23/05/2024
On 14 and 15 May 2024, Lady Justice Davies, Lord Justice Arnold and Lord Justice Birss of the Court of Appeal heard the UK IPO’s appeal against the High Court decision [2023] EWHC 2948 in which Emotional Perception AI Ltd’s application directed to an Artificial Neural Network trained to recommend a musical track was found to be patentable.
The decision will not be released for a number of months. However, our impression of the hearing is set out below. Some background information and discussion on the High Court Judgment can be found in a previous Reddie & Grose article. In the meantime, the UK IPO has revised its guidelines relating to AI inventions. See our analysis here.
The Technology
The invention concerned involves an ANN trained by considering both natural language descriptions of files (one example of which is a music track) and the extracted physical properties of such files. In the case of music files, the natural language descriptions may convey how the music file might be perceived by a human such as based on its genre or lyrical content. The physical properties of music files can relate to properties such as the tone, timbre, speed and loudness of the music they contain. Once the ANN is trained, the extracted properties of target data (e.g., a music track to be analysed) are passed through the ANN. This produces an output that is compared with corresponding ANN-analysed files in a database.
Recommendations are then provided to an end user device by sending any relevant reference files that have property vectors that are measurably similar to the property vector of the target data. The recommendations identify relevant reference files in the database that are semantically similar to the target data. For example, in the case of a music track, the system can identify other music tracks that have natural language descriptions that are semantically similar to each other based on the similarity of the (physical) property vectors between the target music file and the relevant reference music files.
UK IPO Appeal
The UK IPO, as appellant, argued that the claimed invention had made no technical contribution, because the substance of the invention and indeed the only output of the trained ANN was essentially “any old” music file. Even if this were argued to be a subjectively or aesthetically better (in other words a music file that a listener would be more interested in listening to), the music file itself was technically unchanged by the process. The claimed invention was therefore considered to be quite different to the situation considered in the VICOM decision of the EPO, where the output of a computer running a mathematical method, was an image that had been sharpened and therefore technically improved by the process.
An interesting contribution made by Lord Justice Arnold during the hearing was that a recommendation “is still within the realm of aesthetics” even if the system that provides such a recommendation includes elements such as electronics or mechanics that make the overall system technical. The appellant made use of this point and argued that a mechanical version of a recommendation system would be allowable because the “guts of the machine are not excluded” whereas the legislature is such that the “guts of a computer program is excluded”.
The Applicant’s position
By way of contrast, the applicant (as respondent) argued that the claimed Artificial Neural Network was patentable because once trained, the ANN provided a more efficient recommendation system (implemented in hardware) that could be used “to identify swiftly and accurately files from a database which corresponds semantically to a target file, and to provide again swiftly and accurately file recommendations to a user device over a communication network”. This was argued to be a patentable “movement of data” based on adequate technical considerations of an underlying technical system, and therefore consistent with current UK case law.
The applicant also sought to emphasise that, although the contribution provided by the invention considers the accuracy of the file recommendation in terms of whether the target data and the metadata associated with a particular file are “semantically” similar, “the invention works by analysis of the physical properties [of the file]” (in other words characteristics such as, beats per minutes, frequency, amplitude etc).
The applicant highlighted that this functionality is consistent with an example of an allowable invention in a technically related case described in the European Patent Office (EPO) guidelines, which state that “[t]he classification of digital images, videos, audio or speech signals based on low-level features (e.g. edges or pixel attributes for images) are further typical technical applications of classification algorithms”. The appellant countered that there is some necessary context to the EPO’s guidelines because it is also a requirement to look at the technical purpose of the classification.
Summary
If the Court of Appeal decision agrees with aspects of the High Court’s Judgment, then the recent changes to the UK IPO’s Guidelines will likely remain in place, and further revisions will be needed to update the guidance and codify the more lenient approach to the interpretation of the statutory exclusions from patentability. On the other hand, if the Court of Appeal instead upholds the initial view of the UKIPO’s hearing officer to refuse Emotional Perception AI Ltd’s application, then the guidelines will likely revert to their previous form. In either scenario, it is possible that the Court of Appeal will provide further guidance on how to deal with ANN-based inventions in the UK. Practitioners in the UK eagerly anticipate the Court of Appeal’s Judgment.
This article is for general information only. Its content is not a statement of the law on any subject and does not constitute advice. Please contact Reddie & Grose LLP for advice before taking any action in reliance on it.
