In a three part series [1], Mark Vanderveken and I look at how Artificial Intelligence (“AI”) could change the nature and scope of intellectual property (“IP”) rights. Part 1 will look at what exactly is AI.  Part 2 will examine the impact AI may have on legal frameworks governing IP rights. Finally, Part 3 will consider whether AI based technology could be an author or inventor of copyrighted works or patentable inventions.

Part 1: What is Artificial Intelligence?

“Sometime early in this century the intelligence of machines will exceed that of humans. Within a quarter of a century, machines will exhibit the full range of human intellect, emotions and skills, ranging from musical and other creative aptitudes to physical movement. They will claim to have feelings and, unlike today’s virtual personalities, will be very convincing when they tell us so.”

Ray Kurzweil, The Coming Merging of Mind and Machine (Scientific American 2009)

Coined in the mid-twentieth century, “artificial intelligence” (“AI”) is an umbrella term to describe a computer science sub-discipline, which include a variety of related technologies. Countries around the world are racing to invest in the development of AI-based technologies so as to secure continued economic growth. Public and private investments in research and development of AI-based technologies have surged in Canada and the U.S. A recent article by The Brookings Institution pegged equity investments in U.S.-based AI companies at over $25 billion (USD) in 2019, while Canadian AI companies attracted approximately $885 million (USD).

What many of us typically think of as artificial intelligence derives from popular culture; think about the benevolent C-3PO in Star Wars and homicidal systems embodied in the “Terminator” and “H.A.L.” As represented in popular culture, this “general” AI can be thought of as machines or systems capable of carrying out higher cognitive functions in a manner that appears similar to the human intelligence. Current systems have yet to achieve this level of sophistication, however. Current AI based technology is directed to “narrow” or specific applications. Such “narrow” AI applications include successfully understanding human language (verbal and written), image recognition, competing at a high level in strategic game systems (e.g. chess, JEOPARDY and GO), autonomous cars, military simulations, interpreting complex data (e.g. images and videos), etc. While not as fully functioning as general AI, narrow AI based technologies have impacted and will continue to impact our daily lives.

One of the key functions of AI related technology is that it can make inferences in an attempt to accurately predict outcomes based on past outcomes or training data, and improve (i.e. re-train itself) by continuous revision (so called machine learning). AI related technologies can, for the purposes of this article, be broken down into two components: (a) the “Process” wherein engineers/computer scientists develop AI platforms or “engines” to make the predictions; and (b) the “Output” wherein researchers select and train AI tools on a suitable data set to make specific predictions in a particular area. These will be considered in greater detail in Parts 2 and 3.

As AI-based technologies gain in capability, such technologies will play an ever-increasing role in the development of IP. In early 2020, the drug baricitinib was identified by an AI system as a potential treatment for 2019- nCoV acute respiratory disease (see The Lancet, February 4th, 2020). This drug is now being studied as a possible treatment for COVID-19 infections. It is clear that both the Process and the Output may give rise to innovations that are protectable IP, and both the Process and the Output may also result in transgression of IP rights.

In December 1998, the United States Patent and Trademark Office (“USPTO”) issued a patent entitled “Neural Network Based Prototyping System and Method” (U.S. Patent No. 5,852,815) to Stephen Thaler, a computer scientist who developed an early example of AI called the “creativity machine” based on a self-stimulating artificial neural network. According to Thaler, the creativity machine invented the subject matter of this patent. Another example comes from software employing “genetic programming”, whereby the natural processes of biological evolution are simulated computationally. Genetic programming is the core technology in John Koza’s “invention machine”, to which he has attributed the invention behind at least one U.S. patent, which issued in 2005 (U.S. Patent No. 6,847,851). In both of these cases, the role of proto-AI systems in the creation of the inventions appears not to have been disclosed to the USPTO during prosecution of the applications.

A more recent example is IBM’s “Watson” AI system, which is touted as being capable of “computational creativity” in generating and evaluating ideas by processing vast amounts of data. While it does not appear that any of Watson’s outputs have been patented, it is possible that at least some of Watson’s outputs could represent new, useful, and non-obvious inventions.

Much has been written about the disruptive effect of AI based technology on a number of industries. The disruptive effect of AI will likely have a significant impact on the nature and scope of IP rights. This article discusses the potential impact that AI may have on the legal frameworks governing the protection and enforcement of IP rights. The rise of AI related technology will impact what can be protected by IP rights, how those rights will be obtained and how those rights could be enforced and against whom. Some relevant questions include:

• Have we reached a stage where AI related activities could give rise to liability for a breach of intellectual property rights?
• Could AI provide additional insight into whether there is a likelihood of confusion when conducting a traditional trademark confusion analysis or whether an invention is obvious in view of the prior art?
• Have we reached a stage in the evolution of AI where the “machines” could independently create original and/or patentable material?

In the near future, IP regimes will need to recognize and adjust to the answers raised by these questions. As AI begins to mimic aspects of human endeavour that have been traditionally restricted to the hallmarks of human activity, namely creativity and reasoning, it is interesting to consider at what point a computer crosses the line from mere tool to inventor or author.

Conclusion

As AI-based technologies become more ubiquitous and gain in capability, it is likely that they will play an ever greater role in determining the nature and scope of IP rights.

The continued integration of AI into commercial products and services raises questions about ownership of its outputs and liability for infringement of third party IP rights. Stay tuned for Part 2 of our series where we will examine the impact AI may have on legal frameworks governing IP rights.

Once we have reached the point where AI-based technologies can meet the requirements of conception and reduction to practice (some would argue they already have), it is unclear how current IP legal frameworks in the U.S. and Canada will deal with and adapt to the reality of inventions and works of authorship created by non-humans. In Part 3 of our series, we will consider whether AI based technology could be an author or inventor of copyrighted works or patentable inventions.

Fasken’s experienced team of IP lawyers, patent agents, and trademark agents are available to advise on and assist with all manner of IP matters including the protection and use of AI technologies. For more information, please contact the authors or visit our group practice page.