Patent System and Artificial Intelligence: Towards a New Concept of Inventorship?

The objective of this work is to answer the question whether an AI can be considered an inventor, as can a human, through a methodological approach which analyses different documents that are mostly secondary sources, but also case-law and legislation. The answer is negative: there is no such thing as a new concept of AI inventorship for now. In particular, although there have been attempts by some authors - defined as the “classic literature” - to consider AI as creative and thus capable of generating inventions (the so-called “AI-generated” inventions), a more careful “technical” literature states that AI systems operate through a different intelligence than the human one, and this philosophical difference can be practically envisaged not only in the current case-law of the EPO, but also in the way machines operate in our reality. Indeed, the computational problem solving mechanism requires the human contribution, especially in the phases of abstraction/modelling, defining an algorithm and programming. Therefore, even the most sophisticated soft-computing methods, such as ANNs and EAs, cannot be considered autonomous. However, this work will not completely underestimate the possibility that in the future there could be something such as an AI inventorship. Unfortunately, not only the very important incentive justification but also other classic IP theories (fairness, personality, and culture) would not be compatible with this hypothetical AI inventorship. As a consequence, the current patent system should be reformed through the implementation of a tailoring approach. The problem is that, in order to do so, legislators and judges should be aware of the optimal patent strength of each industry. However, the information about R&D costs, risk of failure, and level of innovation, is very difficult to obtain. Given this impossibility to reform the patent system, other ways through which AI inventorship can be protected will be mentioned. This working paper is a part of the "Outstanding LLM Dissertations 2021"