The Machine’s Mind Matters: Using Logic to Disclose Artificial Intelligence Inventions

Abstract Inventors must disclose their inventions in a clear, concise and complete manner to enable skilled artisans to practice the invention without undue experimentation. AI inventions may not fulfil this requirement for various reasons, such as the unintelligibility of AI internal mechanisms. Inventors are currently required to disclose components of AI inventions, such as datasets, algorithms, nodes, parameters, outputs, etc. I term the disclosure of these components as structural disclosure. However, structural disclosure is marred by uncertainties, and it is not clear what exactly constitutes the structure of AI inventions for purposes of disclosure. This article argues that logic – which I conceptualise as a structured understanding of the internal mechanisms of AI systems that unveils the relational link between the inputs and outputs – can resolve stark disclosure challenges. Based on what I term as functional disclosure, I contend that inventors should be incentivised to disclose the logic of AI inventions. This article makes three scholarly contributions. First, it establishes the important role that logic plays in AI inventions and, subsequently, positions logic as the vehicle through which inventors can surmount the limitations of structural disclosure and fulfil the disclosure requirement more satisfactorily. Second, it problematises patent law’s indifference toward the logic of inventions and demonstrates how this indifference undermines the teaching role of patent law. Ultimately, borrowing from the market exclusivity regimes for pharmaceutical inventions, this article lays down a balanced framework for implementing functional disclosure without threatening the bargain embedded in the disclosure doctrine.