The Comptroller-General’s Spotify daylist might be looking a bit “Downbeat rainy Monday morning” following the judgment issued in Emotional Perception AI Ltd v Comptroller-General of Patents, Designs and Trade Marks [2023] EWHC 2948 (Ch). The UKIPO has been forced to abruptly suspend its guidelines on patenting AI inventions following the High Court ruling, which finds that the training and use of an artificial neural network to suggest related media files to a user (e.g. to recommend songs) is a potentially patentable invention.

Section 1(2) of the UK Patents Act 1977 excludes from patentability “program[s] for a computer…as such”. In recent years, this exclusion has proved troublesome for those seeking patent protection for innovative uses of machine learning and artificial intelligence.

Although many UK patents have been obtained for inventions that use or relate to AI, the UKIPO has readily invoked the “computer program” exclusion for patent applications that involve computer-implemented AI techniques such as artificial neural networks (ANNs) unless the use of the computer results in some additional “technical contribution”. In other words, the UKIPO’s approach has been to treat computer-implemented AI inventions as any other computer program running on a computer. Last year, the UKIPO issued detailed guidance on patenting AI inventions, along with several illustrative example scenarios.

The patent application at the centre of the High Court’s ruling, by Emotional Perception AI, is directed to a system which uses an ANN to find semantically similar data files to a target file using their objectively measurable properties (e.g. to suggest subjectively similar songs to a user by measuring objective properties of a target song). Linking objective measurements to subjective qualities is a relatively typical use-case of an ANN.

The “clever” bit of the application (as is so often the case with ANN inventions) is how the ANN is trained. A database of reference media files (e.g. songs) is used, with a textual semantic description of each file provided by a human (e.g. describing a song as “frenetic” or “light”). The ANN is trained by taking pairs of files from this database and adjusting the parameters of the ANN to reproduce a semantic similarity distance between each pair of files (determined by natural language processing of the text descriptions) but using only the objective properties of each file. Once trained, the ANN can then be used on a new target file (without a text description) to find semantically similar data files from the database.

The independent claims considered by the UKIPO Hearing Officer (and the appeal judge) do not contain any explicit reference to “a computer”, “software” or the like. Nevertheless, the Hearing Officer considered that the claimed method was necessarily computer-implemented and refused the application for relating to a computer program as such. Emotional Perception AI appealed to the High Court, with the case being heard before Sir Anthony Mann in July 2023.

Sir Anthony’s judgment reverses the decision of the Hearing Officer and finds that the claimed system does not fall under the computer program exclusion. The judge found that the claimed use of the ANN is not a computer program at all, reasoning that, even when it is implemented in software, the specific parameters of the trained neural network (e.g. connection weights) aren’t directly programmed by a programmer but are instead the product of the machine learning done in the training phase. The judge also remarks that the ANN is “metaphorically” operating at a different level to the underlying software on the computer.

The judge considered the possibility that this conclusion was wrong and that the claimed ANN was in fact a computer program. However, the judge found that the claimed invention would still not be a computer program as such because the sending of the relevant reference files (i.e. the recommendations) to the user is a technical effect that happens external to the computer. Despite the result being a subjective improvement in user experience, the judge considered that the files were being selected in a technical way using the criteria determined by the ANN and that this was sufficient to escape the exclusion.

The judge’s finding that the claimed ANN is not a computer program at all would appear to make patent protection for inventions involving an artificial neural network markedly more obtainable. The judge has also seemingly lowered the bar for what may be considered a “technical” effect external to a computer to include those that lead to only subjective benefits.

Perhaps notably, the judge declined for procedural reasons to decide whether an ANN is a mathematical method that might fall under a different exclusion. For what it is worth, this author has some sympathy for the Hearing Officer’s view that an ANN, abstracted from the hardware or software in which it is implemented, is nothing other than a series of mathematical operations. Of course, the fact that the judge separately found the implementation of the ANN in question to be technical means that this issue is somewhat moot anyway.

In response to the judgment, the UKIPO has suspended the application of the AI invention examining guidelines it issued last year (Examining patent applications relating to artificial intelligence (AI) inventions – GOV.UK (www.gov.uk)). This is not surprising given that the facts of the case were uncomfortably similar to some of the applied AI example scenarios that were presented as non-allowable (even if the finding that an ANN is not a computer program at all is put to one side).

The UKIPO has also issued new guidance that “Patent Examiners should not object to inventions involving an ANN under the “program for a computer” exclusion of section 1(2)(c)”. https://www.gov.uk/government/publications/examination-of-patent-applications-involving-artificial-neural-networks

It remains to be seen whether the IPO will seek to appeal the decision.