The U.S. Patent and Trademark Office (USPTO) held its Artificial Intelligence: Intellectual Property Policy Considerations conference on January 31, 2019. The conference featured six panels of speakers, including policy makers, academics, and practitioners from Canada, China, Europe, Japan, and the United States. As stated by USPTO Director Iancu during his introductory remarks, the purpose of the conference is to begin discussions about the implications that artificial intelligence (“AI”) may have on intellectual property law and policy. In this post, we provide an overview of Director Iancu’s Introductory Remarks and of three of the conference panels that addressed several current and forward-looking issues that will impact patent law and society at large.
Opening Remarks by Director Iancu
The Director noted that governments around the world are adopting long-term comprehensive strategies to promote and provide leadership for technological advances of the future, and that America’s national security and economic prosperity depend on the United States’ ability to maintain a leadership role in AI and other emerging technologies.
The USPTO is using AI technology to increase the efficiency of patent examination. For example, the USPTO has developed and is exploring a new cognitive assistant called Unity which is intended to allow patent examiners to search across patents, publications, non-patent literature, and images with a single click. The Director concluded by stating that one of his top priorities is ensuring that the U.S. continues its leadership when it comes to innovation, particularly in the emerging technologies such as AI and machine learning.
Patenting AI: Views Across the Corporate Spectrum
Christian Hannon (Attorney-Advisor, Office of Policy and International Affairs, USPTO) moderated a panel which included Yoon Chae (Associate, Baker McKenzie), Kate Gaudry (Partner, Kilpatrick Townsend & Stockton LLP), Manny Schecter (Chief Patent Counsel, IBM), and Scott Streit (Chief Technology Officer, Open Inference). The panel considered a host of issues surrounding patenting and AI. Additional information about conference speakers can be found here.
The panel members agreed that one of the most critical issues surrounding AI patent applications is whether the invention is patent-eligible subject matter under 35 U.S.C. § 101. The two-part test under § 101 involves: i) determining whether a patent claim is directed to a patent-ineligible concept such as an unpatentable law of nature, natural phenomena, or abstract idea, and ii) if so, determining whether the elements of the claim, considered both individually and as an ordered combination, transform the nature of the claim into a patent-eligible application. Kate Gaudry explained that USPTO rejections under § 101 have become increasingly prevalent after the Federal Circuit’s Electric Power Group decision.
Some panel members stressed that the Supreme Court’s § 101 jurisprudence is too difficult to administer, making software-related patents difficult to obtain and difficult to enforce. These panel members explained that, because AI is relatively new, it can be difficult to ascertain what is conventional in the art.
Manny Schecter noted that IBM supports the USPTO’s recently revised § 101 guidance. Its subsequently submitted comments urge Congress to amend § 101, suggest ways to clarify the guidance, and recommend that the USPTO continue its efforts to educate examiners.
The panelists provided several suggestions for preventing and overcoming subject matter eligibility rejections, such as including in patent applications salient details that could be included in claims that the USPTO might allow. In addition, Yoon Chae stated that Federal Circuit decisions, such as Enfish and Berkheimer, can be helpful in combatting § 101 rejections.
As we explained in a previous post, the Federal Circuit in Enfish found that claims directed to “a specific improvement to the way computers operate” are not directed to an abstract idea and are thus patent-eligible under § 101. As explained in another post, the Federal Circuit in Berkheimer held that when a claim is directed to an abstract idea, the question of whether a claim element or combination of elements is “well-understood, routine and conventional to a skilled artisan in the relevant field”—which bears on whether a claimed abstract idea can be transformed into a patent-eligible application—is a question of fact. Furthermore, the panelists explained that the mental steps doctrine is frequently cited by examiners in Office Actions that pertain to AI-related patent applications, and prosecution counsel needs to be able to explain why the invention cannot practically be performed by the human mind.
