AI is a field of technology – to most of us, this statement would not be controversial As reported in a recent survey by McKinsey, 55% of participants (1,843 participants representing the full range of regions, industries, company sizes, functional specialties, and tenures) said their organisation had adopted AI in at least one function. The most common use cases included service operations optimisation and product enhancements. For more than a quarter of respondents, AI adds at least 5% to earnings and resulted in at least 10% cost savings for 79% of respondents. Two thirds of respondents say prospects for more investment in AI remain strong. What is more, good engineering practices seem to be one of the main drivers distinguishing AI high performers from the rest of the pack. (AI high-performers are defined by the survey as those companies that gain 20% on their earnings due to AI) from the rest of the pack. Clearly that is technology at its best making big differences to adopters’ bottom line (if not without its ethical and regulatory challenges as is increasingly part of the conversation about AI).
If you are not familiar with the intricacies of protecting software inventions at the EPO, you might be surprised that this is not the view the EPO takes. The EPO guidelines for examination have this to say:
“Artificial intelligence and machine learning are based on computational models and algorithms […that] are per se of an abstract mathematical nature”
That sounds fair enough but in the context of how computer-implemented inventions are examined at the EPO, it means that improvements to AI technology itself are difficult to protect broadly as we need to show that an advance over the state of the art is motivated by the internal functioning of computers. Otherwise, application-independent improvements in AI are not per se taken into account for the assessment of inventive step (if you want to know more about what that may mean and why it is important, you can look at some of the previous stuff I have written). However, it does not have to be that way because AI being treated as abstract math is not dictated by EPO statute (the European Patent Convention), nor does it seem to be set in stone in the case law of the EPO’s Boards of Appeal.
What is more, comparing AI technology with relational database technology, or as the EPO puts it “database management systems” in its guidelines. Of interest here is that the EPO guidelines treat database management systems as “technical systems on computers to perform the technical task of storing and retrieving data using various data structures for efficient management of data”. As a result, “features specifying the internal functioning of database management systems […] are taken into account for the assessment of inventive step”.
Why is there a different (easier) patenting standard for database management technology compared to AI technology? Both technologies retrieve data, in one case from a structured data store, in the other case by adapting a model to retrieve information present in its input in an unstructured way (for example identifying an image of a cat as a cat or a spam email as spam). Both technologies have enormous societal and economic impact and require specialised engineers to implement and develop. At which point do database management and AI part company so that one becomes a technical field and the other abstract math?
It took about 30 years from the widespread adoption of database management systems in the 1980s to gain their recognition as technical by the EPO. Here is hoping that realising the nature of AI as an engineering discipline will happen quicker so that applicants and society can better reap in the benefit of broad patent protection for an already ubiquitous field of engineering, without having to rely on clever legal argument.