The current Guidelines for Examination at the EPO consider AI and Machine Learning to be pure mathematical methods (see Guidelines, G-II,3.3.1), in contrast to other areas of computer technology such as database technology (see Guidelines, G-II3.6.4). While I have argued previously that this differential treatment ignores the technological reality of our days, in this article I will take a closer look at an aspect of the provisions in the Guidelines (specifically G-II, 3.3) in relation to the patentability of mathematical methods, and hence at present AI and ML. ​​ G-II, 3.3 provides two routes to patenting an invention based on mathematical methods. Mathematical methods are, as such, not considered to be technical and hence not capable of supporting an inventive step. Mathematical methods that have technical character in that they involve technical considerations and contribute to the solution of a technical problem are, however, considered to contribute to an inventive step. The two routes for a technical character to be recognised for a mathematical method are a technical purpose (viz. the application of the method) or its technical implementation. I will come back to the former some other time and focus on the latter here. ​​ ​​On the technical implementation route, G-II, 3.3 provides the following:

A mathematical method may also contribute to the technical character of the invention independently of any technical application when the claim is directed to a specific technical implementation of the mathematical method and the mathematical method is particularly adapted for that implementation in that its design is motivated by technical considerations of the internal functioning of the computer (T 1358/09).

This route is attractive in that it allows in principle to protect inventions of a fundamental nature that are not limited to specific purposes or applications. However, the reference to a specific technical implementation that the mathematical method is particularly adapted risks being interpreted narrowly leading to a restrictive approach in regards to this route. While practice on this seems to be not settled, the example given in the guidelines and the notion in the remainder of the relevant section that computational efficiency is not enough if the implementation of the mathematical method does not go beyond a generic technical implementation seems indeed to have led to a restrictive approach taken, for example requiring specifically adapted hardware in some cases.

Where does this restrictive interpretation of a specific technical implementation come from and is it justified by the case law of the EPO’s Board of Appeal? ​​ ​​The guidelines refer to T 1358/09 in respect to the specific technical implementation. However, there is no basis for this notion in this decision, which is indeed much broader in its consideration of technical implementations (see 5.5 of the reasons). Applying this decision correctly, the paragraph cited above would read much more broadly:

A mathematical method may also contribute to the technical character of the invention independently of any technical application when the claim is directed to a specific technical implementation of the mathematical method and the mathematical method particularly suitable for being performed on a computer is particularly adapted for that implementation in that its design is motivated by technical considerations of the internal functioning of the computer (T 1358/09).

​​

​​Is there anything else in this decision that would justify the more restrictive language in the Guidelines?

Contrary to the remainder of the section in the Guidelines, T 1358/09, 5.5 of the reasons for the decision, explicitly recognises that:

​​…not all efficiency aspects of an algorithm are by definition without relevance for the question of whether the algorithm provides a technical contribution.

This would seem to indicate that the approach of requiring a specific technical implementation in the absence of which any efficiency gains, such computational resources in terms of speed or memory requirements, are to be ignored is more restrictive than envisaged by the Board of Appeal. In the context of this, the T 1358/09 refers to the earlier Board of Appeal decision T 258/03 and the Enlarged Board of Appeal decision G 3/08. In the former, the efficiency gain that was to be ignored stemmed from changing the rules of an auction and in the latter the relevant passage states that although

​​… all computer programming involves technical considerations since it is concerned with defining a method which can be carried out by a machine, that in itself is not enough to demonstrate that the program which results from the programming has technical character; the programmer must have had technical considerations beyond "merely" finding a computer algorithm to carry out some procedure. (emphasis added)

While it is clear that the established case law of the Boards of Appeal requires more than just finding a (any) computer algorithm to carry out some procedures, the restrictive approach of ignoring all algorithmic (computational) efficiency consideration in the absence of a technical purpose or a specific technical implementation goes beyond what has been envisaged by the Boards of Appeal. This risks encouraging an approach to examining mathematical method inventions, and in particular AI and ML inventions, that is too restrictive and risks excluding from patentability many methods that are the result of technical considerations beyond merely finding a computer algorithm to carry out some procedure.

It is an important aspect of the Enlarged Board of Appeal to ensure the consistent application of the European Patent Convention (in full the “Convention on the Grant of European Patents” - note convention on the grant, not refusal, of European Patents). It is therefore of concern that the Guidelines take a more restrictive approach than that which the Enlarged Board of Appeal has found to be applied by the Boards of Appeal without divergence. Is this stance justified by case law since G 3/08?​

The latest word on this is another Enlarged Board of Appeal decision, G1/19. This decision specifically concerns itself with computer simulation but, importantly for the question at hand, takes again stock of the case law relating to the patentability of computer software and mathematical method. This decision sets the tone by noting that “the term "technical" must remain open, not least in anticipation of potential new developments”. G1/19 considers in non-exhaustive form how and when "technical effects" or "technical interactions" may occur in the context of a computer-implemented process, so that they can contribute to an inventive step. Relevant to the question of “technical implementation”, G1/19, at 85 of the reasons, notes that

​​[a]daptations to the computer or its operation, which result in technical effects (e.g. better use of storage capacity or bandwidth), are also examples of features that may contribute to inventive step … technical effects can occur within the computer-implemented process (e.g. by specific adaptations of the computer or of data transfer or storage mechanisms)

​​​ G1/19 goes on to refer to T 697/17, Reasons 5.2.5, for a list of examples and references to the relevant board decisions. In brief form, the examples in that decision include:

  • ​​reducing the amount of data to be stored or transmitted

  • ​​searching for low-cost query execution plans using a cost estimate for the computer resources

  • ​​the purposive use of information modelling in the context of a solution to a technical problem

  • ​​an index file used for the purpose of controlling the computer "along the path leading to the desired data"

  • ​​a RAM-based hash table of fingerprints of stored URLs was used, in the context of web crawling

  • ​​search indexes used to provide access to stored data

  • ​​specific bit strings and matrices and respective operations chosen to efficiently perform in parallel the steps of a method for evaluating selection conditions​

​​More generally, T697/17, endorsed in G1/19, finds that

​​features make a technical contribution if they result from technical considerations on how to for instance improve processing speed, reduce the amount of memory required, improve availability or scalability, or reduce network traffic, when compared with the prior art or once added to the other features of the invention

​​​but noting that

​​[o]n the other hand, such effects and the respective features are non-technical if the effects are achieved by non-technical modifications to the underlying non-technical method or scheme (for example, a change of the business model, or a "pure algorithmic scheme", i.e. an algorithmic scheme not based on technical considerations.

​​In summary, then, it is the established case law of the Boards of Appeal that algorithms based on technical considerations, such as processing speed, memory requirement, availability and scalability, must be considered for their contribution to an inventive step. What must be disregarded as non-technical are modifications of an underlying non-technical method or scheme, that is an algorithmic scheme not based on technical considerations. This is perhaps best illustrated by the example in T 258/03, specifically reasons 5.7 and 5.8, in which changing the rules of an auction circumvented or rendered moot, rather than solved the technical problem, and was found not to contribute to an inventive step. ​​ ​​The current approach set out in the Guidelines risks encouraging Examiners to ignore all aspects of computational efficiency for the purpose of assessing inventive step. This risks ignoring adaptations to computer operations that result in technical effects such as better use of storage or bandwidth, contrary to the analysis in G 1/19. It risks penalising algorithms that are based on technical consideration that go beyond merely finding a computer algorithm to carry out some procedure, contrary to G 3/08. It remains to be hoped that the Guidelines will be revised in time to take account of this discrepancy and that in the meantime, Examiners apply this section in a way that is consistent with the case law of the Boards of Appeal.