Source: Norton Rose Fulbright

This publication by Norton Rose Fulbright analyzes a recent Patent Trial and Appeal Board (PTAB) rehearing decision concerning a foundational machine learning invention by DeepMind Technologies. The PTAB upheld rejections of patent claims under U.S. patent law sections 101 and 103, reaffirming that the invention was directed to an abstract idea without sufficient technical improvement to qualify for patent eligibility.

The decision highlights the Board’s approach to evaluating AI and machine learning patents, emphasizing that algorithmic steps and generic computing elements alone do not constitute patentable technological improvements. The article discusses the implications for AI developers, particularly those working on continual learning, transfer learning, and multi-task learning systems. It advises patent applicants to focus on specific technical solutions and detailed documentation to meet eligibility criteria and navigate examination challenges.