PATENTING ARTIFICIAL INTELLIGENCE: AN ADMINISTRATIVE LOOK INTO THE FUTURE OF PATENT LAW.

• Author: Stamatis, Nicholas James

Table of Contents Introduction 329 I. The Functionality and Patentability of Artificial 333 Intelligence A. What is an Artificial Intelligence System? 333 B. Why Patent Artificial Intelligence? 336 II. The Federal Circuit Approach to AI Patentability 339 A. Subject Matter Patentability under [section] 101 339 III. The Administrative Authority of the Patent and Trademark Office 345 A. History of the PTO 345 B. PTO Rule Promulgation 346 C. Oil States v. Greene's Energy: The Potential for Expanded Administrative Deference 349 IV. Where to go after Oil States 353 A. The Court Deems IPRs Constitutionally Valid 353 B. The Court Holds IPRs to be Constitutionally Invalid 354 C. AI's Place in the PTO's Administrative Future 357 Conclusion 359 Introduction

The overwhelming breadth of technology is expanding exponentially each year, with yesterday's science fiction novels becoming today's newest breakthroughs. (1) Along with these constant technological advances we are forced to examine the reliability of our laws and ensure that they are still applicable today. This idea is particularly true in the area of patent law.

Under the United States Constitution, Congress has 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." (2) This clause established the patent as a substantive vehicle for inventors to prevent unauthorized use of their hard-earned inventive works. (3) Accordingly, the United States Patent and Trademark Office ("PTO") was put in place by Congress as the administrative body responsible for granting exclusionary patent rights to inventors throughout the country. (4) As technology races forward, the statutory definition for what is or is not patentable subject matter is constantly being reinterpreted by the Federal Circuit Court and Supreme Court. (5) Perhaps one of the fastest growing areas of technology is the rapid development of artificial intelligence (AI) systems, which will certainly create a dramatic shift in how we interact with the world around us. (6) The patentability of these AI systems is also becoming an important facet of the law as inventors race to not only implement these systems into all aspects of everyday life, but also maintain ownership and receive just compensation for the various costs associated with their conception. (7)

Despite the myriad of potential uses, the primary concern underlying AI patentability is whether such an invention qualifies as patentable subject matter under 35 U.S.C. [section]101. (8) While this decision is traditionally left to the judicial discretion of the Federal Circuit Court, there has been a recent push to expand the substantive authority of the PTO, and grant it greater administrative and judicial deference which could play a large part in future subject matter litigation. (9)

To determine the patentability of AI systems, Part I of this Comment examines the functionality of AI, its prevalence as an emerging technology in society. (10) Part II discusses the elements that must be satisfied under 35 U.S.C. [section] 101 in order for AI to be patentable in the eyes of the courts. (11) Part III discusses potential administrative changes and the granting of greater judicial deference to the PTO through the Supreme Court's recent decision in Oil States (12) Finally, Part IV details an alternative method through which the PTO could expand its authority should the Court narrowly interpret the Oil States holding in the future, and how future AI patents may fit into this scheme. (13)

1. The Functionality and Patentability of Artificial Intelligence

   1. What is an Artificial Intelligence System?

      For many the words "Artificial Intelligence" conjure up images of a far-off dystopian future where computers have become indistinguishable from humans. (14) While this very well could be where we as a species are heading, today the field is much more optimistic and not so reliant on science fiction tropes. (15) Today, AI is recognized as a subfield of computer science which seeks to simulate an expansion and extension of the human mind through computerized neural networks. (16)

      In the last fifty years, AI has seen tremendous advancement from being able to play chess, to being the seminal technology found in self-driving cars. (17) Most often AI systems are designed to simulate human sensory perception through the implementation of specialized algorithms which allow the machine to "learn" and adapt to the problem it is given. (18) The various methods of developing computational intelligence can be separated into three main categories: neural networks, fuzzy systems, and genetic algorithms. (19)

      Neural networks are large conglomerates of artificial neurons that are analogous to biological neurons in their structure and function. (20) Each artificial neuron within the computer system is designed to have a biological counterpart which simulates how a human's neurological network would function. (21) By utilizing computerized neural networks, large sets of complex data can be analyzed more similarly to how the human brain analyzes the world around it, thus allowing the computer system to "learn" accordingly. (22)

      Fuzzy systems create an algorithm for the computer system which rely on computational "uncertainties" to map decisions and learn. (23) Most computers analyze data in a true/false, yes/no structure using Boolean logic. (24) Using fuzzy logic allows the computer to evaluate data that may be considered uncertain at the time, and thus registers it differently than it would if a simple true/false problem arose. (25) Being able to analyze these uncertainties helps simulate human mental processes further and advances the overall goal of AI systems. (26)

      Finally, genetic algorithms may be used to simulate the evolution and mutation of chromosomes themselves in an attempt to create a constantly adapting computerized intelligence network. (27) These algorithms function by assigning variables to act as simulated chromosomes which are then able to "reproduce" based on the number of favorable computational results which that "chromosome" has obtained. (28) This algorithmic process is repeated again and again until the most successful variables and processes are mapped out to solve that particular problem. (29)

      While this information is dense and deeply rooted in the realm of computer science, it is important to understand the end result. (30) Through the simulation of human intelligence utilizing a variety of the methods and algorithms discussed above, we are able to apply these problem-solving machines to a variety of real-world processes in an attempt to automate and further refine the world around us. (31) As technology continues to rush by us the applications of AI systems truly seem endless. (32)

   2. Why Patent Artificial Intelligence?

      Patents represent exclusionary property rights for an inventor. (33) The patent prevents others from copying the inventor's work and allows him to profit from the technology which he has invented, thus providing a global incentive to invent and improve upon technologies. (34) These concepts are especially important in the field of AI as society becomes more and more saturated with these systems and the benefits they provide. (35) In order to better understand why inventors want to patent these systems, we must further examine the primary incentives for inventors to invent. (36)

      Innovation is born out of necessity. (37) The idea that there is always an easier way or that there is always something that can be done better underpins the entirety of human ingenuity. (38) The patent system helps by incentivizing the inventors who identify and address these societal problems. (39) An inventor may file for a patent the moment that their invention has been reduced to practice and claim a variety of exclusionary property rights encompassed by their invention. (40) The patent application itself is comprised of two components: the specification and the claims. (41) The specification discloses the claimed invention and informs the readers of the particulars of that invention while the claims describe the metes and bounds of the inventor's property rights. (42) The most common incentive for patenting technology falls under the reward theory which allows the inventor to recoup research and development costs, as well as make a profit, by either selling the invention or licensing out the technology so that others may capitalize on it. (43) Another powerful tool in the patent holder's arsenal is the ability to bring infringement actions against parties who either directly or indirectly copy the claims of the patented invention in an attempt to make their own profit. (44)

      The ability to commercialize and defend one's invention and claimed property rights fits in well with the overall goal of AI. (45) These AI systems are being designed to revolutionize our societal landscape and automate many of our day-to-day processes, and being able to protect the rights of those who take the time to create and develop these systems is an important concern in any inventive community. (46) There is little doubt that the patenting of AI would allow inventors to receive just compensation and protect them against competitors who would otherwise infringe upon their work. (47)

2. The Federal Circuit Approach to AI Patentability

   1. Subject Matter Patentability Under [section] 101

   Under 35 U.S.C. [section]101, an invention is patentable if its subject matter is worthy of protection. (48) Traditionally there is a bar against patents which attempt to claim biological material, abstract mental processes, laws of nature, and certain types of software. (49) Finding exactly where AI fits in within these subject matter requirements is difficult because of its nature as a computational simulation of human thought process and application to...