APAA e-Newsletter (Issue No. 25, October 2021)

Is My Smartphone Smart? Australian Court Decides AI Can Be an Inventor

Paul Warden-Hutton, Houlihan² (Australia)

The Decision by the Australian Federal Court in Thaler vs. The Commissioner of Patents [2021] FCA 879

Probably the most important detail of this decision is that it is being appealed. In relation to the appeal, the Commissioner of IP Australia has stated “…that the legislation is incompatible with permitting an AI to be an inventor… .”

While both Australia and South Africa have recognised that an Artificial Intelligence (AI) can be an inventor, the South African case can be distinguished because there was no substantive assessment of the named inventor. In Australia, the Patent Office issued a decision stating that Australian Patent Law “…is inconsistent with an artificial intelligence machine being treated as an inventor.” A Federal Court decision then set aside this decision and rejected it in favour of deciding that an AI system can be an inventor. Significantly, the Australian decision itself is concerned with examining only formalities matters regarding whether, or not, an inventor was named. The answer was yes, in that an AI entity was named as the inventor.

While the decision was chiefly concerned with a formalities matter, the Australian judge also considered the provisions in the Australian patent law concerning who can own a patent and definitions for the terms “invention” and “patent”. Interestingly, Australia’s patent law does not provide a definition for the term “inventor”.

The Australian judge also commented that there are no specific provisions in Australian patent law that expressly prohibit an AI from being an inventor and that there is also no specific aspect of patent law, unlike copyright law, involving the requirement for a human author. The judge also posited that while other Australian judges have commented that the concept of patentable subject matter must be widened in the face of modern technology, there is no reason the same flexibility should not be afforded the concept of “inventor”. Consideration was also given to the Australian patent law not being focussed on the inventor, but instead being focussed on inventive step.

The invention

To determine inventorship, the contribution to the invention has to be assessed. To do that, we have to first be familiar with the invention. Australian patent application number 2019363177 is entitled: “Food container and devices and methods for attracting enhanced attention”. The Australian application matured from International patent application number PCT/IB2019/057809. While there are various aspects to the invention, claim 1 is directed to a food or beverage container that has a wall with a fractal profile with corresponding convex and concave fractal elements on corresponding interior and exterior surfaces wherein the convex and concave fractal elements form pits and bulges and the fractal profile of the wall permits coupling by inter-engagement of a plurality of said containers together.

This appears to mean that an AI has designed interlocking food containers that are easy for robots to grasp.

Is my smartphone smart?

We use the term “smartphone” to distinguish from a “feature phone”. A smartphone has stronger computing functions, provided by hardware, which run software to permit internet access and multimedia functionality. However, does this make a phone “smart”?

AI is commonplace in our daily lives. Internet searches, virtual assistants, like Siri and Alexa, and recommendation systems, like those used by Netflix to suggest programs, all use AI. AI relies on a dataset, from which it learns how to provide a result. The AI also requires an algorithm, which comprises the series of steps used to arrive at the result. The AI receives an input to produce the output, which is the result. For example, if you want to be able to determine if a food is Japanese or Korean, you provide a dataset with photos of as many Korean foods as possible and as many Japanese foods as possible. You provide the computer with a series of steps that allow the classification of each photo as being either Korean or Japanese. Then, when you come across a new food and do not know whether it is Korean or Japanese, you take a photo and input it to the computer. The computer will then provide an answer. However, does that make AI intelligent?

Technological convergence

The invention in the relevant patent specification provides a result in the area of food and beverage packaging. As a graduate student, the writer studied proteins and how they fold into their various three-dimensional structures. Protein structure is important because proteins are the tools, machinery and structural building blocks of plants and animals. Solving a protein structure was at one time a great achievement requiring weeks, months or years of experimental work. The impact might have been as significant as sequencing a new gene.

Today, AlphaFold, an AI system developed by DeepMind, a company owned by Google, can solve a protein structure in seconds and regularly with an accuracy competitive with that obtained through experiment. What this means is that with digital data available, or able to be produced, for almost all areas of human endeavour, we will see AI applied to an increasingly wide array of technologies.

If your current practise does not yet include AI inventions (or inventors), it soon might.