Since the late 1800s, people have been intrigued by robots. There’s something strange, wonderful, but sometimes scary about walking, talking, thinking machines, especially when in human form. Talking about “whether a robot can hold a patent” is bound to intrigue humanoids. 

Mute the Robot Sound Bite

In June 2021, we started considering the provocatively titled podcast transcript, “Can a Robot Be Arrested? Hold a Patent? Pay Income Taxes?”, posted on the IEEE Spectrum site. Steven Cherry interviewed Ryan Abbott, physician, lawyer, and professor, about these topics and referencing his 2019 book, The Reasonable Robot: Artificial Intelligence and the Law. Our previous discussion, “Can a Robot be Arrested and Prosecuted?”, addressed criminal liability for crimes committed by artificial intelligence (AI) systems. Now we consider: “Can a robot hold a patent?”

The sober (boring) lawyer has to answer the patent question with “No.” Under current U.S. law, only human beings can be identified as the “inventor” in a patent application. Only humans can be granted a patent – not a corporation, not a government agency, and not C-3PO, Sonny, Ava, Colossus, Rhoda, Lisa, Uncle Simon, or any other robot. 

Artificial Intelligence Inventors

Answering the simple “robot” question doesn’t end the conversation, however. Abbot’s Reasonable Robot book expands upon some real questions about AI and intellectual property law. For example: When human beings use an AI system to help invent a new hardware circuit, a previously unknown biological antibody, or a new flu vaccine, who is considered the “inventor” to own the initial patent rights?

Sci-fi fans may clamor to confer patent rights upon robots and other thinking machines. But consider the tech realities. Some humans devised the hardware, and some humans devised the complex software, which together comprise the AI system. Then there are humans who describe the task for the AI system to perform, and they input the data to define what “success” looks like for the system to seek.

Recall several powerful AI systems. Deep Blue, the famous chess-playing system that beat Garry Kasparov (the world’s best chess player at that time) was a hardware+software system designed and built by humans. John Koza and his team showcased his “invention machine” in the 2000s, which reputedly invented a system to improve factory operations and routinely “outperforms humans” in design and operational projects. IBM’s Watson AI system platform “is a data analytics processor that uses natural language processing, a technology that analyzes human speech for meaning and syntax,” and then “performs analytics on vast repositories of data that it processes to answer human-posed questions, often in a fraction of a second,” returning “answers to questions that companies could never solve before.”

Yet human fingerprints appear all over the AI-assisted invention process. All such systems required humans to design, build, and interconnect hardware components on both micro and macro levels. The systems all required extensive human intelligent design to create the software to perform the tasks, whether to write songs, imagine food recipes, or cure cancer. If the human made the machine, and the machine developed a new device or biological product, then the first cause is still the human. In addition, the machine would not know to do anything until someone switched it on, told it what to look for, fed in the data, and pushed “compute.”

AI: Assistant or Inventor?

In Reasonable Robot, Abbott agrees first that “at one end, AI is a tool assisting an inventor.” But he also observes that sometimes AI “functionally automates conception” (p. 78). In other words, there are times when AI systems start with input data and desired goals, and produce results that a team of experts might never have sufficient time and resources to ever obtain. In those cases, have humans “invented” the new thing? Or did humans just create the conditions for creativity, and the AI system actually did the inventing?

One prong of that question leads to discussions about “What is creativity?” and “How does creativity arise?”, whether in humankind or machine. Such a discussion goes to the heart of what it means to “invent” something and who should be able to own, control, or take credit for it. There is a body of law, chiefly recognized in Europe, concerned with preserving the dignity and personhood of creators by providing them with certain “moral rights,” including a “right of integrity” that gives artists the legal right to prevent buyers and owners of their works from altering those purchased works at any time without the creator’s approval. In the U.S., a copyright statute, 17 U.S.C. § 106A, recognizes and protects some moral rights concerning portraits, sculptures, and certain other works of visual art. The creator of protected works retains the rights to claim authorship, prevent the use of his or her name on someone else’s work or on any work that has been modified or mistreated in a way that damages the creator’s honor or reputation. 

The more immediate concern for multi-billion-dollar industries, however, is: “Can we patent the results of AI computation?” If the owner of the AI system cannot get patent protection for an invention that came via the AI system, then using an expensive AI system may turn out disadvantageous.

