"What is the computer, and where is the program?" AI and the Judge

While browsing LinkedIn yesterday, I stumbled upon an intriguing case, initially shared by Martin Ebers , president of the Robotics & AI Law Society - RAILS .

In the United Kingdom, the Patents Act 1977 section 1(2)(c) states that computer programs are not eligible for patent protection. Nevertheless, the Emotional Perception AI Ltd case brought forth, in the words of the judge Sir Anthony Mann, "new questions because it involves deciding whether the use of an aspect of Artificial Intelligence, namely an Artificial Neural Network ("ANN"), in the circumstances of this case, engage[d] the exclusion".

What is the invention in question?

The patent in discussion is about an innovative system for recommending media files, such as music tracks, to users. Unlike conventional methods that sort music by genre, this system employs an ANN to suggest music based on human perception and emotional response, irrespective of musical genre or other users' preferences - offering a more intuitive and personalized user experience.

"A pair of music files is taken, each of which is accompanied by a natural language description [...] of the type of music [...] in terms of how that music is perceived by a human. At its simplest the music might be described as happy, or sad, or relaxing."

In essence, the ANN in question is designed to recommend music by interpreting both the emotional content of songs, as described in natural language, and their physical characteristics, such as tone and tempo. The system utilizes two ANNs: the first classifies tracks into a "semantic space" based on their emotional descriptors — like "happy" or "sad". Concurrently, a second ANN analyzes the same music for physical properties, placing them in a "property space".

A unique aspect of this technology is how the second ANN adjusts the physical coordinates to align with the emotional, semantic space through back-propagation — in essence, it corrects itself to match the emotional understanding of music. This self-learning approach enables the ANN to identify the emotional resonance of a track without explicit programming instructions.

Once trained, the ANN can then provide recommendations that are emotionally akin, not just based on genre or physical attributes, but on a deeper semantic level. It does this by comparing the physical vectors of a user-provided track with those in its database, suggesting semantically similar music. This represents a more intuitive approach to music discovery, leveraging AI to tap into the emotional undercurrents of music preferences.

"The ANN has learned how to discern semantic (dis)similarity from physical properties. It has not done so because any human (programmer) has told it how to do it. It has done it by producing results, being provided with information reflecting its degree of error, adjusting its own internal assessment parameters, reprocessing the files to reduce the error and repeating this process until it gets it sufficiently right sufficiently often."

After considering all aspects of the previous decision, the judge pondered two fundamental questions, previously unaddressed: "What is the computer, and where is the program which is said to engage the exclusion?"

He further explored these questions by first defining a computer as "a machine [...] intended to operate a computer program". Then, considering this definition, is a hardware ANN a computer? Is a software-emulated ANN the same as the machine itself? Since it processes data, it seems likely that it is indeed a computer, and in the case of a software emulation, it must run on a computer.

However, does a program run itself in the case of the emulated ANN? Certainly, at least at the input-based training stage. But what about the subsequent internal training?

"There was no program at that point because no person had given a set of instructions to the computer to do what it does - the ANN had trained itself. What it was operating was not a set of program instructions at all. It was applying its own weights, biases and so on to produce relevant vectors or co-ordinates."

As someone deeply engrossed in generative AI and ANN literature, this case felt like reading a thriller - the suspense was palpable. How do we classify something that, at some point, exists independently, be it virtually or physically? Regardless of the packaging (an existing computer or a specialized black box), this case delved into uncharted territory. Despite the perhaps unattractive guise of intellectual property protection, legal proceedings often lead the way in understanding science, as regulations - and judges - scramble to keep pace with technological advances.

The judge ultimately concluded that the ANN must be considered separately from the underlying software because it operates on a "metaphorical" different level (similar to a hardware ANN). The original patent claim focused not on the input-based training stage "which initiates the training", but rather on the distinct "idea of using pairs of files for training", the precise step where the ANN is [no longer] implementing a series of human-prescribed instructions, but "operating according to something that it has learned itself".

The court recognized the uniqueness of the trained ANN's method of operation — how it learns and applies this knowledge — as a significant part of the invention. This transcends the traditional concept of a computer program executing predetermined instructions; the inventive aspect lies in how the ANN utilizes its training to make recommendations, not merely in the underlying computer program. This reflects a broader principle in patent law: while mere computer programs are not patentable, inventive technical contributions that extend beyond standard programming, particularly those involving significant human conceptual input, can be patentable. Confronted with the IP law's definition of patentable "creations of the mind", we now have an intermediary level of consideration between the human creator and their machine creation, perhaps best termed as a "machine creator". And it remains to be explored, across various philosophical disciplines, how we will define this new kind of "mind".

Are you intrigued by the ever-evolving world of AI and cybersecurity? Do you thrive on staying ahead of the curve in these dynamic fields? If your answer is a resounding "yes", then I invite you to join me on a journey of discovery and professional growth.

?? Follow me on LinkedIn for insightful discussions, the latest trends, and thought-provoking content that delves deep into the realms of artificial intelligence and cybersecurity. Together, we can explore the cutting-edge developments shaping our digital future.

But why stop there? Extend your network further by following the official LinkedIn page of the The LHoFT - Luxembourg House of Financial Technology . We have a vibrant community of experts, visionaries, and innovators, all dedicated to advancing the world of FinTech and beyond.

#ArtificialIntelligence, #Cybersecurity, #AInnovation, #TechTrends, #LegalTech, #PatentLaw, #ANN, #MachineLearning, #TechEthics, #DataScience, #IntellectualProperty, #FutureOfTech, #TechCommunity, #DigitalTransformation, #Robotics, #AILaw, #EmergingTech, #AIResearch

Nasir Zubairi Alex Panican Mélina Guilli Anna Sobków Symeon Papadopoulos Annabelle Albert-Fonteneau Luca Mancuso ???? Romain Iovalone Dilek Ayaydin- Batal Christophe R. Vojtech Seman Emilie Allaert Paola Grosjean Carcamo Claudio Bareato Rapha?l MANSUY Roberto Garcia Thomas Taschler