Remarks delivered at the Artificial Intelligence & Big Data Innovation
Director of the U.S. Patent and Trademark Office Andrei Iancu
November 15, 2018
Thank you, Judge Porter, for your kind introduction. And good afternoon, everyone. It’s an honor to be here with all of you at this outstanding annual event. I’d like to thank the board, the officers, and the staff of the Federalist Society—especially Dean Reuter—for the invitation, and for organizing today’s panel discussion on the future of big data and artificial intelligence (AI).
It’s an honor to be here with all the panelists, and I’m especially honored to share the stage with my immediate predecessor, Michelle Lee, who—among many other things—initiated a Big Data program that not only supports what we currently do at the USPTO, but also serves as the foundation for future AI development at the agency.
As Judge Porter noted in his introduction, I have the honor of leading the USPTO, where nearly 13,000 employees—including some 9,000 examiners—work tirelessly, every day, to secure the intellectual property rights of inventors and brand owners.
As you can imagine, the USPTO has vast reserves of scientific data contained in the more than 1 million patent and trademark applications we receive every year. Indeed, our patent, trademark, and other related data are among our most important assets. Many depend on our data. Whether it’s an independent inventor working in her garage or lab to better understand her innovation compared to the existing landscape, or a large multi-national corporation considering whether to invest in developing new strategic technologies, creating new brands, or acquiring existing patent portfolios, just about everyone uses the USPTO’s data to keep up with the pace of innovation and entrepreneurship.
We, too, as an agency, are leveraging this invaluable data so we can more efficiently and effectively fulfill our constitutional mandate “to promote the progress of science and useful arts.” Let me share with you just a few ways we’re doing this.
At the USPTO, we have a “Big Data Reservoir” that contains over 8 million patent office actions. This empowers us to harness data to measure work product consistency across our entire patent corps and systematically focus our quality improvement efforts.
For instance, our “Big Data Reservoir” has enabled us to answer fundamental questions such as: How many—and what types of —§101 rejections are our examiners making and consistently applying throughout the examination corps? How can examiners more effectively use non-patent literature in prior art rejections? And what impact has our guidance and training had on examination outcomes?
Efforts like these, as well as other patent quality studies, have resulted in re-allocating millions of dollars in training expenses to more localized areas for optimal rate of return.
Moreover, by identifying how and what prior art is being used by our examiners and comparing that to, for example, the outcomes of AIA trials before the Patent Trial and Appeal Board (PTAB), we can begin to measure and quantify the accuracy of the searches we conduct during examination, as compared to the art an opponent might find during a dispute after issuance.
Enhancing search is one area, in particular, that we expect AI could yield tremendous returns. Indeed, it’s a tool we hope can help us narrow the gap between the search done during examination and the search done post-issuance.
To that end, we’ve developed—and are actively testing—a new ”cognitive assistant” called “U” or “Unity,” which leverages AI and machine learning in a way that augments our existing next-generation patent tools. For example, the tool is intended to allow patent examiners, through a single-click, to conduct a “federated search” across patents, publications, non-patent literature and images. And, through AI and machine learning-based algorithms, this would present to the examiner the results in the form of a “pre-search” report.
We’re also exploring semi-automated tools for “search query expansion,” trained to mine technology-specific synonyms with the help of crowd- or “examiner-sourcing.” This new capability holds the potential to promote consistency in searching and to more quickly surface prior art that may be located in any of several disparate databases.
And that’s important, because one of the benchmarks of a high-quality patent is whether it can withstand fair challenges down the road. Surfacing the best prior art early helps to increase the likelihood that this will happen. AI can help us do that.
We’re also testing new AI tools and techniques such as robotic process automation (RPA) that could potentially generate smart office action templates that are automatically populated based on the interactions between examiner and attorney, saving our examiners time from some of the more tedious clerical tasks when generating office actions.
