The Semantic Labyrinth: Navigating Patentability in the Era of AI

Introduction

The intellectual property law finds itself in constant catch up mode in the quickly advancing technological world. In the news of the last little while, the patentability of artificial intelligence comes under the spotlight through the legal saga surrounding Emotional Perception AI Ltd. (EPL). It makes for an interesting and sometimes bewildering look at this dispute of the struggle over patentability of artificial intelligence. It is much more than a technical dispute over law as it represents a landmark case which could relegate the historical geography of innovation as it is difficult to conceive what “invention” implies in terms of what exactly an invention is.

The crux of the issue directly pertains to a patent application by EPL of a system utilizing an artificial neural network (ANN) to suggest media files, including music files, based on a so-called semantic similarity as opposed to the mere genres or artists as metadata. Consider an application capable of recommending a classical music song to a person with an affection to heavy metal not on single basis of sharing a common designation, but having a common emotional intonation or complicated music contextualization. That is a generic, non-obvious functionality which was sought to be protected by EPL.

Having received the decision of the UK Intellectual Property Office (UKIPO), the High Court, and, lately, the Court of Appeal, the case stands as an exemplary guide to the use of the old law on the new technologies. The debate itself centres on a single, on paper quite innocent provision: Section 1(2) of the UK Patents Act 1977.

The Elephant in the Room

The Patents Act 1977 Section 1(2) is an imperative passageway to patent applications within the UK. It enumerates a variety of things, which are not inventions when it comes to the Act. Included in these exceptions, there are those that do not include a program of a computer as such, a mathematical method as such, a literary, dramatic, musical or an artwork, or any other aesthetic invention. This was the basis upon which the UKIPO had earlier on declined to grant the patent to EPL as a patent application that was deemed as hitting the computer program as such exclusion mark directly.

[Image Sources: Shutterstock]

According to UKIPO and subsequently the Court of Appeal, a trained ANN, however complex and good at self-learning, actually remains nothing more than a collection of instructions on how a computer should solve a problem. The weights and biases that the ANN is trained to represent were regarded to be equivalent to the code of a traditional computer program, such as a set of instructions which give direction to the computer programming. A judgment of the Court of Appeal given by Lord Justice Birss expressly made the point that the argument that ANNs were almost unique was mistaken and that a computer program just meant a set of instructions to make a computer do something.

Aerotel and the Technical Contribution

This is the point where the legal action really becomes exciting, because it deals with a well-practised judicial test in action. Their method in the course of assessing the patentability of computer-implemented inventions follows a 4-step regime commonly known as the Aerotel/Macrossan test that was instituted in the case of Aerotel Ltd v Telco Holdings Ltd [2007] RPC 7. These steps include:

1. Correctly interpret the allegation: What does the invention really do?

2. Locate real contribution: What is inventive concept?

3. Inquire as to whether the identified contribution only includes the excluded matter: Is the invention a computer program, or a mathematical method, etc.?

4. Determine the actual nature of the contribution: is the contribution a technical one by solving a technical problem in a non-obvious manner?

The question in the case of EPL turned out to be: had their invention, though perhaps presumably a computer program, had a technical contribution which put it beyond the exclusion? This is the statutory get-out of computer implemented inventions. A simple computer program which simplifies a business task is not patentable, however, a program which causes a computer to work with greater efficiency or one which provides feedback into a non-computer based technical task can be patented.

The history of the concept of the technical contribution as a concept that has passed legal barriers is long. The case of a digital image processing method is a landmark precedent in the case of European Patent Office (EPO) Board of Appeal, T 208/84, VICOM Systems Inc. The EPO considered that a method of mathematics is not patentable but that one that achieves a technical effect (in the present case a technically better or otherwise processed image) is not barred. Such decisions have provided guidance by the UK courts over a long period.

EPL argued that its invention injected technical contribution because its invention offered suggestions that were better compared with those provided by the current systems. In the initial hearing by the High Court, the same was found that the semantic similarity, which is the capability of the invention to generate a file with some attributes, was a technical and a truly real-world effect. This was an audacious and potentially game altering reading, which was likened to the case of Protecting Kids The World Over v. UKIPO , a content-filtering system was held to have more than an incidental technical effect (i.e. it is more than a mere computer program).

