Artificial intelligence (AI) technologies are poised to further disrupt a wide-ranging span of industries, not least of which includes medical care and research. There are already patents granted for AI technologies directed at drug discovery, disease diagnosis, medical imaging, and precision medicine, among many others.
Innovations in machine learning and deep learning have helped propel these advances, and research entities, health care institutions, and innovative companies are seeking to harness the power of AI to ameliorate patient health and care and produce medical and basic science research faster and with new insights.
Patent protection is an early and critical consideration for many of these entities. It is an open question whether intellectual property law sufficiently accounts for all the aspects of AI or must adapt in view of this rapidly evolving and uniquely complex field of computer science.
To probe this issue especially as it relates to patent protection, the United States Patent and Trademark Office (USPTO) posed twelve questions to public stakeholders in a request for comment published in the U.S. Federal Register this past August. The purpose was to help determine whether the USPTO should issue new guidance on how it will examine patent applications related to AI inventions and/or whether new forms of intellectual property protection should be developed for AI inventions. In its request for comment, the USPTO shows an appreciation that many of the factors involved in developing, training, and using AI technologies are unique amongst the other computer science disciplines and that it may be appropriate to accordingly define a uniquely tailored application of the law or develop new laws to protect AI technologies.
In its questions, the USPTO can be seen grappling with the issues of inventorship (can an AI technology contribute to the conception of an invention?), ownership (can parties involved in the training of the AI technology own an invention created by the AI?), the level of written description disclosure required in the patent application (how much detail must be disclosed for deep learning algorithms having hidden layers and evolving weights?), enablement, subject-matter eligibility, and other issues having legal implications.
Many of the questions posed have philosophical underpinnings. Traditionally in Canada, only humans are recognized under the law as capable of being inventors. However, AI technologies increasingly challenge the notion that some capabilities are exclusively “human”. What happens when an AI technology creates a new insight that forms a part of a new invention? Who should be the inventor when an AI technology identifies a new biomarker for diabetes based on characteristics of tissue samples that no doctor would have previously considered for diagnosing the disease or selecting a course of treatment? Can a person skilled in the art – the mythical [1] (human) creature typically deemed by law to be the addressee of patent documents – be defined by the knowledge and capabilities of a (non-human) AI entity? It stands to reason whether a “person” or “inventor” can or should be defined as including an AI technology.
This is also not the first time the USPTO has specifically considered the patenting of AI technologies, which you can read more about here.
A couple of months later in October, the USPTO issued a second request for comment published in the U.S. Federal Register. This second request for comment posed thirteen questions to the public relating to intellectual property protection for artificial intelligence innovation. More specifically, it probed the adequacy and ability of copyright, trademark, or trade secret law to protect AI innovation and for their regimes to survive new AI technology. There are questions relating to defining who the copyright author and owner are in works produced by or involving AI, as well as the protection and possible infringement of copyright relating to the use of data to train AI technology. The USPTO’s two requests for comment provided a unique opportunity for stakeholders developing or otherwise invested in AI innovation to shape how patent and other intellectual property law and policy can be used to protect AI.
The USPTO’s second request for comment appeared to echo some of the considerations recently reviewed by the Canadian government in its evaluation of Canadian copyright law.
The Canadian government recently completed its review of the Copyright Act – a review mandated by statute and having the stated purpose of evaluating the currency of Canada’s copyright regime in view of advances in digital technology. Notably, the resulting report by the Standing Committee on Industry, Science and Technology specifically considers works created by AI technology without human intervention and works made by humans using AI technology. The Standing Committee includes a recommendation that the Copyright Act be amended or new legislation be enacted to “provide clarity around the ownership of a computer-generated work”, as well as a recommendation that the Copyright Act be amended to “facilitate the use of a work or other subject-matter for the purpose of informational analysis”.
These recommendations were made specifically in view of facilitating the development of AI technology. For example, an exception to copyright protection for the use of data to train the AI technology could be permitted as being informational analysis. It remains to be seen whether the legislative regime will be amended in response to these recommendations.
It is an exciting time for AI, and if not already, one day soon we may be able to call on our virtual assistants – “Hey AI Bot” – to help solve our hardest questions with insights not previously understood by humans in healthcare and beyond.
[1]The touchstone point of reference under Canadian patent law remains a “mythical creature”: see e.g., Apotex Inc. v. Sanofi‑Synthelabo Canada Inc., [2008] 3 S.C.R. 265, 2008 SCC 61
by Larissa Leong