Article image created using StableDiffusion 2.0
The decision BL O/938/22 of the UKIPO hearing officer considers the patentability of optimisation / simulation methods, the results of which have an effect outside the computer. While the claim is not specifically about machine learning, clearly the concept of optimising something for use in the real is an important one for many machine learning inventions. This particular invention runs a simulation to optimise an athlete’s garments for a given route (disclosed embodiments relating to road cycling). An athlete then wears the clothing on the route for which it has been optimised. They question was whether this amounts to an effect outside the computer that would take the invention outside the exclusions from patentability.
The UKIPO approach to computer-implemented inventions in general, and machine learning specifically, has received less exposure in my newsletter and is generally less familiar to practitioners outside the UK, as compared to the EPO’s approach. So, a quick overview to get us started seems appropriate.
As at the EPO, the UIKIPO’s starting point is that machine learning is a mathematical method and hence excluded from patentability per se. More akin to the approach in the US under s101 35USC than the EPO’s approach, the UKIPO deals with the question under inherent subject matter eligibility, rather than under inventive step, which is assessed in the normal way if the subject matter is found not to be excluded. To do this, the English case law has developed a structured four step approach:
properly construe the claim;
identify the actual contribution;
ask whether it falls solely within excluded subject-matter;
check whether the actual or alleged contribution is actually technical in nature
The crucial step is step (iii), similar to the EPO asking whether there is a technical effect or not, since in practice step (iv) is already answered if step (iii) is answered in the positive. As to the first two steps, there was little controversy about how the claim should be construed and agreement that the actual contribution of the invention was the identification of the most suitable athletic garments based upon a simulation for a specific route and a resulting improvement in the performance of the athlete when performing on the specific route.
The disagreement, then, was about whether the actual contribution falls solely within excluded subject matter or not (step iii). English case law provides a number of signposts that can indicate that an invention is not excluded. The one relevant here asks whether the invention has a technical effect on a process carried out outside the computer. This is where the applicant and examiner disagreed and the point on which the hearing office therefore focused.
The applicant’s position was that “simply because the claimed method involves a simulation, that does not mean that it is automatically excluded from patentability”, that the current invention was comparable to the Haliburton decision where a method of designing a drill bit was held to be patentable (even without making the drill bit) and with the PKTWO decision, where a method of generating a more rapid and reliable notification was considered patentable. Although the garment itself may not be novel, the athlete was wearing the garment during performance of the specific route because it has been identified by the computer as having the most favourable performance indicator based on specific data relating to the garment, the specific route, and the athlete. This was not known and thus forms part of the contribution to human knowledge made by the application. The examiner disagreed and found that because the simulation is run on a standard computer and produces/controls nothing external to the computer, the contribution falls solely within the definition of excluded matter (specifically a computer program). The application was distinguished from Haliburton because that decision related to designing and manufacture of a new technical thing (a drill bit) and from PKTWO because there was no improvement in the selection method outside the computer in the case in question.
Considering these argument, the hearing officer sided with the applicant, noting that nothing in the Haliburton decision necessarily required a technical contribution to include design or manufacture. Regarding PKTWO, the hearing officer noted the need expressed in that decision to assess the invention as a whole and that, while merely producing a different screen display was too abstract to escape the exclusion, this is not the case if the effect outside the computer could fairly be described as a physical concept, process, or effect. As explained in the Haliburton decision, the deciding question was whether the effect outside the computer was itself within the exclusions (for examples a method of doing business) or not.
As a result, the hearing officer held that “once an in-computer simulation produces any form of technical solution applicable outside of the computer, then the objection that the invention is a program for a computer will fail except when the contribution falls within another category of excluded matter.” In the application in question, the invention did “not concern a method of doing business or presenting information or any other notionally abstract process” and hence the application was held not to be excluded subject matter and sent back to the examiner for examination of novelty and inventive step.
The decision is a useful clarification of the UKIPO’s position when it comes to inventions that involve optimising a real-world process inside a computer. It seems reasonable that this applies to doing so using machine learning, not just simulations per se, in line with the recent examination guidance for machine learning inventions (which is well worth a read). The decision clarifies that if there is an effect outside the computer and that effect is not itself in an excluded field, then the invention is not excluded from patentability. This should come in handy to get patents for machine learning inventions that are designed to affect or being taken into account in a real-world process that is not itself excluded subject matter.