A recent court case has highlighted just how complex patents can be when they involve AI. If you’re working with an Artificial Neural Network or a similar computer program, then this may useful to understand.
In a recent appeal, the High Court overturned a decision by the UK Patent Office to refuse a patent application for an invention that uses an Artificial Neural Network (ANN). In a case known widely as the “Emotional Perception” case (after the applicant name), the UK Patent Office’s position was that the invention is just a new computer program that does make any technical contribution to the art and, as such, is not patentable.
However, the High Court judge disagreed, stating that a) there is no computer program in a trained ANN, an b) the invention does involve a technical contribution to the art.
Both of these aspects are quite interesting in themselves, and have sparked some controversy, especially now that UK Patent Office has appealed the decision and lawyers and innovators in this technical field are waiting to see what the Court of Appeal makes of them.
The invention
The invention is a method for providing media file recommendations similar to an input file, based on the similar tastes of other humans. An ANN, trained using a training set including media files that can be described objectively and subjectively, is used in a method that is able to identify similar media files by merging an assessment of the objective and subjective criteria together, recreating the way in which a human may do so. A file recommendation is then output to the user and this file recommendation is deemed to be “better” (i.e. the technical contribution lies) in that it is more likely to correspond with what the human user would have selected.
Is there a computer program in a trained ANN?
A trained ANN applies weights and biases in its individual nodes. These weights and biases are determined by the training process, and are not based on a computer program written by a human. On that basis, the High Court judge decided that there is no computer program in a trained ANN that could contravene the computer program exclusion in the UK Patents Act.
Does the invention make a technical contribution?
Having decided this, there should have been no call for the judge to go on to assess the invention in terms of the 4-stage “Aerotel” test, which as its final step requires that the invention makes a technical contribution to the art. Nevertheless, the judge went on to say that even if the invention was a computer program, it would be patentable on the grounds that it makes a technical contribution.
In essence, the judge said that the technical effect is the transfer to the user of the output file produced by applying the ANN model; even though the only “improvement” is that the recommendation is only “better” if viewed through the eyes of a subjective human.
Our thoughts
The High Court judge’s decision is surprising to say the least, especially in view of established earlier case law in the UK and the European Patent Office (EPO). If the decision is upheld by the Court of Appeal, it will mean far more than just that UK patent applications for inventions involving trained machine learning models will have a much greater chance of success. It will also mean that many more computer-implemented inventions may be considered to involve the requisite technical contribution if the only requirement is that a file is transferred and even though the benefit can only be determined by a subjective human (i.e. in this case, how relevant is the recommendation based on their specific preferences?).
It has long been the case that the ‘contribution’ of a patentable invention must be technical and not be grounded in one of the non-technical exclusions (which include, for example, methods of doing business, mathematical methods, mental acts and methods of displaying information). We are not entirely sure how a better recommendation, which can only be judged by a subjective human, can be considered a technical contribution.
No corresponding European application was filed for this invention, but we can’t help thinking that the outcome of such an application at the EPO would have been very different. The EPO would have been looking for a technical solution to a technical problem, and we think the EPO would have viewed the problem of providing (subjectively) “better” file recommendations as non-technical, but in this case, we’ll never know for sure.
Nevertheless, for now at least, it looks like potentially good news for UK innovators in this technical space, and we await with interest what the Court of Appeal will say. In the meantime, the UK Patent Office has issued updated guidelines in which patent Examiners are directed not to object to any invention involving ANNs under the computer program exclusion; and defines an invention involving ANNs as “an invention that claims an ANN itself or that includes claim limitations to training or using an ANN”.
This does not mean that all inventions including an ANN will automatically be patentable, because one or more of the other exclusions may still apply. In other words, the mere use of an ANN will not necessarily make an invention patentable. However, in our opinion, the scope of what may be patentable in this technical field does seem to have been broadened…for now. For more certainty, we’ll have to wait and see what happens once the Court of Appeal issues its decision.
Watch this space for the next instalment…
If you’d like to explore whether your technical innovation is patentable, then please book in a free, no obligation chat with Vicki, where we can discuss together what potential there might be.
