Remarks delivered at the AI and Patent Data Workshop
Hosted by Carnegie Mellon University
Deputy Director of the U.S. Patent and Trademark Office Laura Peter
December 9, 2020
Virtual
As prepared for delivery
Hello everyone, and thank you, Dean Alderucci for that kind introduction, and thank you to both you and Sean Tu for organizing this wonderful event. It is an honor to be here today, speaking among such extraordinary academic, public, and private sector leaders. I am especially thrilled to be speaking at a Carnegie Mellon University (CMU) hosted event—as CMU has positioned itself as a leader in many technologies of the future, including software engineering, artificial intelligence (AI), robotics, manufacturing, energy, and autonomous transportation.
Pittsburgh as a whole, whose iron and steel industries were once so vital to our national economy, is today one of the world’s premier cities for research. In fact, according to the Pennsylvania Center for Workforce Information and Analysis, in 2018, the number of Pittsburgh area private-sector jobs in the scientific and research and development (R&D) sectors, excluding academic positions, exceeded those in the mills that were once the lifeblood of the economy, by 41%. In addition, 59 artificial intelligence startups call Pittsburgh home. Larger companies like Google, Uber, and Facebook have also all made great investments in the region.
Indeed, Pittsburgh has really become a hub of innovation for AI, and this cooperation and vibrant partnership between industry, academia, and government, demonstrated in Pittsburgh, will continue to be critical to the development of all innovation ecosystems. This is why the AI patent data conference we are participating in today is so important.
Jeff Bezos remarked a few years back, “We’re at the beginning of a golden age of AI. Recent advancements have already led to invention that previously lived in the realm of science fiction—and we’ve only scratched the surface of what’s possible.” I agree with him.
Artificial intelligence is now a pervasive force that is transforming the global economy. In fact, earlier this year our Office of the Chief Economist studied the millions of patent applications that have come through our office since 1976 to determine the distribution of AI technology in U.S. Patents. What we found was astounding.
The subsequent report published in October, titled, “Inventing AI,” found that the number of patent applications received annually with AI subject matter more than doubled from 2002 to 2018, from 30,000 to 60,000 applications. Also, a whopping 42% of our technology classes contained AI-related subject matter. This is a clear indication of both the importance of the technology and of how much it has permeated our society.
As the world adopts 5G, as sensors and transmitters continue to be embedded into every known product, and as quantum computing is commercialized, this number will only rise. Currently, we see AI being utilized in biotech to help create drugs and vaccines at lightning speed.
We have reached an age where AI is now required for almost all companies that conduct business on a national or global scale. The same can be said for intellectual property (IP).
Patents and the need for strong IP portfolios are becoming increasingly important, even in areas beyond traditional technology sectors. Companies, universities, and government entities alike can all have portfolios with hundreds or even thousands of patents. And, their IP portfolios are often key to their success.
As all of us here well know, for each one of these patents, an extensive search of mountains of prior art must be performed, both on the side of the applicant and in the examining office for each pending application. Also, data needs to be sifted through for due diligence in R&D investments, licensing determinations, and during litigation, including infringement cases and the like.
As you may be aware, the 2020 U.S. Chamber of Commerce’s Global Innovation Policy Center’s (GIPC) International IP Index ranked the U.S. as the second strongest intellectual property system in the world. Singapore holds the number one spot overall, but for trademarks and copyrights, we rank number one. For the U.S. to remain a technological leader, we must take an active role in protecting and fostering the creation of future technologies.
This issue is of great importance to the current Administration and Congress. Last year, the President signed an Executive Order on Maintaining American Leadership in Artificial Intelligence.
The order noted that “Continued American leadership in Artificial Intelligence is of paramount importance to maintaining the economic and national security of the United States,” and the American Artificial Intelligence Initiative was launched. This past week, on December 3, another Executive Order was signed entitled, "Promoting the Use of Trustworthy AI in the Federal Government."
During the past two years, from the beginning of this charge, the United States Patent and Trademark Office (USPTO) has actively engaged with other government partners to keep pace with the thousands of applications that pass by our examiners every day.
Use of AI tools was critical in developing the Economist’s report I mentioned earlier. The investigation of millions of patents could not have been performed without using AI. Artificial intelligence tools are also being deployed within our Patents and Trademarks business units to assist in the production process. I will give a quick summary of these efforts, but you will hear more on them during the panel discussion from the USPTO’s Technology Center Director, Matthew Such.
In 2020, we released two requests for information regarding commercially available AI tools that might assist in patent examination and search. The first, in March, sought software products that can assist applicants with improving the quality of patent applications at or around the time of filing. The second, in September, sought AI software products to assist with image searching in the context of design applications. At this time, review of these submissions is still underway, but I can report that the response has revealed very promising technologies that continue to grow in sophistication.
