'AI-Generated Inventions': Time to Get the Record Straight?

Abstract

This article attempts to clarify the notion of an ‘AI-generated’ invention, an issue which has triggered an intense debate on the future of patent law and policy. While there is a general consensus that such inventions are incompatible with the concept of human inventorship, it remains largely unclear to what extent concerns regarding ‘non-human’ ingenuity can be justified. Most uncertain is how AI ‘autonomously generates’ inventions, and in what way ’AIgenerated’ inventions differ from inventions developed with the aid of AI. Drawing on the extensive literature review, this article depicts AI techniques as methods of computational problem solving. It emphasises that such methods should not be equated with a computer’s ‘cognitive autonomy’. Further, it clarifies that the types of AI that have been most debated in the patent law literature – artificial neural networks and evolutionary algorithms – essentially require detailed instructions that determine how the relation between inputs and outputs is derived through computation. Accordingly, it is argued that, as long as computers rely on instructions defined by a human as to how solve a problem, the separation between human and non-human (algorithmic) ingenuity is, in itself, artificial. Ultimately, the article calls for a broader technical inquiry that would elucidate the relevance of the currently debated normative concerns over ‘non-human inventorship’ against the background of the technological state of the art.