AI and IP Economics
Andrew Toole (Chief Economist, Office of Policy and International Affairs) moderated a panel which included Iain Cockburn (Richard C. Shipley Professor of Management, Chair of Strategy and Innovation, Questrom School of Business, Boston University), Rory MacFarquhur (Director for Global Economic Policy, Google), and Daniel Spulber (Elinor Hobbs Distinguished Professor of International Business and Professor of Strategy, Kellogg School of Management, Northwestern University). The panel considered, from a law and economics perspective, whether current IP laws sufficiently incentivize innovation in AI.
Daniel Spulber argued that inventors must be permitted to obtain patent protection for AI-related inventions. In Spulber’s view, § 101 jurisprudence can make it difficult for inventors to secure patent protection on software-related inventions, AI-related inventions therefore should not be subject to the same scrutiny under § 101.
Iain Cockburn offered a somewhat different perspective, reasoning that because AI technology is so new, it is currently difficult to determine whether the current U.S. IP system will appropriately incentivize AI innovation. Cockburn argued that initial patent protection for both DNA and business methods was overly broad, which risked hindering innovation in those fields by creating an “anticommons” where multiple rights-holders with overlapping patent rights can often exclude each other and others from engaging in valuable future research. In the case of DNA, there was an initial rush of patent applications filed based on findings that were said to be relatively routine. The USPTO subsequently increased scrutiny of these applications under the utility requirement. Second, in the case of business method patents, inventors initially obtained overly broad patent protection on, for instance, implementing known business practices via a web browser. If patenting for AI-related inventions turns out to be similarly overly broad, a similar correction may be needed.
Global Perspectives: AI and IP Policies Around the World
Susan Allen also moderated a panel discussing the global perspectives of AI and IP policies around the world. The panel included Qian Wang (Professor of Law, East China University of Political Science and Law), Alain Strowel (Professor, University of Louvain, Belgium), Matsuo Nonaka (Senior Director, Patent Examination Department), and Daryl Lim (Professor of Law and Director, Center for Intellectual Property, Information & Privacy Law at John Marshall Law School), who respectively offered Chinese, European, Japanese and Singaporean perspectives.
The European Patent Office issued new guidelines that cover artificial intelligence and machine learning. As we explained in a previous post, the EPO Guidelines state that artificial intelligence and machine learning “are based on computational models and algorithms” that “are per se of an abstract mathematical nature.” To qualify for a patent, an invention must first have a “technical character,” in that it relates to a technical field and is concerned with a technical problem. The guidelines provide examples of AI inventions which include the requisite technical character, such as the use of a neural network in a heart-monitoring apparatus for the purpose of identifying irregular heartbeats.
The UK Copyright, Designs and Patents Act defines “computer-generated” works as those “generated by computer in circumstances such that there is no human author of the work.” The Act also describes the effect on authorship: “the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken.”
The patentability of AI-generated inventions has not yet been strongly debated in China, but will possibly intensify as patent applications are filed that are based solely on AI. Mr. Wang stated that there are no known Chinese patent applications directed to an invention created solely by an AI independent from human activity. Whether such applications will be allowed is in doubt. Chinese copyright law excludes machine-generated works from copyright protection, which is hotly debated. Professor Wang expressed support for such an exclusion because copyright law is intended to incentivize innovation, and humans—not AIs—are incentivized by the enticement of exclusive rights.
Mr. Nonaka explained that Japan has not reached a conclusion as to whether AI-generated works should receive patent protection, and theta further discussion is necessary as the technology evolves. The Japanese Patent Office has recently provided, in both Japanese and English, case examples pertinent to AI-related technologies.
Professor Lim offered Singapore as an example of what a small country can achieve with the right mix of policy, politics, and the “X-factor.” Mr. Lim described Singapore’s IP policies as middle-of-the-road (i.e., not maximalist or minimalist) and argued that such policies have appeared to help Singapore thrive as a country over the past 50 years. Notably, Lim explained that Singapore does not have an expansive subject matter eligibility threshold like in the United States. Rather, Lim opined that Singapore “let[s] 1000 flowers bloom.” Lim argued that Singapore’s lack of an expansive subject matter eligibility threshold is advantageous because patents attract the excitement and investment that’s necessary to stimulate innovation. Singapore released a framework for ethical use of AI, which recognizes that AI will only reach its full potential if the public trusts it.