Written by Dennis Crouch
Recent eligibility decisions include: AI visualization and nuance__ F.4th __ (Fed. Cir. 2024), pause to consider the question of eligibility of AI inventions more generally. When does the design or creation of an AI system element qualify as an eligible invention? In a recent article, Professor Nikola Datsoff writes of what we have all been thinking: [and are] It will revolutionize our society. ” Nikola L. Datsov The role of patent (medium) status in promoting artificial intelligence innovation, 92 UMKC L. REV. 1, 4 (2023).
in AI visualizationFederal Circuit Court sided with the accused infringer in finding that the asserted claims were ineligible under the two-step doctrine Alice Framework. AI Visualize claimed four related patents that facilitate the use of low-bandwidth web portals for visualizing 3D/4D medical scans. The key here is to use virtual views and the system to determine which views have already been downloaded. Some claims require a unique and identifiable key for each view. Others use a stepwise approach, first sending lower quality frames for immediate display, then higher quality frames. U.S. Patent Nos. 8,701,167, 9,106,609; 9,438,667; and 10,930,397. The district court granted the asserted claims directed to patent-ineligible subject matter under 35 USC § 101 and invalidated the action ab initio. AI Visualize, Inc. v. Nuance Commc'ns, Inc., 610 F. Supplementary 3D 638 (D. Del. 2022). On appeal, the Federal Circuit affirmed.
In Step 1, the court concluded that the claim was directed to the abstract concept of “retrieving remotely stored information requested by a user.” Although the claim speaks of creating a virtual view “on the fly,” the court held that “the claim language does not imply that the 'creation' of a virtual view is accomplished by manipulating a portion of an existing view.” “This makes it clear that this is the case,'' he reasoned. [data set]”
In Step 2, the Federal Circuit agreed with the district court that the asserted claims involved nothing more than “abstract ideas themselves” and conventional computer technology. The court found that the creation of virtual views is an abstract idea known in the art, as acknowledged in the patent specification and subsequent oral argument. Therefore, that restriction fails to provide the inventive concept necessary to transform the claims into patent-eligible subject matter. During oral argument, Rajkumar Vinnakota, an attorney for AI Vis, provided more technical explanations, including the explanation that “basically, new frames are created in response to user requests to fill in the gaps.” was also included. For tables, etc., those frames are sent locally to his processor and combined with the stored frames. ” (Oral argument, 04:30-04:47). However, in its decision, the Federal Circuit concluded that these technical details were not part of the claimed invention. Additionally, during oral argument, Chief Justice Moore explained his view that the specification “states the following.''[es] There is a general concept of users creating virtual views, but no details about how the server actually implements them. ” (Oral argument, 05:53-05:59).
The final case focuses on how claim language defines the “creation” of virtual views, with interesting implications for the eligibility of AI systems, particularly generative AI (GenAI) technologies . And at first glance, AI visualization GenAI suggests that applicants seeking functionally strong claims may face an uphill battle for patent protection.
An important factor is likely to be how the “generation” of new AI content is defined in the claims. If, like, AI visualization, if a claim requires manipulating an existing data set to generate new content, there is a good chance that a Federal Circuit will find that the claim is directed to an abstract idea. . However, the solution is to recite certain innovative technological solutions for creating new content. The closer your claim is to a concrete technical solution for how AI improves the capabilities of computers, the better your chances of eligibility.
In his paper, Datsov suggests considering AI eligibility through a stack approach, or in academic terms, a “three-tier taxonomy” of data, software applications, and hardware systems. As you can imagine, the deeper you dig into the stack, the more you qualify. However, many computer hardware system claims have relied on traditional computer components and have been disqualified where the true innovation is found in the functional data output. In other words, at all three levels, the key to eligibility is to make clear claims about how the technology works, not just its function or purpose. I have heard from patent attorneys that they are sending more invention disclosure information back to inventors than ever before, asking for further development until improvements in technology become apparent.
One conclusion of all this is that innovative AI systems, while patented, must either provide (and claim) a technical solution or overcome a technical problem. This naturally results in narrower claims, but that's the way the patent system works. We don't have 11 million patents, each a typical collection. Still, the second reality is that with current approaches, application layer developers are unlikely to achieve patent-eligible inventions, whereas developers building the underlying models and AI systems are less likely to collaborate and layer. It means there are far greater opportunities for developers who tune in. Finally, we must recognize that AI remains a buzzword without a strict definition. According to some statistics, more than 10% of patent applications in the United States are related to AI. What this means is that the types of inventions can change, at least in ways beyond my imagination. As a result, creative and skilled patent drafting is more important than ever.
