Understanding how AI inventions will be treated thus requires an understanding of how CI inventions are currently treated by the EPO. Indeed, the application is a step closer to successful prosecution if the invention is applied to a field of technology, for example, signal processing or medical analysis. What happens however, when AI is applied to “non-technical” fields like business and commerce?
Although the guidelines are clear on the treatment of CI or the mixed-type inventions, we see a general hesitation to recognise the technical contribution of allegedly non-technical features, even if they combine with the remaining technical features of the claim and produce a technical effect. In many instances, such non-technical features relate to business related fields like advertising, banking transactions, social networking etc. This occasionally results in the problem-solution approach for inventive step assessment boiling down to a comparison of the invention to a general purpose computer.
The COMVIK decision (T0641/00) has led to the creation of a benchmark in the assessment of CI inventions, to the extent that “COMVIK approach” has become a familiar term in art. Although the expression “computer-implemented” may be trivially used to escape the bounds of Article 52 (2) EPC regarding exclusion of patentability, the Applicant often encounters a tough time trying to convince the Examiner of Article 56 EPC, the inventive step. Indeed, the objection merely shifts from one article to another in these cases.
A widely seen approach of the EPO Examiner in the assessment of inventive step of a mixed-invention type claim, is to strike off (literally) those features in the claim that are technical. These may be a server, processor, memory, computer and the like. An example from recent days is the G1/19 referral, where the Examining Division had struck off the expression “computer-implemented” in claim 1 of the main request, and implied that the remaining subject-matter was a mental act requiring no involvement of a technically skilled person.
Now, if the technical features in the claim which were struck off merely amount to a processor or a memory, the closest prior art is almost always chosen as a general purpose computer. The rest of the argumentation then mostly rests on the technical implementation being an alleged automation of the non-technical method using a computer. As if things could not get worse, an allegation of notoriously known hardware produces no prior art, which means that the search fee paid by the Applicant goes down the drain! A few examples of common general knowledge may be cited, the documents comprising a computer with basic functions.
The process of patenting an AI or a CI invention in one of the non-technical domains can be made relatively easier, if attention could be paid to a few aspects while drafting the application. Focus can be given, for example, to the technical implementation of the invention. In a recent decision T 0697/17, the Board of Appeal (BoA) of the EPO reversed the decision of the department of first instance regarding inventive step of the invention. The decision concerned a method which related to a database management system (DMS) and wherein the independent claim disclosed specific technical means, viz., a parser, a query optimiser and a query execution engine, to support update operations on data stored and managed by the DMS. This implementation was found by the Board to make a technical contribution over a general-purpose computer, which was (and not surprisingly) considered the closest prior art by the Examining Division. According to the BoA,”[6.1].. Starting from a general-purpose computer, the skilled person would first have to add the technical functionality of a relational database system and then add more features to it…. Since a general purpose computer is not even remotely related to the specific problem of the invention and does not include any of the claimed features, it is not a realistic starting point for an attack on inventive step.”
While detailing the technical implementation, it is therefore useful to emphasize sub-modules of the technical means which steers the invention, and how these sub-modules functionally interact with one another to bring about the technical effect(s) of the invention. After all, an AI algorithm that is specially adapted to carry out this specific implementation can then not be any “suitable” algorithm selected for automation of a “non-technical” method!
Secondly, emphasizing technical effects in the application can clearly be of advantage. It should be borne in mind however, that an inventive step objection will likely surface if the effect were to reside in considerations involving user preferences. Furthermore, technical effects which result from considerations inherent to the non-technical application are not considered sufficient to establish an inventive step. Consider a technical effect of resource economy, which results from having to process a lesser amount of data than that conventionally required. Assume now that the data relates to a non-technical application, say, calendar data, in a method which deals with organising business calendars. In this case, the effect will be considered as the result of a consideration which lies inherently in the non-technical scheme of calendar organisation.
On the other hand, an algorithm with a technical effect which is a result of considerations involved in the [technical] implementation of the "non-technical" scheme or those involved in the internal functioning of the computer (further technical effects) is taken into account for assessment of inventive step. In a mixed-type business-related invention, such specific technical considerations cannot be present as part of the business specification that a business person may give a technically skilled person or a computer programmer as such to implement (T 0144/11, T 1463/11). After all, a business person cannot possibly be aware of the specifics in the involved technology, and a computer programmer as such can merely find a suitable computer algorithm to automate the allegedly non-technical method!
Thus, the Applicant can benefit from a more effective prosecution of an AI (or CI) business-related invention, if the application details considerations inherent to the specific technical implementation of the business method, and effects which follow these [technical] considerations.