Separately, we are also in the process of deploying AI tools for our utility patent and trademark examiners. So many hours of examiner’s efforts are lost weeding out irrelevant prior art and zeroing in on relevant art instead of the sophisticated intellectual analysis of what is patentable under intellectual property laws. AI tools might provide relief from this mundane work. We believe that these technologies will create a more positive work environment for all our staff, and also improve stakeholder satisfaction.
Back before the time of computer searching, patent examiners would search through paper card catalogues that we at the USPTO called the “shoes.” Computer searching changed all that, and the sophistication and scope of prior art searching exploded. We foresee that AI tools will bring more efficient and high-quality searches as our examiners sift through the exponentially growing databases of art. This in turn will lead to more stable, reliable, and predictable IP rights.
In fiscal year (FY) 2020, we evaluated an AI prototype for utility patent searches that not only helps identify potentially relevant art, but also takes the next step by making suggestions for other areas of search. These capabilities are now being integrated into next generation patent examiner search tools, and we plan to deploy them during the current fiscal year.
In addition, I am pleased to announce that this month we launched our initial deployment of an AI-based auto-classification system. This is a huge step, and begins automation of assignment of technology classification numbers to new applications.
On the Trademarks side of the house, we have two big projects in the works. Trademark filings continue to break records and, in fact, September was our biggest filing month in history for trademarks. We are motivated to bring new technologies in to help bear the caseload and to make our processes more efficient and cost-effective.
These include an Image Search tool and an Automated Specimen Analysis Project for detecting doctored specimens. Both are high priorities because they will help to counteract fraudulent filings.
The search prototypes compare trademark image logos and the like, suggest the correct assignment of mark image design codes, and identify acceptable identifications of goods and services. We began beta testing in November.
I am pleased to report that the tool for detecting fraudulent specimens is being implemented in full production this month. It is able to detect images that have been manipulated, and we will be using it to screen the images that registrants need to submit showing their products or services that are used with the trademark.
Let me state this clearly: even though we are incorporating AI tools into our operations, we do not foresee AI replacing our examiners, or in making decisions as to what is patentable or registerable as a trademark. The human ability to interpret drawings and nuanced information associated with a patent invention disclosure, and to determine the appropriate legal protections for both patents and trademarks, is not replaceable by machine. So, as always, the best way forward is through high quality examination by our examiners, and we continue to encourage our stakeholders and examiners to retain open lines of communication throughout the prosecution process.
On a related note, the USPTO has recently confronted the issue of whether a patent can be awarded where the inventor of an invention is claimed to be a machine. Recent patent applications have claimed that the invention was created by computers trained in techniques of artificial intelligence, and without any obvious human intervention. So, earlier this year, we sought the public’s input into how to handle patent applications for inventions created by AI.
We found widespread support for our determination not to grant patents to computer programs or machines. Only a human is responsible for invention. Human sentience provides a machine with the ability to create new products, and the same is true in patent examination matters. I might add that IP offices around the world share this view as well. AI is quickly evolving, and humans are directing its course.
As we continue to navigate new territory, not only with respect to technology implementation into our operations, but in almost every aspect of our lives as we have seen during the current pandemic, we must make special effort to incentivize innovation, and ensure our IP system is equipped with the tools that will enable us to continue this fight. It is the protection of inventors, provided by the U.S. patent system, that precipitates dynamic competition, and that creates the industries which have revolutionized global commerce.
When I think of the advancements that are sure to come in both our patent system and society as a whole with the advancements of AI technology, I am reminded of what Vint Certf, who is known as one of the “fathers of the internet” and one of the co-inventors of internet-protocol, once said. He said, “AI and natural language processing will surely make the internet far more useful than it is today.” He also wisely summarized, “At some point, you can’t lift this boulder with just your own strength. If you need to move a bigger and bigger boulder up hills, you will need more and more help.”
Similarly, implementing the right AI tools in our operations and patent processing tasks, especially those that rely on reading and assessing the meaning of hybrid legal and technical documents, are such a hill.
The USPTO is proud to be a global leader in the dissemination of Open Data for patents, enabling much of the groundbreaking research you will hear about later today. However, we cannot climb to the top of this hill without coming together with our partners in academia, industry, and the rest of the government, to determine the next steps forward.
We are eager to engage everyone here today at this conference in both the academic and data science communities in defining our next-generation suite of AI capabilities, and as we continue this journey. For as glorious as our past has been, I believe our future will be even brighter, as we stand on the brink of truly remarkable technological and scientific advancements, including AI.
Thank you for inviting me here today, and I look forward to our continued dialogue.