Computer implemented inventions: Comptroller General of Patents v Emotional Perception AI Limited

Those with a keen eye on all things AI and patents will recall that, in a judgment Emotional Perception AI Ltd v Comptroller-General of Patents, Designs and Trade Marks [2023] EWHC 2948 (Ch) delivered in November 2023, the High Court had considered for the first time whether the use of an aspect of Artificial Intelligence (AI), namely an Artificial Neural Network (ANN), engaged the statutory exclusion on obtaining patent protection for a "program for a computer…as such". Adopting perhaps a more permissive approach to the statutory exclusions than we might have previously observed, Mann J held that the statutory exclusion did not apply to the invention as claimed. The Comptroller of Patents, having previously refused Emotional Perception AI's application for patent protection, appealed. The appeal was heard in May 2024. In allowing that appeal, the Court of Appeal has upheld the UKIPO's original decision to refuse to grant a patent for Emotional Perception AI's ANN driven music recommendation tool. In doing so, the Court of Appeal has sought to remind us that AI related inventions are, at their heart, simply a continuation and evolution in the field of software patents, to which many of us in the patent world have dedicated considerable time. In delivering judgment, it has ruled that whilst ANN implemented inventions may be patentable, they must be assessed in the same fact-specific way as any other computer-implemented invention. In delivering judgment, Birss LJ (with whom Davis LJ and Arnold LJ agreed) held that Emotional Perception AI's ANN music recommendation tool was, indeed, a computer program and did not otherwise make a technical contribution that might take it outside the reach of the computer program exclusion, and therefore was not patentable.

Given the investment into AI solutions across a range of industry sectors, we would anticipate that the question of patentability of those inventions is such that consideration of the issue by the Supreme Court will not be too far away.

The invention

The invention described in Emotional Perception AI's patent application concerns an ANN which is trained to perceive semantic similarity or dissimilarity between media files. It claims an improved system for providing such media file recommendations to an end user. In practice, we might envisage the invention being used, for example, by a music website where a user may be interested in receiving music similar to a track they had previously enjoyed. The advantage of the invention is that it can offer suggestions of similar music in terms of human perception and emotion irrespective of genre and taste.

An aspect of the invention involves a first ANN which is trained to characterise certain music tracks based on a description of how they are perceived by a human, e.g., happy, sad, relaxing. Co-ordinates are produced in a notional "semantic space" for each track in a pair of music files. Two tracks which are semantically similar will have co-ordinates closer together.

A second ANN analyses the physical and measurable properties of the same two tracks, such as tone, timbre, speed, loudness, and produces co-ordinates in a notional "property space". Similarities or differences are reflected in the proximity of the co-ordinates. The second ANN is then trained to make the distances between pairs of the property co-ordinates converge or diverge in alignment with the distancing between them in the semantic space. This training is achieved by a process called back-propagation in which the “error” in the property space is corrected using an algorithm provided by a human.

Correction is achieved by the ANN adjusting its own internal workings and it learns from experience without being told how to do it by a human. The ANN therefore learns by repetitive correction, how to detect semantic similarity or dissimilarity from physical characteristics in a music track in order to recommend semantically similar music files.

High Court decision and UKIPO response

After detailed consideration of the authorities on excluded subject matter, and contrary to the position taken by the UKIPO, the High Court concluded that an ANN was not a program for a computer and should, in effect, be treated as a piece of hardware, irrespective of whether it was directly implemented as hardware or as an emulated ANN. Accordingly, it held that the statutory exclusion to patentability in s1(2)(c) Patents Act 1977 ought not to be invoked, either because the invention as claimed was not "a computer program as such" or, if it was, because the invention as claimed demonstrated a technical effect.

Following the High Court judgment, the UKIPO updated its guidelines for examining patent applications relating to inventions involving ANNs to confirm that the computer program exclusion should not be invoked for inventions which involve ANNs implemented as physical hardware or emulated using software. This meant that for a brief period, the UK approach to certain AI inventions differed from the approach adopted by the European Patent Office.

Court of Appeal decision

What is a computer program and is there a computer program in an ANN?

The first point that the Court considered was whether Emotional Perception AI's ANN was a computer program. The definitions of a "computer" and "computer program" were discussed at length, which highlights the lack of an established definition for these concepts for the purposes of the Patents Act 1977, which is particularly challenging considering the significant evolution of computer technology since 1977.

Emotional Perception AI's position was that the weights and biases of an ANN (whether hardware or software) were not a computer program. It argued that a computer program takes the form of serial, logical 'if-then' type statements defined by a human programmer which define exactly what it is that the programmed computer does and therefore the weights and biases of an ANN were not a computer program.

The Court of Appeal applied its definition of computer, which it found to be "a machine which does something, and that thing it does is to process information in a particular way" and computer program, which it found to be "a set of instructions for a computer to do something". It rejected Emotional Perception AI's argument on the basis that:

• referring to a human programmer was not relevant or helpful and that there was no justification for drawing a distinction in law between instructions created by a computer and those created by a human;
• the fact that ANNs aim to solve problems which are not easy to solve with conventional computers was irrelevant as both conventional computers and ANNs can aim to solve problems which are difficult for humans to solve unaided;
• whether the program for a given computer is fixed in a permanent form or not does not alter the fact that the program represents a set of instructions for a computer to do something;
• the fact that the set of weights and biases does not take the form of a logical series of 'if-then' type statements is irrelevant, the set of weights and biases as a whole instruct the machine to process information it is presented with, in a particular way.

Therefore, the Court of Appeal found that the ANN based music recommendation tool was a set of instructions for a computer to do something, and therefore a computer program.

Does the claimed invention involve a substantive technical contribution?

Having determined that Emotional Perception AI's invention, as claimed, was for a computer program, the Court of Appeal went on to consider whether the invention made a technical contribution that would take it outside the reach of the exclusion, and therefore make it patentable. The Aerotel test was central to this discussion as was whether the actual contribution was either the sending of an improved recommendation message, or the training process of the ANN rather than the output it provides.

The Court found that the process of training the ANN was "in effect, part of the creation of the program", and therefore not part of the technical contribution and that the provision of a recommendation message was the "presentation of information" which was also unpatentable unless it went further and involved a technical contribution.

As a final point, the Court found that the improved file recommendation was based on semantic, cognitive or aesthetic qualities, rather than technical ones and the non-technical quality of the contribution meant that it could not be held to be a system which produces a technical effect outside the excluded subject matter, and therefore was not patentable.

Comment

If the High Court judgment back in November 2023 had perhaps signalled the beginning of a new approach to the patentability of AI in the UK, the Court of Appeal ruling might be said to have sought to remind us that the patent system in the UK has been wrestling with the complexities of computer implemented inventions for decades. AI related inventions, while having an extraordinary potential to advance innovation in a range of industry sectors, are merely another step along that journey.

Emotional Perception AI has indicated that it will appeal the decision of the Court of Appeal to the Supreme Court. For now, ANN-based inventions will adopt the same position under patent law as other computer implemented inventions. They will only be patentable if they make a “technical contribution”. This is and always has been a high hurdle to overcome in the case of computer programs, and it remains to be seen whether the rapid evolution of ANNs and other AI related inventions will mean that in time that approach will change and potentially lead to a divergence of approach to patentability of AI inventions across jurisdictions.

If you found this article interesting, you might be interested in our monthly data protection bulletin for which you can register here.

You might also be interested in our previous article:  

The Neural Network – July 2024