2021 Year in Review: Artificial Intelligence – Part 2

This article is part two of a two-part series highlighting 2021’s most important developments in the field of artificial intelligence (AI) and intellectual property (IP). In the first part, we looked at the new developments on the topic of AI and inventorship. In this article, we recap the latest developments on the topic of subject matter eligibility of computer-implemented inventions, including AI inventions.

AI Inventions

In a 2020 report published by the United States Patent and Trademark Office (USPTO) titled “Inventing AI: Tracing the diffusion of artificial intelligence with U.S. patents” (October 2020), it was noted that from 2002 to 2018, the number of AI patent applications filed before the USPTO rose by more than 100%, and the percentage of all U.S. patent applications containing AI grew from 9% to nearly 16%.^[1] A Canadian Intellectual Property Office (CIPO) report similarly outlined a considerable growth in patent filings in AI, noting a 31% increase, on average, annually, in worldwide filings. The report noted a steady increase in the number of patented inventions, with over 18,000 AI inventions patented in 2017 worldwide, suggesting that AI-based patents are increasingly becoming a reality for many inventors.^[2] More specifically, in the same report, natural language processing (NLP) related patent applications were noted as the most prevalent type of AI specialization in both Canada and the U.S.

With the rise of AI patent filings, concern about how AI inventions should be examined has also grown.  While there is no consensus, in the USPTO report, many commenters agreed that AI is best viewed as a subset of computer-implemented inventions, and therefore current guidance on subject matter eligibility [of computer-implemented inventions] should apply.^[3]

Canadian Guidance

In response to the Yves Choueifaty v. Attorney General of Canada^[4] (Choueifaty) decision issued in the summer of 2020, in which the Federal Court set aside the decision of the Commissioner and ordered the Commissioner to consider the application afresh, CIPO issued a practice notice (“Patentable Subject-Matter under the Patent Act”) in the fall of 2020, clarifying the application of claim construction to computer-implemented inventions.

The new practice notice states that CIPO examiners must apply the purposive construction test set out by the Supreme Court in Free World Trust v. Électro Santé Inc^[5], considering the entirety of the specification, and presuming all claim elements as essential unless it is established otherwise or is contrary to the language used in the claims. However, the practice notice also states that an element deemed essential using purposive construction may not form part of the actual invention if it has no material effect on the working of the invention or, in other words, it does not cooperate with other elements of the claimed invention.

The practice notice further emphasizes that the mere fact that a computer is identified to be an essential element of a claimed invention does not necessarily mean that the subject-matter of the claim is patentable. Instead, it is necessary to consider (i) whether the computer cooperates together with other elements of the claimed invention and thus is part of a single actual invention and, if so, (ii) whether that actual invention has physical existence or manifests a discernible physical effect or change and relates to the manual or productive arts. The practice notice suggests that one way for the computer to be considered as forming part of a single actual invention with the algorithm is if running the algorithm improves the functioning of the computer.

In January 2021, the Patent Appeal Board (PAB) issued its decision and concluded that certain claims of the Choueifaty application comprised patentable subject matter within the meaning of section 2 of the Patent Act. The PAB applied the new CIPO practice notice and concluded that Choueifaty’s invention included steps which “…are not merely for yielding information, but for permitting the computer to carry out the portfolio optimization procedures with significantly less processing and greater speed … algorithm improves the functioning of the computer used to run it … [and] the computer and the algorithm together form a single actual invention that has physicality and solves a problem related to the manual or productive arts” (emphasis added).

Notable 2021 Canadian Patent Appeal Board Cases

Throughout 2021, the PAB continued to clarify the guidance issued on computer-implemented inventions through Decisions of the Commissioner of Patents and several applicants were successful in overturning previous rejections and obtaining allowances.

In ExxonMobil Upstream Research Company (Re)^[6], a case involving improving the stability and efficiency of computer simulations of hydrocarbon reservoirs, the PAB easily found that the claims disclosed a physical process that produced discernable physical effects, as the calculations set out in the claims were explicitly applied to effect physical drilling of a well. Indeed, the amended method claim included the step of “causing a well to be drilled based on the solution produced by the model”. The PAB applied similar reasoning in Landmark Graphics Corporation (Re)^[7] and found that the computer-implemented method of calculating the impact of economic uncertainties associated with petroleum production planning had sufficient physical existence, because the computer simulation analysis was, as explicitly recited in the claims, used in physical well perforation and completion processes.

In a subsequent case involving an application for a betting system for race events^[8], the PAB somewhat expanded this logic and found that the step of receiving intermediate race results by a device that is located at intermediate points at the race event, produced a sufficiently discernable physical effect or change. The PAB noted that even if the focus of the claimed subject matter was considered to be placing bets and resolving those bets, claim 1 also involved the reception of measured race results, suggesting that the element providing physicality need not be the main focus of the claim.

