Webinar: Patenting perfumes and fragrances
Register for this 45 minute webinar on 12th September 2024.
In a significant ruling, the Court of Appeal recently (19th July 2024) addressed the patentability of Artificial Neural Networks (ANNs) in the context of the UK Patents Act 1977. This judgment clarifies the legal interpretation of what constitutes a “program for a computer” and whether ANNs as such fall within the exclusions from patentability.
The Court of Appeal found that the claimed invention, which involved recommending files based on their semantic similarity, is excluded from patent protection. In doing so, the Court of Appeal overturned the decision of the High Court and thereby upheld the original decision of the Hearing Officer.
The case at hand involved a patent application for an ANN-based system designed to provide recommendations for files (such as music) based on their semantic similarity. The application was initially rejected by the Hearing Officer on the grounds that it fell within the exclusion for computer programs “as such” under section 1(2) of the Patents Act 1977. Emotional Perception AI Limited (EPL), the applicant, appealed this decision to the High Court, where Sir Anthony Mann ruled in their favour. The Comptroller then appealed this decision to the Court of Appeal.
The central issue was whether the ANN’s set of weights and biases constituted a “program for a computer“. The Comptroller argued that these weights and biases are indeed a set of instructions for the computer, thus falling within the definition of a computer program.
The court examined various dictionary definitions of “computer program” and concluded that a program is essentially a set of instructions that a computer follows to perform a task. This interpretation was supported by previous cases such as Gale’s Application [1991] and Aerotel [2007].
EPL contended that the unique aspect of ANNs lies in their ability to solve problems that are intractable for traditional programming. They argued that since the structure and function of an ANN are determined through a training process rather than explicit human-written instructions, it should not be considered a computer program.
The court dismissed this argument, emphasizing that the manner in which instructions (weights and biases) are generated—whether by a human programmer or through machine learning—is irrelevant to their classification as a computer program.
The court examined whether the output of the ANN (file recommendations) constituted a technical effect. The Hearing Officer had concluded that the improved recommendation based on semantic similarity was subjective and cognitive, therefore not providing a technical contribution.
The High Court judge had initially found that the system’s method of selecting and recommending files involved a technical contribution. However, the Court of Appeal disagreed, stating that the semantic qualities of the recommended files are aesthetic and cognitive, thus not amounting to a technical effect.
The judgment considered analogies with previous cases like Protecting Kids the World Over [2011] and Gemstar v Virgin Media [0009]. It concluded that these cases did not provide useful parallels due to the superficial similarity and the non-technical nature of the file recommendation process in the present case.
The key part of the judgment comprises paragraph 79 of the decision, which summarises the reasoning and also suggests a level of conformance between the UK and EPO approaches:
“79. In my judgment the Hearing Officer’s conclusion is the right one. What makes the recommended file worth recommending are its semantic qualities. This is a matter of aesthetics or, in the language used by the Hearing Officer, they are subjective and cognitive in nature. They are not technical and do not turn this into a system which produces a technical effect outside the excluded subject matter. I note that the same view was expressed by the Technical Board of Appeal of the EPO in Yahoo T 0306/10, at paragraph 5.2 in holding whether song recommendations are “good” or “bad” does not amount to a technical effect. EPL make the point that this case was concerned with inventive step but that is only an artefact of the difference in the way the EPO approaches patentability from the manner in which it is approached in this jurisdiction. It does not undermine the relevance of the Board’s observation.”
The Court of Appeal upheld the original decision that the ANN-based recommendation system is excluded from patentability. The court clarified that an ANN’s weights and biases, whether implemented in hardware or software, constitute a computer program. The recommendation process, driven by semantic similarity, does not involve a technical effect that could escape the exclusion.
This ruling emphasizes that the mere fact that an invention involves advanced computational methods or machine learning does not automatically render it patentable. The key takeaway is that for computer-implemented inventions, including those using ANNs, the focus must be on whether the invention provides a technical contribution beyond the excluded subject matter.
Applicants seeking patents for inventions involving ANNs or similar technologies should carefully consider how their claims are framed. Emphasizing technical contributions that go beyond mere data processing or subjective improvements is crucial. This judgment serves as a guidepost for navigating the complex landscape of patent exclusions and ensuring that applications meet the necessary criteria for patentability.
You can read the full judgment here.
If you have a question about ANNs or for any other advice on patent applications please contact one of our attorneys.