And, in an effort to reduce the costs of manually classifying patents, we’re exploring the use of AI technology to ensure that we route the “right case to the right examiner.” This, in turn, enables us to organize our workforce more effectively and as a result conduct more effective examinations.
So, these are just a few of the ways we’re using Big Data and AI within the institutional walls of the USPTO. There are, of course, many other AI efforts underway at the USPTO, including the potential of engagement with industry to help us identify the most advanced search tools.
Outside of our agency, AI has significant implications for the law, the economy, and America’s position as the global innovation leader. Not surprisingly, AI is changing the landscape of intellectual property policy, and in doing so, it’s raising real legal, regulatory, ethical, and moral questions for us to grapple with.
I am sure other panelists will address many of these issues in their remarks. Some IP-related examples include: Will the legal concepts of inventor, author, and creator be fundamentally changed by AI? Does use of copyrighted works to “train” AI systems constitute fair use? How will firms, both large and small, protect AI-related inventions and how does patent subject matter eligibility impact those strategies? What are the disclosure requirements in a patent for a machine-learning algorithm, when the human inventor may not know exactly how the machine will perform a given task after it has learned from training?
Such questions cut across industrial sectors and national boundaries, and many do not have viable answers yet. But how we choose to answer them will have major national economic implications. The good news is that we are working on them as we speak.
In fact, these and similar issues will be examined at an all-day conference called “Artificial Intelligence: Intellectual Property Policy Considerations” organized by the USPTO, to be held on December 5. Please join us.
Needless to say, AI has evolved from the obscure to the mainstream, and it’s taking the use of computers to a new level, at an awe-inspiring speed.
Some have even characterized this fusion of technologies that blur the lines between the physical, digital, and biological spheres as a “Fourth Industrial Revolution.” As with prior industrial revolutions, these new technologies, which include robotics, autonomous vehicles and quantum computing, among many others, hold the promise to improve and lengthen lives, generate higher income levels, dramatically increase productivity and efficiency and, importantly, vastly increase the speed of innovation itself. But they also pose substantial risks, particularly if the United States is left behind in the innovation race.
Countries around the world are adopting and implementing long-term, comprehensive strategies designed to increase their prominence and leadership in innovation. For example, the “Made in China 2025” initiative is aimed at transforming China into a global leader in strategic industries, such as AI and 5G, that are critical to competitiveness and innovation in the Fourth Industrial Revolution.
In recent years, there has been dramatic growth in Chinese patent filings in such key technologies as next generation information technology, computerized numerical control and robotics, and advanced transportation. When we look at patent applications in the technology areas largely covered by the “Made in China 2025” initiative, we see that filings by Chinese nationals to China’s IP office have grown at an annual rate of 24 percent between 2006 and 2016. By comparison, such applications filed by U.S. nationals to USPTO grew at an annual rate of only 3 percent. Patent filings are not fully determinative of innovation, and some have questioned the quality of some Chinese applications, but these statistics are one measure and a potential leading indicator. Some other indicators point to similar trends.
And China is not the only innovator in the technologies of the Fourth Industrial Revolution. From the smallest countries like Singapore to the largest like China, many nations around the world have become extremely competitive in the innovation ecosphere.
Only by innovating faster and in key areas will the United States continue to lead. We must harness our long history of innovation, born of our nation’s founding document and perpetuated by our people’s innovative spirit since then, and apply the same spirit to this new Fourth Industrial Revolution.
As director of the USPTO, one of my top priorities is making sure the United States remains the market of choice when it comes to innovation, especially in the emerging technologies of the future, including AI and machine-learning technologies.
This administration is committed to protecting and promoting American innovation and entrepreneurship, as symbolized by President Trump’s recent signature of Patent 10 Million. This was only the second time a president signed a patent since John Quincy Adams, and it evidences the importance of American intellectual property in today’s economy and the Administration’s commitment to it.
Thank you again for the invitation to speak on this important topic, and I look forward to continuing the conversation with all of you during the panel discussion.