Do Robots Dream of Fame and Fortune?

To “patent” an invention nowadays means to obtain exclusive rights of sale and use, which translate into a medium-term monopoly on the first sale and use of the invention. A registered patent allows the patent owner, for the period of the patent, to file lawsuits against anyone who uses the invention without permission. In the U.S., a design patent lasts 14 years, a utility patent lasts 20 years, and neither is renewable. 

Governments confer and protect patent rights for straightforward reasons: (1) to give incentives to people to create and invent things that benefit other people in society; and (2) to make publicly available the new invention ideas and technologies so that others in society can perhaps create even better devices and solutions. Both of these results of patent protection work toward raising society’s standards of living. The U.S. Constitution, under Article I, section 8, empowers Congress to create patent and copyright laws:

Congress shall have power… 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.

When people have a strong belief that their creative work will be protected in a way that profits them and/or their companies, people with inventive minds and abilities will be willing to invest effort, time, and money to fashion new things that other people will want and pay for in the future. For example, a study published in 2017 reported the average cost of a new patented pharmaceutical from development to market was over $2.8 billion. A medium-term monopoly allows the innovating companies to recoup costs and return a profit. Without a monopoly, the innovators would face price competition from other companies that either obtained the formulae or reverse-engineered the product – both methods far less expensive than the original invention.

The incentives to invent include: (1) the profit motive; (2) the enjoyment of fame or industry recognition and leadership; and (3) the desire to solve problems and thereby help other people. These incentives do not excite AI systems to think and create. But, as Abbott points out, even though a “patent would not directly motivate AI to invent, it will inspire the people who build, own, and use AI.” (p. 72)

Keep It Simple for Robots

Currently, humans use AI systems to apply heuristic methods, vast data storage sampling, and brute force computation to obtain results that lead to potential utility or design patents. Some systems use “evolutionary” procedures that attempt millions of possibilities and save the ones that best fit the specified goal. In all such systems, humans design and build the hardware and software, supply the sources of data to scan, and direct the AI system toward achieving human-defined goals. Abbot describes the present scenario:

[H]uman inventors are widely augmented by AI. A person might design a new engine using AI top perform calculations, search for information, or run simulations on new designs. In these scenarios, the AI does not meet inventorship criteria but does expand the capabilities of a researcher in the same way that human assistants can help reduce an invention to practice. In fact, researchers may rarely be unaided by AI, depending on the industry they work in and the problems they are trying to solve.

Ryan Abbott, The Reasonable Robot: Artificial Intelligence and the Law, pg. 98

As spectacular as the AI systems can be, giving the appearance of imagination and creativity, the results of AI computing still owe their existence to humans at all key stages. AI systems are tools created by and for humans. Thus, Abbott suggests a clear answer to “who owns an AI system invention?”:

Ownership rights to AI-generated inventions (in the event they are patentable) should vest in an AI’s owner because it would be most consistent with the way personal property (including both AI and patents) is treated.

Ryan Abbott, The Reasonable Robot: Artificial Intelligence and the Law, pg. 87

Abbott’s rationale remarkably accords with Western civilization’s long-standing starting point for property rights:

If a person owns a machine that produces property, then he would own that property whether it is a loaf of bread or a trade secret[.]

Ryan Abbott, The Reasonable Robot: Artificial Intelligence and the Law, pg. 87

As with any device, recipe, or trade secret, the owner of the initial rights can contract with people who buy, lease, or obtain licenses for the rights. A company that wants to use IBM’s Watson system, for example, could buy a copy of the AI system and retain by contract or license the rights to anything the system made. A wide variety of voluntary win-win arrangements could be developed to make AI systems available and cost effective for all concerned.

It does excite the sci-fi senses to imagine robots or massive computer networks fighting in courts for their rights to own patents to their inventions. As it stands today, humans still own and control the robots, reaping the fruits of AI-assisted productivity. May it ever be so. 

You may also wish to read the first post in this series:

Can a Robot be Arrested and Prosecuted? An Uber driver is held liable if he runs over someone. But what if a driverless taxi ran over someone? A legal culture should be developed now to hold people accountable for AI software code designed to steal, cause damage, or even kill. (Richard W. Stevens)