This decision was however, overturned by the Court of Appeal. It believed that the improvement was not a technical one. The reason the court gave is that the recommendation is founded on an elaborate procedure, but it is still a question of taste or subjective or cognitive comparison, it is not the solution to a technical fulfilment. It is an important difference: what the court meant by that it is not a clever algorithm, how complex and innovative it might be, and nothing more. It has to be a solution to a technical problem, and a more aesthetic solution would not serve the role of a solution. The decision of the court is reminiscent of decisions like the one in which Gale was granted an application in the case of Gale v Advocate General  , which maintained that simply because a program had caused a computer to run faster, it could not be termed as a technical contribution.

The rule provided in the case of the Court of Appeal is congruent with the structure used in AT&T Knowledge Ventures Application , which provided five so-called signposts to guide courts in deciding whether an invention is a technical contribution. These are whether the claimed effect technologically impacts a process beyond the computer, whether it acts at the level of the computer architecture, or whether it causes a different way in which the computer is used. The invention of EPL, based on the decision of the Court of Appeal, failed to accommodate these requirements in that, the impact of the invention was limited to processing of data in a non-technical subject matter.

The DABUS Debacle

When speaking of the Emotional Perception case, it is impossible to exclude another important legal case that pre-determined the UK attitude toward the AI patents: the DABUS case (Thaler v Comptroller-General of Patents, Designs and Trade Marks). Although it is a different case in a different area, DABUS offers an important context.

DABUS was about a patent application in which the sole inventor was an AI system. The Supreme Court of the United Kingdom by a unanimous decision decided that based on the Patents Act 1977, an inventor has to be a natural person, that is, a human being. The ruling reaffirmed that our present-day system of law is not prepared to give birth to inventorship to an AI, regardless of the degree to which it may have created a thought in an autonomous manner. The application of the statutory provisions by the court was based on plain reading of the statutory texts of the respective authorities, especially Section 7 of the Patents Act 1977 that provided that a patent could be granted to any person, and Section 13(2) can be applied to mean that the applicant must indicate the person in his opinion that is the inventor. A view was taken that a person in this matter means a natural person. This ruling brings the UK into conformity with European Patent Office and US legal thinking and the legal jurisdiction of other large nations.

Although the DABUS case is concentrated around the persona of the inventor, it falls under the same wider discussion as Emotional Perception. These two instances are indicative of the conflict between the fast-evolving, and even general, aspect artificial intelligence and the human-focused, stagnant, aspect of the current intellectual property law. They emphasize that although AI is a potentially useful device of inventions by human beings, neither is an AI itself, nor does it currently count towards a human inventor, a legal person entitled to patenting rights under the law as the provider of technically contributory innovations on its own.

The Indian Context

Despite a striking difference in statutory language, the nature of law surrounding AI and computer-related inventions is strikingly similar in its course to that of the UK. This is more so in the interpretation of the Section 3(k) of the Indian Patents Act 1970 that omits a mathematical or business method or a computer programme per se, or algorithms. Having the imperative terms, such as per se, has been in the pivot of Indian legal discourse where a technical effect test was evolved owing to there being no technical effect test, as expressed by the Delhi High Court in Ferid Allani v. Union of India & Ors., (2019). The court stressed that the invention should not be forbidden in this option provided that it has a technical impact or technical contribution. As a result, to be patentable an AI invention in India must be cast in a light that would reflect a tangible, non-obvious technical impact on a tangible system or a real world process, in line with the UK approach and which reflects an issue common across the globe in response to disruptive AI technology

The Road Ahead

Emotional Perception AI saga is more than a footnote in legal history, it is a potent lure to innovators, and legal professionals. The case decided by the Court of Appeal, in particular its findings of rejection of improved recommendation as a technical contribution, has not shut the door on AI patents, but rather it has replaced an otherwise simple lock with a much more complicated variant. To the new breed of inventors, the way into patentability of technology enhanced with ANNs and other highly developed algorithms has become a complex one with the legal fix lying well-grounded on the Aerotel/Macrossan test. It must be a careful transition of emphasis, from the cleverness of the code, to the physical, technical effect it achieves, a measurable advantage in computer processor performance, a new algorithm to process physical data, or an answer to a problem in some other non-computer field. The precedent that has now been cemented raises a question which is of great significance, is your AI the bright philosopher of aesthetics or is it a gifted engineer, getting technical issues done? This difference is at the heart of the future of AI patent law, not just in the UK but more widely and the EPL case has raised the stakes to anyone who wants to be the inventor in this braved new world.

Author: - Manannya Bhandari, in case of any queries please contact/write back to us via email to chhavi@khuranaandkhurana.com or at  Khurana & Khurana, Advocates and IP Attorney.