Home: Maucher Jenkins

Intellectual Property


Patents | Trade Marks | Designs

News & Commentary

UK Patent Office takes a step ahead of the EPO in patenting deep learning inventions

Date: 1 December 2023

Following our report of Emotional Perception AI Ltd. v Comptroller General of Patents, Designs and Trade Marks here, we can further report that the UKIPO has published a new Practice Note on the patenting of Artificial Neural Networks, available here.


The Practice Note simply says Examiners should not object to inventions involving ANNs under the “program for a computer” exclusion. The Manual of Patent Practice and the office’s guidelines for examining patent applications relating to ANNs will be updated in due course.


Let us see what we can glean from this.


There are two limbs to the Emotional Perceptions judgement: (1) an ANN is not a program for a computer; and (2) in the facts of the case, sending a file (selected by the ANN system) to a user device amounts to an external technical effect.


What we need to ask ourselves is how far-reaching is this judgement? Is one of these two factors enough to avoid the exclusion? Or is the exclusion avoided on account of both factors?


On the first factor, Mann J. reasons that an ANN can be implemented in hardware or software. If implemented in software (as is more usual), the computer program is a simulation of the hardware. Clearly, if implemented in hardware, the ANN is not a “program for a computer” and is not excluded from patentability. This being so, there is no reason to treat a software implementation any differently. By this rationale, anything that could be implemented in hardware is not deprived of patentability merely because if could be implemented in software. Voice processing and the like is typically implemented in software, but could conceivably be implemented on a hardware analog computer. Discrete Fourier transform processing of radio signals can be implemented either way.  But these sorts of processes have not been in doubt as to patentability. Their patentability has been secure ever since Vicom T208/84. The doubt has always been where the effect of the computer program has been in an excluded field (mental act, business method, presentation of data etc.) and has therefore been deemed non-technical.


In other words, it is not at all radical that an invention implemented on an ANN can be patentable. It is, however, a change in UK law if this is a sufficient reason to set aside the exclusion of section 1(2)(c) such that any invention implemented on an ANN passes that hurdle. If so, what if there is no novel and inventive technical effect? What if the only novel (and inventive) effect is non-technical?


Before answering that question, let us consider the second limb of the decision.


The Court was satisfied that sending a file to a user was an external technical effect. Heretofore, the UKIPO has considered a network of computers to be “a computer” such that a program that sends a file from one computer to another is no less of “a program for a computer”. Would it be enough that the ANN outputs a link to a file? Is it critical that the file (or the link) is sent to another device?  Might it be enough that the link is available to that other device (which is practically and technically almost equivalent)? The UKIPO practice note is silent on this. For the UKIPO it seems the first limb of the test is enough. The “involvement” of an ANN in the claim is enough to avoid the exclusion. This is consistent with the decision, in which the first limb was sufficient and the second limb was considered merely on an “if I am wrong” basis (para 62 of the judgement).


Might the second limb alone be sufficient (viz. that the computer spits out a new file)? The Practice Note does not say this, and neither does the judgement. The judgement says:


It is not just any old file; it is a file identified as being semantically similar by the application of technical criteria which the system has worked out for itself. So the output is of 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.


So it is clear that the UKIPO door is now wide open for any invention that involves an ANN, and the door is wider than that of the EPO, which applies the Comvik approach (T0641/00) of setting aside novel non-technical features and asking whether the remaining novel technical features involve an inventive step. In setting aside non-technical features, the EPO disregards the cognitive content of data.  It is unlikely the EPO would have viewed this invention in the positive light of the UKIPO, but the EPO does indeed grant many applications for AI systems, and it should not be ruled out.