While in some of the cases allowed in 2021, the physical effect was substantial, in Desk Chair Learning Systems, Inc. (Re)^[9], the PAB found that measuring time intervals between user inputs was not a generic computer input or output of information and constituted subject matter that had physicality, indicating that physical time measurements may meet the physicality threshold.

The decisions of the Commissioner of Patents also highlighted the importance of a comprehensive disclosure. In BGC Partners, Inc. (Re)^[10], a case involving a method and system for protecting users from erroneous price entries in electronic trading, the PAB relied in part on a paragraph of the description to find that the claimed solution solved a user interface problem and improved data entry, and recommended allowance. The PAB concluded that the claimed method steps were operating to improve the functionality of the trading computer by reducing or eliminating erroneous data entry caused by the computer’s user interface.

Specifically on the topic of AI, on June 9, 2021, the PAB issued its decision in Qiagen Redwood City, Inc. (Re) ^[11], finding that the AI-related application constituted allowable subject-matter under subsection 27(8) of the Patent Act. The application disclosed systems for evaluating genomics data using a built library of biological pathway profiles generated from an ontology database, a knowledge database and profile generation criteria. An ontology is a representation of knowledge as a set of concepts within a domain and the relationships that connect them. The ontology can be used to infer, in this case, classifications of genomics data based upon biological interactions of interest. This type of inference using ontologies is a widespread technique in AI which aims to mimic the human ability to build a living connected map of knowledge. The PAB found that “the new database structure permits the analysis of complex interactions occurring within complex biological pathways and optimizes the production of a library of biological pathway profiles … with significantly less processing and greater speed”, thereby improving the functioning of the computer akin to what is described as allowable in the practice notice.

Not all applicants, however, were as lucky and numerous applications were refused by the Commissioner due to their lack of or insufficient discernable physical effect or change, or their use of generic computers operated in a well-known manner. For example, in Bio-Rad Laboratories, Inc. (Re)^[12], which involved an application relating to determining an optimum quality control schedule for testing reference samples and patient samples, the PAB contrasted the claims that positively recited the analyte measurement module with those that did not, stating that the understanding that the input data came from an analyte measurement module without more was insufficient and only of intellectual significance. 

While in some cases, the specification contributed to a finding of allowability, in others, the specification had the opposite effect. Although the specification weighed in favour of BCG Partners in its first case before the PAB, in the second BCG Partners^[13] decision, the PAB pointed to the specification and figures as suggesting the use of any suitable computer and network technologies to implement the methods for trading within a spread market in question, without any particular configuration being significant, and without any challenges in using them to implement the proposed solution. Similarly, in Smart Technologies^[14], a decision involving assessment grading methods, the description contributed to a finding that the computer elements were generic. In reviewing the specification, the PAB found no evidence that the claimed invention was an improvement over previous computer systems, finding that the specification instead indicated that devices and software that were well-known before the claim date could be used to implement the claimed algorithm.

In United Parcel Service of America, Inc. (Re)^[15], with the independent claims referring to a computer system for selectively printing invoices, relying in part on the title of the application, and the field and background sections of the description, the PAB found that the context of the invention was not limited to paper-based invoices and therefore the terms “selectively printing” did not necessarily produce any physical effect. The proposed solution additionally did not address a computer problem or an improvement to the functioning of a computer, as the specification and figures generally described a conventional computer system.

Throughout the year, the PAB continued to stress that the presence of computer elements having physicality is not sufficient to make an application patentable if the computer elements are generic and used in a well-known manner. For example, in Home Depot International Inc. (Re)^[16], involving systems and methods for automatically identifying products for which sales are disproportionately dispersed, the PAB found that the only element having physicality was the computer. However, while the computer was essential, it was merely used to carry out calculations and did not form part of the actual invention.

Key Takeaways

Though the full impact of CIPO’s 2020 guidelines still remains to be seen, the practice notice as well as the PAB decisions on subject-matter eligibility issues provide some clarity, predictability and optimism to patent applicants for AI and computer-based inventions. From the various 2021 PAB decisions recapped above, it can be said with sufficient certainty that a claim directed to a computer that merely processes an algorithm in a well-known manner, without solving any problem in the functioning of the computer such as speed, efficiency, security etc., will continue to be rejected. It also appears that the PAB will consider the application as a whole, including the title and the description, beyond just the claims, to determine whether a computer problem is being addressed. A similar approach can be seen in determining whether the physicality requirement of the actual invention has been met. Ultimately, a carefully planned-out and strategically drafted patent application is going to play a very important role in pushing the AI and computer-based inventions closer to allowance.