyers working on the AIP. In his article, I Think, Therefore I Invent: Creative Computers and the Future of Patent Law,[19] he concludes that computers independently meet the requirements for inventorship, but the statutory language requiring inventors to be individuals and judicial characterizations of the invention as a “mental act” present barriers to computer inventorship.[20]
If a machine can think, it would seem logical that it could compose or design. If AI does create something, such as a piece of music, and it is impossible to tell by hearing whether it was written by a computer or by a person, one might wonder if the notion of machine authorship ought to be accepted.[21] The U.S. has not made reforms to address this, but some parts of the world have initiated actions to keep the field of copyright adaptive with the advances in technology. As early as 1988, the U.K. Copyright, Designs and Patents Act recognized a “computer-generated” work as one without a “human author” and granted such work copyright protection. In 2017, the European Parliament advocated giving autonomous robots the legal status of “electronic persons” for copyright protection purposes. Also, the Delegation of the European Union to Japan met in Tokyo to discuss, among others, whether AI-generated works are copyrightable under Japanese and European law.[22]
The U.S. perspective in inventorship and authorship remains rooted in the traditional belief that only humans are capable of intellectual creation. However, this conclusion was reached only by analyzing creation issues concerning animals, corporations, and the state. Intellectual property considerations in works of machine creators vis-à-vis human creators are yet to be tackled by U.S. courts.
That is not to say that there has been no movement for the recognition of AI inventors in the U.S. The USPTO acknowledged the emergence of creative technologies and their capability to produce otherwise patentable inventions.
It hosted an event entitled “Artificial Intelligence: Intellectual Property Policy Considerations (January 2019).” Due to the unique issues with advanced technologies, the USPTO is considering whether it should make any changes to how it handles examination of these applications. For its analysis, the USPTO issued a request for public comments on the protection and study of these inventions. It even expanded the scope of its inquiry to cover copyright, trademark, and other intellectual property rights impacted by AI.[23]
However, absent any new legislation or case law recognizing AI inventors, the USPTO is unlikely to retract its opinion in the DABUS petition. Even with the policy considerations, this is actually a safe approach for the USPTO.[24] It would prefer to follow the plain language of the law and leave it to Congress or the courts to direct the office to do otherwise.
With the proliferation of AI capable of conception, how must the legal landscape adapt?
Until the patent system is reformed, AI developers will tend not to disclose a computer’s participation in the inventive process to successfully obtain a patent. But this poses serious risks. For one, if the patent office is aware that the human is not responsible for the invention, the application should be denied.[25] Another problem here is exposing the applicant to liability. The listed inventor must sign an oath or declaration that such individual believes himself or herself to be the original inventor.[26] A false oath or declaration carries a severe punishment of a fine or imprisonment.[27]
If a person who discovers the subject matter of the invention can be an inventor, why not list him or her instead rather than the machine? Abbott thinks that this would entail questions. AI may be functioning more or less independently, and it is only sometimes the case that substantial insight is needed to identify and understand a computational invention. In cases where a person just thinks of a problem to be solved, but the computer solves it with no contribution from the human, giving the person an incentive would be unfair.[28] There is also a logistical problem since the first person to notice and appreciate the solution would be deemed the inventor, regardless if he or she owns the AI or had any right to see the created technology at the outset.[29]
According to Abbott, treating computational inventions as patentable and recognizing creative computers as inventors would be consistent with the Constitutional rationale for patent protection.[30] We already know that the foundation of patent law is the incentive theory. If the present law continues not to recognize machine inventors, innovation would be hindered without any form of legal benefit available to would-be developers of such machines. But if creative machines are accepted, although AI would not be motivated to invent by the prospect of a patent, engineers would be encouraged to develop more of these machines.[31]
The financial benefit is particularly important since developing AI demands much resources.[32]Moreover, patents not only confer a right to use the invention but also a right to stop third parties from using or exploiting said invention. Of course, AI systems are already developing at a rapid pace and are generating many new technical solutions, patentable or otherwise. But if the patent system is re-wired to adapt to this new era, the appropriate party would have some form of control over the exploitation and dissemination of inventive ideas generated by AI systems.[33]
The “appropriate party”, i.e. owner of the patent, is not the computer. A patent remains to be property that only persons can own. Man may be too proud at present to recognize personality in machines. We will be keeping all our rights to ourselves as long as we can. Thus, the patent owner should still be an entity capable of owning property, whether it be a person, a corporation, or a state.
