Trained ‘artificial neural networks’ can clear patent hurdle, says UK court

Sir Anthony Mann said that even if trained ANNs constitute computer programs, it may still be able to be shown that they make a “technical contribution” that lifts them out of an exclusion on patentability that applies to computer programs under the UK’s Patents Act 1977.

The judge considered the issue in a case in which a company behind a media recommender system, Emotional Perception AI Ltd (EPAI), successfully overturned an earlier decision by a hearing officer at the UK’s Intellectual Property Office (UKIPO) who determined that their claimed invention was not patentable.

Patent attorney Harry Muttock of Pinsent Masons said the ruling provides “a clear direction to examiners at the UKIPO that trained artificial neural networks should not be automatically considered excluded subject matter”.

EPAI has developed a system that it claims makes music and other media recommendations to users in an improved way, based on how the users have been categorised by trained ‘artificial neural networks’ (ANNs).

An ANN is described by the High Court as an aspect of AI system and as being “a black box which is capable of being trained as how to process an input, learning by that training process, holding that learning within itself and then processing that input in a way derived from that training and learning”.

Two ANNs operate in EPAI’s system. The first is a system that receives media files, together with natural language descriptions of how a human perceives those files – such as happy, sad, relaxing. The ANN analyses how the files have been described and, based on human instructions, makes assessments of how similar or different the respective media files are.

The second ANN analyses the media files, via parameters set by a human, to determine their characteristics – such as, in the case of music, its tone, speed, and loudness – and again makes assessments of how similar or different the respective media files are. The second ANN is also trained to cross-analyse the similarities and differences established by both ANNs and adjust its own internal assessment parameters in order to ultimately make recommendations that more closely align with what it perceives a user likes.

EPAI applied to the IPO for a patent for its invention. Dr Phil Thorpe, IPO hearing officer, considered, however, that the EPAI system as a whole was a computer program and that its provision of “semantically similar file recommendations” was “not technical in nature”, meaning he considered it was excluded from patentability under the Patents Act.

In overturning that ruling, however, the judge, Sir Anthony Mann, considered that the patent claimed by EPAI was not for a computer program at all, and therefore that the associated exclusion to patentability under the Patents Act did not apply.

The judge said: “What is said to be special is the idea of using pairs of files for training, and setting the training objective and parameters accordingly. If that is right, and I consider it is, then the actual program is a subsidiary part of the claim and is not what is claimed. The claims go beyond that. The idea of the parameters itself is not necessarily part of the program. On this footing as a matter of construction the claim is not to a computer program at all. The exclusion is not invoked.”

In case he was wrong on that point, however, Sir Anthony, went on to consider the extent to which the EPAI system was patentable even to the extent it constituted a computer program.

In this regard, the judge first considered that Thorpe had been wrong to evaluate EPAI’s system as a whole. He determined that it was appropriate to “decouple” the part of the system used to implement the trained ANN and the part of the system used to train the ANN. He considered that it was only “the program which achieves, or initiates, the training” which constitutes a computer program.

Notwithstanding that he found that what EPAI claimed under the patent application was not a computer program at all, Sir Anthony went on to consider whether the program which achieves, or initiates, the training – which he characterised as the ‘hardware ANN’ – provided a ‘technical contribution’, which would have lifted it out of the Patents Act’s exclusion on patentability for computer programs. He said it did, disagreeing with Thorpe’s assessment of the matter, on the basis that the system produced an output – a file – that “would not otherwise be selected”.

“That seems to me to be a technical effect outside the computer for these purposes, and when coupled with the purpose and method of selection it fulfils the requirement of technical effect in order to escape the exclusion,” Sir Anthony said.

“The invention is not just one depending on the effect of the computerised process on the user. There is more than that. There is a produced file with (it is said) certain attributes. The file produced then goes on to have an effect on the user (if the thing works at all) but one cannot ignore the fact that a technical thing is actually produced. It would not matter if the user never listened to the file. The file, with its similarity characteristics, is still produced via the system which has set up the identification system and then implemented it,” he added.

In making that assessment, Sir Anthony suggested that a computer-implemented method for selecting non-technical criteria may fulfil the requirement of a technical effect if the selection deriving from the application of the criteria is technical in nature. Muttock said that Sir Anthony’s views on trained ANNs and technical effect could prompt the UKIPO to revisit guidance it has issued on examining patent applications relating to AI inventions.

The UK government previously opted against making changes to UK patents law to account for growth in the development and use of AI systems, and it also reversed plans to introduce a new UK data mining copyright exception to support AI development following pushback from the creatives industry.