Tue 28th May 2024

UKIPO appeals Emotional Perception decision

Service: Patents

Sectors: AI and data science

The UK’s approach to patenting AI inventions could change very soon! The High Court’s unexpected decision in the case of Emotional Perception AI Ltd v Comptroller-General of Patents, Designs and Trade Marks has been appealed by the United Kingdom Intellectual Property Office (UKIPO).

The invention concerns an Artificial Neural Network (ANN) which is trained to find semantic similarity from the physical properties, such as tone and pitch, of a media file, to make music recommendations to users. The patent claims recited a method of training and the use of a trained ANN for prediction, as well as a system comprising the trained ANN. Before the UKIPO, the Hearing Officer considered the subject matter to be excluded under the “computer program” exclusion. Emotional Perception AI appealed to the UK High Court which found that the implementation of the trained ANN did not constitute a computer program because the ANN is able to learn via the training process, rather than merely implementing code.


This decision has had important implications for patenting Artificial Neural Networks in the UK.  The UKIPO, bound by the Judgement from the UK High Court, has updated their guidelines for examining patent applications relating to artificial intelligence (AI), confirming, that the computer program exclusion should not be engaged for inventions involving ANNs, implemented as physical hardware or emulated using software. According to the new guidelines, to escape the computer program exclusion, the applicant would have to “either claim an ANN itself or include claim limitations to training or using an ANN”. Thus, a claim to a generic machine learning model or the training thereof would not escape this exclusion. Other exclusions from patentability may apply to ANNs, such as the mathematical method exclusion.


Following this Judgement, the UK approach to AI inventions now differs from the approach adopted by the European Patent Office (EPO). The High Court decision is now being appealed by the UKIPO, and thus the UK approach to AI inventions could change yet again. On the 14th of May, Emma and Shameena attended the appeal proceedings at the Court of Appeal where the case was heard by Lady Justice Nicola Davies, Lord Justice Arnold and Lord Justice Birss.


The UKIPO first presented their arguments and their grounds of appeal. In particular, the UKIPO discussed the definitions of a “computer” and a “computer program” at length. The UKIPO argued that a generic ANN is itself a form of computer, with the set of weights and biases of the trained ANN serving as a computer program that instructs the generic ANN computer to perform a task. Thus, the UKIPO’s first ground of appeal was that the Judge in the High Court was wrong in finding that the computer program exclusion is not engaged at all.  


The UKIPO further presented their first arguments around the mathematical method exclusion, which they had not presented at the High Court because of procedural reasons. The UKIPO argued, in the third ground of appeal that, the Judge in the High Court was wrong in not considering the mathematical method exclusion. The UKIPO discussed that if an ANN is decoupled from a computer program, then it is abstract enough to fall under the mathematical method exclusion.


According to the “Aerotel/Macrossan test” used at the UKIPO to assess excluded subject matter, if an invention provides a technical contribution, it may avoid exclusion. In this context, the UKIPO argued that the contribution of the invention, which is the sending of music file recommendations, is non-technical, and thus does not allow the invention to escape either exclusion.


In response to the arguments put forward by the Comptroller, Emotional Perception disputed the use of the approach taken by the comptroller when assessing whether the claimed invention falls within the exclusions. They submitted that an incorrect approach was adopted by the Comptroller in their skeleton arguments.


Specifically, Emotional Perception submitted that the Comptroller was allegedly looking at the claimed invention, extracting the inventive step, or ‘clever bit’, and only then asking whether that falls within the exclusions. Emotional Perception stated that this specific approach had been rejected in Aerotel. However, such an approach was commonly, albeit incorrectly, adopted when considering applications of this sort. They submitted that the UKIPO commonly focuses on inventive step and disregards related features.


Emotional Perception sought to clearly explain the technology of the invention as defined in the claims. There was a particular focus on defining what can be considered the key components of a trained ANN such that the trained ANN can be recreated from an ‘off-the-shelf’ neural network. The Comptroller submitted that any trained ANN can simply be defined in terms of the weights and biases learned through a training process. Emotional Perception disputed this definition, stating that one cannot separate the weights and biases from the links and nodes of an ANN. They submitted that the weights, biases, links and nodes are all intertwined such that the topology of the ANN can be redefined in training. They argued that it would therefore require more than inputting just the weights and biases into an ‘off-the-shelf’ neural network to recreate the ANN as claimed.


The definitions of a ‘computer’ and a ‘program for a computer’ were also discussed. It was accepted that a computer can be defined using the dictionary definition as included in the High Court judgement.


Emotional perception submitted that the claimed ANN cannot be considered a program for a computer such that a computer program can be considered a series of instructions that cause a computer to perform a task.  They stated that for a (human) programmer to write a program, the (human) programmer must be able to define the problem to be solved and generate a sequence of instructions to solve it. Emotional perception submitted that it is not known how the features of a music track relate to its semantic nature as determined by the claimed ANN. Therefore, a human would not be able to prescribe how to analyze musical characteristics and the problem can be considered an intractable problem. Lord Justice Birss questioned this argument by referring to impure functional programming where functions themselves create programs.  


Judgement was reserved at the end of proceedings. The first and third grounds of appeal are of particular significance for AI Inventions and thus the CA decision is a much awaited one. If the Judgement from the CA finds that the trained ANN is neither “a computer program” nor a “mathematical method”, inventions concerning trained ANNs, and possibly many more types of trained AI could avoid both exclusions, paving the way for these inventions to benefit from patent protection in the UK.  


This briefing is for general information purposes only and should not be used as a substitute for legal advice relating to your particular circumstances. We can discuss specific issues and facts on an individual basis. Please note that the law may have changed since the day this was first published in May 2024.


Our offices


+44 20 7831 7929

Bedford House, John Street, London, WC1N 2BF, United Kingdom


+44 20 7831 7929

Suite 3.05, Platform, New Station Street, Leeds, LS1 4JB, United Kingdom


+44 20 7831 7929

Generator Hub, The Gallery, Kings Wharf, The Quay, Exeter, EX2 4AN, United Kingdom


+49 89 5150 5800

Widenmayerstr. 10, D-80538 München, Germany