Abbott suggests that, subject to contract, a computer’s owner should be the default assignee of any invention because this is most consistent with the rules governing property ownership and would most incentivize innovation.[34] He likens it to an employer-employee relationship, where the owner is the former and the machine is the latter. To illustrate, if Thaler is the owner and DABUS is the inventor, Thalus owns the patent to any of DABUS’ creations.[35] Like any of his properties, Thaler is of course free to sell, lease, license, or even donate, DABUS or any of his patents. His lawyers can be as creative as possible to craft terms to protect his interest. What is important is that at the creation of DABUS, Thalus is properly incentivized for producing an AI-inventor.
With the evolution of AI and emergence of machine-created technical solutions, there is a need to address intellectual property issues that many innovators are facing. They have this dilemma among them on who should be listed as the inventor of products created by their AI. They are not comfortable in claiming all credit for themselves but at the same time, they fear exposing the invention to the public domain without the protection of a patent. With the haunting uncertainties, regulators must act to define rights arising from AI-created inventions. There will come a day when society will be looking to machines to come up quickly and efficiently with solutions. The incentive theory of patent law must remain as an existing policy to promote the creation of these kinds of AI. With so many challenging problems today, we need now more than ever to change the rules to encourage innovation.
[1] See Robert Jehan, Should an AI system be credited as an inventor, http://artificialinventor.com/should-an-ai-system-be-credited-as-an-inventor-robert-jehan/.
[2] Id.
[3] See Ryan Abbott, I Think, Therefore I Invent: Creative Computers and the Future of Patent Law, 57 B.C.L. Rev. 1079 (2016), http://lawdigitalcommons.bc.edu/bclr/vol57/iss4/2, citing Mark A. Lemley and John Stuart Mill.
[4] See Bereskin & Parr LLP — Isi Caulder and Ray Kovarik, USPTO: AI Cannot Be Named as Inventor on Patent, April 28, 2020, https://www.lexology.com/library/detail.aspx?g=73ac519f-7fbb-4206-9aba-c16e1bf731df.
[5] See Angela Chen, Can an AI be an inventor? Not yet., January 8, 2020, https://www.technologyreview.com/2020/01/08/102298/ai-inventor-patent-dabus-intellectual-property-uk-european-patent-office-law/.
[6] 35 U.S.C. §101.
[7] 35 U.S.C. §115(a).
[8] 35 U.S.C. §115(b).
[9] 35 U.S.C. §115(h)(1).
[10] Visit https://www.uspto.gov/sites/default/files/documents/16524350_22apr2020.pdf for the full text of the USPTO decision.
[11] 734 F.3d 1315 (Fed. Cir. 2013).
[12] The Federal Circuit explained in this case: “Conception is the touchstone of inventorship, the completion of the mental part of invention. It is 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. Conception is complete only 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. [Conception] is a mental act…”
[13] 990 F.2d 1237, 1248 (Fed. Cir. 1993).
[14] See Rebecca Tapscott, USPTO Shoots Down DABUS’ Bid For Inventorship, May 4, 2020, https://www.ipwatchdog.com/2020/05/04/uspto-shoots-dabus-bid-inventorship/id=121284/.
[15] See Naruto v. Slater, №16–15469 (9th Cir. 2018).
[16] See Ryan Abbott, The Artificial Inventor Project, December 2019, https://www.wipo.int/wipo_magazine/en/2019/06/article_0002.html.
[17] See CONTU Final Report (1979) at 44.
[18] Id.
[19] See Abbott, supra note 3.
[20] Id. at 1083.
[21] See Pamela Samuelson, Allocating Ownership Rights in Computer-Generated Works, 47 U. Pitt. L. Rev. 1185 1985–1986.
[22] See Edward Klaris, Copyright laws and artificial intelligence, December 2017, https://www.americanbar.org/news/abanews/publications/youraba/2017/december-2017/copyright-laws-and-artificial-intelligence/.
[23] See Artificial Intelligence (AI) Patents — Will the Patent Office Change the Rules?, January 7, 2020, https://www.natlawreview.com/article/artificial-intelligence-ai-patents-will-patent-office-change-rules.
[24] The USPTO stated: “Lastly, petitioner has outlined numerous policy considerations to support the position that a patent application can name a machine as an inventor. For example, petitioner contends that allowing a machine to be listed as an inventor would incentivize innovation using AI systems, reduce the improper naming of persons as inventors who do not qualify as inventors, and support the public notice function by informing the public of the actual inventors of an invention. These policy considerations notwithstanding, they do not overcome the plain language of the patent laws as passed by the Congress and as interpreted by the courts.”
[25] See USPTO decision in the DABUS petition, supra note 10.
[26] 35 U.S.C. § 115(a) and (b).
[27] See Robert Jehan, supra note 1.
[28] See Abbott, supra note 3.
[29] Id.
[30] See Abbott, supra note 3.
[31] Id.
[32] Id.
[33] See Robert Jehan, supra note 1.
[34] See Abbott, supra note 3.
[35] See Angela Chen, supra note 5.
