Patents for computer programmes

Earlier this year, we looked at the case of Thaler v Comptroller- General of Patents, Designs and Trade Marks in which the Supreme Court considered whether artificial intelligence (AI) can be an inventor for the purposes of the Patents Act 1977 (Civil Litigation Case Update February 2024). The inventions in the Thaler case concerned a food container and a light beacon. It was undisputed that these were patentable, i.e., there was no issue around novelty, for example, and the inventions did not fall within categories that are excluded from patentability. But what about the patentability of the AI system itself? Under section 1(2)(c) of the Patents Act 1977 ‘a programme for a computer … as such’ is excluded from patent protection. Essentially, the position is that one cannot obtain a patent for a computer programme in itself; however, if the computer programme provides a ‘technical contribution’ to the real world, then it is patentable. The Court of Appeal recently considered this statutory provision in the context of AI in the case of Comptroller – General of Patents, Designs and Trade Marks v Emotional Perception AI Limited [2024] EWCA Civ 825. This follows an appeal from the High Court ruling in the case of Emotional Perception AI Ltd v Comptroller-General of Patents, Designs and Trade Marks [2023] EWHC 2948 (Ch).

Emotional Perception

Emotional Perception had applied to patent an artificial neural network (ANN), an aspect of AI. The ANN was said to be capable of providing improved media file recommendations. Taking the context of music websites for example, where a user wants to receive music similar to music they already have, traditional tools would recommend music tracks based on similar categories of music (e.g., rock), those categories having been tagged as such by humans. Instead of taking the ‘category’ of music as the criteria for recommending similar music tracks, the ANN based system is said to identify similar music tracks based on human perception and emotion. In brief, the system works as follows: it takes a pair of music files, which have been given a semantic label, e.g., ‘happy’, ‘relaxing’, and so on. Using Natural Language Processing, the files are plotted in a ‘semantic space’, with the distance between the files indicating their semantic similarity. In addition to their semantic properties, the files are also analysed according to physical properties such as tone, timbre, speed etc., and again plotted in a ‘property space’. Using back-propagation, the property space is corrected in order to reflect the semantic space, so that semantically similar tracks are close in the property space, whereas semantically dissimilar tracks are farther apart in the property space. The operational ANN is then able to take a music track, determine its physical attributes, plot these against the physical attributes of other music tracks in a music library or database, and by looking for those tracks which are most proximate in terms of physical characteristics, it can recommend semantically similar tracks.

An officer for the UK Intellectual Property Office (UKIPO) refused grant of the patent on the basis that the ANN system was considered to be ‘a program for a computer’ and that the patent application related to that ‘computer programme as such’. Emotional Perception appealed to the High Court challenging the decision by the UKIPO to refuse grant of the patent. The matter came before Sir Anthony Mann, J, who considered whether i) the ANN was ‘a program for a computer’ therefore falling within the statutory exclusions to patentability, and ii) if it was, whether there was a technical contribution which meant it fell outside the exclusionary regime.

Differences between hardware and software ANNs

On the first point, Mann J, differentiated between hardware ANNs and software emulated ANNs, and concluded that neither qualifies as ‘a programme for a computer’ and therefore neither was excluded from patentability. In the case of hardware ANNs, it was accepted by the parties that there is no ‘programme’ and therefore this would not fall within the exclusions. ‘The hardware is not implementing a series of instructions pre-ordained by a human. It is operating according to something that it has learned itself’ [54]. In the case of software emulated ANNs, there were two aspects in which computer programming plays a role, one being the training stage, and the other being the software platform which enabled the computer to carry out the emulation. As regards the latter, Mann, J considered that this can be de-coupled from the ANN: ‘It seems to me that it is appropriate to look at the emulated ANN as, in substance, operating at a different level (albeit metaphorically) from the underlying software on the computer, and it is operating in the same way as the hardware ANN. If the latter is not operating a program then neither is the emulation’ [56].

The court found that the ANN, in itself, was not a computer programme because it was not operating a set of programme instructions given to it by a human. The ANN had trained itself, applying its own weights and biases. It was emulating a piece of hardware which had physical nodes and layers, and was no more operating or applying a program than a hardware system was. With respect to the computer programme involved at the training stage, which sets the training objectives and parameters in which the ANN is to operate, the court concluded that this fell outside the actual invention that is claimed. The invention was not a claim to the computer programme at the training stage; the invention related to the idea of using pairs of files for training, and setting the training objective and parameters accordingly. The claim therefore went beyond the actual computer programme.

As explained above, even if an invention were to be a claim to a computer program, it may still be patentable if it provides a “technical contribution” outside the computer program itself. Given his conclusion that the ANN was not a computer programme, Mann J, did not need to consider the question of technical contribution, but he nevertheless did. Following a review of caselaw on what constitutes a ‘technical contribution’, Mann J found that the sending of a file recommendation to an end user is a matter external to the computer and amounts to a technical contribution, i.e., the ANN has a real world effect outside of the computer.

ANNs come within the purview of the exclusion from patentability.

The UKIPO appealed to the Court of Appeal (CoA). The unanimous CoA judgment, given by Lord Justice Birss, overturns the first instance decision of the High Court and upholds the decision of the UKIPO officer. The CoA first considered whether ANNs are computer programs and observed that the meaning of ‘program for a computer’, as it appears in section 1(2) of the Patents Act 1977, is a question of law. The CoA concluded that ‘a computer is a machine which does something, and that thing it does is to process information in a particular way. The program is the set of instructions which cause the machine to process the information in that particular way, rather than in another way’ [61]. In view of this definition, the CoA held that ANNs, irrespective of how they are implemented (in hardware or software form), constitute computer programs: ‘Turning to an ANN … however it is implemented, such a machine is clearly a computer – it is a machine for processing information. Focussing on the weights of an ANN, in my judgment irrespective of the manner in which an ANN is implemented (hardware or software) … these weights are a computer program. They are a set of instructions for a computer to do something. For a given machine, a different set of weights will cause the machine to process information in a different way’ [68]. As such, ANNs come within the purview of the exclusion from patentability.

Emotional AI’s ANN

The CoA then considered whether Emotional AI’s ANN involves a substantive technical contribution so as to lift it out of the excluded subject matter. The CoA concluded that the sending of an improved music file recommendation is not a technical effect: ‘What made the recommended file worth recommending were its semantic qualities. That was a matter of aesthetics … they were subjective and cognitive in nature. They were not technical and had not turned it into a system which produced a technical effect outside the excluded subject matter … The system went about its analysis and selection in a technical way but that was because it was an ANN, i.e. a computer. The fact the computer was using properties it could measure to make this semantic recommendation made no difference’ [79 – 80]. Although Emotional AI’s ANN was judged to be excluded from patentability, the CoA emphasised that the fact that section 1(2) of the Patents Act 1977 is engaged in a case of an ANN implemented invention, as much as it would be in any computer implemented invention, does not mean it is unpatentable. ANNs can be lifted out of the exclusion if they can be shown to make a technical contribution.

The UKIPO has currently suspended its guidelines for examining patent applications relating to AI. However, in view of the CoA judgment, UKIPO is likely to treat inventions involving AI as computer implemented and therefore applications would have to be considered under the computer program exclusion exemption, i.e., whether the invention produced a technical contribution. The position in Europe is similar with the European Patent Convention considering inventions involving AI as computer-implemented inventions which would only become patentable if they are applied to solve a technical problem in a field of technology.

The Court of Appeal judgment can be accessed here.

For any other enquiries, or to discuss instructing the Civil Team at Atlantic Chambers, please contact clerks@atlanticchambers.co.uk