Remarks delivered at Trust, But Verify: Informational Challenges Surrounding AI-Enabled Clinical Decision Software
Deputy Director of the U.S. Patent and Trademark Office Laura Peter
January 23, 2020
As prepared for delivery
Good afternoon! And thank you, Arti, for your warm welcome. I am delighted to be here today to discuss the hot topic of intellectual property (IP) and artificial intelligence (AI).
Years ago, the great American philosopher Ralph Waldo Emerson said, “Do not go where the path may lead; go instead where there is no path and leave a trail.” Today, this statement rings true as we forge ahead into the uncharted territory of artificial intelligence technologies. This afternoon, I would like to take the opportunity to share some thoughts about the direction of AI and the vision for our path forward.
First, thank you to the Center for Innovation Policy at Duke Law for your leadership in this field and for organizing today’s conference. AI technology innovation has accelerated and changed over the years with the advent of faster processors and more database availability. New AI developments across a wide range of industries are being integrated seamlessly into our daily lives. Machine learning is being used in areas including healthcare, financial services, home personal assistants, and retail marketing, while new discoveries in autonomous vehicles, robotics, computing, and gaming are shaping the world we live in today and in the future.
Without the continuous innovations we have seen in AI, none of these advances would be feasible. As with all trailblazing of new frontiers, there is much to be excited about, as well as much to give one pause.
The latest news in mainstream media touts how artificial intelligence technology can help doctors read mammograms more accurately. Using the artificial intelligence tool, false positive rates of malignancies were reduced by nearly 6%. False negatives were lowered by more than 9% in the U.S. As artificial intelligence tools in the health arena become more sophisticated, more lines will be affected positively.
We at the United States Patent and Trademark Office (USPTO) are at the forefront of intellectual property protection of artificial intelligence innovations. We are on this journey with you. Today, AI and related technologies rapidly advance and pose more numerous and more challenging questions for the bar, industry, patients, and government. These questions are at the forefront of the public debate and today’s news headlines and are on the minds of venture capitalists and investors. Questions such as: “Can an entity other than a natural person be an inventor on a U.S. patent?”
Let me take a quick survey.
How many of you in the audience believe that an AI machine that contributes to the conception of an invention should be allowed to be an inventor on a patent?
Second question: How many of you think that an AI machine that contributes to the conception of a drug discovery process should be allowed to be an inventor on a patent?
One such question in the USPTO Request for Comments on the patentability of AI inventions is:
Can an AI machine be the inventor of an invention described in a U.S. patent?
Interestingly, this past December, both the European Patent Office (EPO) and the United Kingdom Intellectual Property Office (UKIPO) refused two patent applications in which a machine, “DABUS,” was designated as the inventor.
The patents were refused on the grounds that they do not meet the requirement of the European Patent Convention that an inventor designated in the application is a human being, not a machine. So far, the answer is no, according to the EPO and the UKIPO. The same issues face us at the USPTO.
While the USPTO cannot comment on pending applications, it has been reported that a similar application is pending in the U.S. The urgency is not lost on government policymakers. This afternoon, I want to suggest a path forward, with the USPTO as the partner to forge ahead together. Your input is critical to this dialogue.
Many of the cutting-edge questions around intellectual property and artificial intelligence technologies are undecided matters leading to uncertainty. Currently, the U.S. legal system does not have easy answers to many challenging questions raised by AI. Likewise, copyright, trademark, and trade secrets are similarly challenged. Some propose that the current system can handle these issues, while others say we need to create new forms of intellectual property. So how do we protect and incentivize the creations of AI innovations? This is where the United States Patent and Trademark Office turns to you.
Last August, we published a Request for Comments related to the impact of artificial intelligence technologies on the patent system. Then, in October, we issued a second AI-related Federal Register Notice requesting comments on the fields of copyright, trademark, database protections, and trade secret law. Both comment periods are now closed.
It was very exciting to see the range of the responses to our Federal Register Notice. We heard from more than 90 groups, including many U.S. stakeholders. We also heard from a great number of international organizations, including the EPO, the Japan Patent Office (JPO), and Korean trade associations. At the USPTO, we are in the process of reviewing the compelling insights and compiling a report that should be published this spring.
So, to whom should authorship and rights be granted? The artificial intelligence machine? The person who set the policies that dictate how the machine thinks? The person who set the parameters of the database used to train the AI machine? And what if the AI’s decisions that led to an invention cannot be explained? How can we grant a patent to something if there is little control and minimal understanding of the process, as required by current patent law?
Also, to receive a patent, the inventors are required to clearly explain the process to make the invention so others can learn from, replicate, and improve on the method and tech. These new AI technologies demand careful consideration in light of current intellectual property laws .It is essential that the United States be at the forefront of strong IP protection for AI technologies to incentivize and accelerate innovation in the U.S. If we fail to do so, other nations will move ahead without us. These are the difficult questions we need to be asking.
The dialogue with industry experts like you about the questions and challenges of AI is essential, and we look forward to continuing our productive dialogue. I want to highlight some of the USPTO’s forthcoming AI milestones.
In the next month or so, we will launch an internet portal concerning all intellectual property-related initiatives and content on artificial intelligence technologies, including Federal Register Notices, research and reports, and news stories. This will offer valuable content in one place for your convenience.
Later this spring, we will release our report summarizing the responses to the Federal Register Notices, which may include relevant conclusions. Also later this spring, our Office of the Chief Economist will release a report on the patent landscape of artificial intelligence-related inventions. This report will help explain patenting activity, such as grants, across a number of technology fields.
And finally, we may have an opportunity to further discuss patent applications on inventions for which a machine is claimed to be an inventor. My request to you: Please continue to think critically about artificial intelligence technologies and invention—most importantly, what you think it truly means to be an inventor. Please continue this dialogue, and help us blaze the trail for the future.
Thank you so much for inviting me to join you today.