Landmark Patent Appeal Decision Strengthens Protection for AI and Machine Learning Innovations

On September 26, PTAB’s Appeals Review Panel (ARP) issued a significant decision addressing the patent eligibility of claims of a patent application filed by Google’s DeepMind Technologies directed to methods for training ML models. The decision vacates a prior ground of rejection under 35 U.S.C. § 101, affirming that the claims at issue, when considered as a whole, are not directed to a patent-ineligible abstract idea but instead reflect a technological improvement in the field of machine learning.

Claimed Invention

The patent application at the center of this decision concerns a computer-implemented method for training a ML model on multiple tasks. The method involves determining the importance of model parameters to a first task, then training the model on a second, different task while protecting performance on the first task. This is achieved by adjusting model parameters to optimize an objective function that incorporates a penalty term based on the previously determined importance measures.

The specification highlights several technical advantages, including reduced storage requirements, lower system complexity, and the ability to preserve performance across sequential tasks, which addresses the well-known challenge of “catastrophic forgetting” in continual learning systems.

Procedural History

The Board initially affirmed the rejection of all pending claims under 35 U.S.C. § 103 and introduced a new ground of rejection under § 101, finding the claims directed to an abstract idea, specifically, a mathematical calculation. The applicant sought rehearing, arguing that the claims provided a technical improvement in ML technology.

Legal Framework

The ARP applied the two-step framework established in Alice Corp. v. CLS Bank and further articulated in the Manual of Patent Examining Procedure § 2106. The analysis first considers whether the claims are directed to a judicial exception (such as an abstract idea) and, if so, whether additional elements integrate the exception into a practical application.

Key Findings

• The ARP agreed that the claims recite an abstract idea in the form of a mathematical calculation. However, the analysis did not end there.
• Upon review, the ARP found that the claims, when considered as a whole, integrate the abstract idea into a practical application by improving the operation of the ML model itself.
• The decision emphasized that the claims address technical challenges in continual learning, such as preserving knowledge from previous tasks and reducing storage needs, which constitute improvements to computer technology.
• This decision underscores that claims directed to ML methods may be patent-eligible when they provide a specific technological improvement, rather than merely reciting abstract mathematical concepts or generic computer implementation.
• The decision also highlights the continued relevance of §§ 102, 103, and 112 as the primary statutory tools for assessing the scope and validity of patent claims, rather than relying on § 101 to categorically exclude innovations in AI and ML.

Implications

While the decision addresses the legal framework for patent eligibility of AI inventions, its broader significance lies in the clear policy direction set by the USPTO’s leadership regarding the patentability of AI and ML inventions. With this decision, the USPTO signals strong support for protecting AI innovations through the patent system.

This approach encourages inventors and companies to pursue patent protection for advances in AI and ML, fostering increased investment and growth in these fields. The decision reflects a commitment to ensuring that US patent policy keeps pace with technological developments, maintaining America’s leadership in AI by providing robust incentives for innovation and reducing uncertainty around the eligibility of AI-related inventions. As a result, the USPTO is likely to see a continued rise in AI patent filings, reinforcing the nation’s position at the forefront of emerging technologies.