by Dennis Crouch
The recent eligibility decision in AI visualization versus Nuance__ F.4th __ (Fed. Cir. 2024), gives me pause to consider the broader eligibility questions of AI inventions. When is the design or creation of AI system elements considered an eligible invention? In his recent article, Professor Nikola Datzov wrote what we are all thinking: “Innovative applications of AI are everywhere we look (and are revolutionizing our society). » Nikola L. Datzov, The role of patent (in)eligibility in promoting artificial intelligence innovation, 92 UMKC L. REV. 1, 4 (2023).
In AI visualizationthe Federal Circuit sided with the accused infringer in deeming the claims advanced ineligible under the two-step procedure Alice frame. AI Visualize had claimed four related patents that facilitated the use of a low-bandwidth web portal to visualize 3D/4D medical scans. The key here is to use virtual views and a system to determine which views have already been downloaded. Some claims require a unique identifiable key for each view; others use a tiered approach: first sending lower quality images for immediate viewing, followed by higher quality images. U.S. Patents No. 8,701,167; 9,106,609; 9,438,667; and 10,930,397. The district court dismissed the case outright, finding that the asserted claims were for subject matter ineligible for patenting under 35 USC § 101. AI Visualize, Inc. v. Nuance Commc’ns, Inc., 610 F. Supp. 3d 638 (D. Del. 2022). On appeal, the Federal Circuit affirmed.
At the first step, the court concluded that the claims addressed the abstract idea of ”retrieving remotely stored information requested by the user.” Although the claims refer to the creation of virtual views “on the fly,” the court found that “the language of the claim makes clear that the ‘creation’ of virtual views is achieved by manipulating a portion of (the existing dataset.
At the second step, the Federal Circuit agreed with the district court that “the allegations advanced involved nothing more than the abstract idea itself” and conventional computer technology. The court held that creating a virtual view was an abstract idea also known in the art, as admitted in the patent specifications and later in oral argument. Therefore, this limitation could not provide the inventive concept required to transform the claims into patent-eligible subject matter. During oral arguments, AI Vis attorney Rajkumar Vinnakota provided additional technical explanations, including how “new frameworks are going to be created to fill in the gaps, essentially, of what the user requested.” These frames are passed back to the local processor, in the case of a table or something like that, and these are then combined with the stored frames. (Oral jokes, 04:30-04:47). In its decision, however, the Federal Circuit concluded that these technical details were not part of the invention as claimed. Additionally, during oral arguments, Chief Justice Moore explained her view that the specification “describes the general idea of a user creating virtual views, but I see no details of how the server creates them.” actually implemented.” (Oral arguments, 05:53-05:59).
The latest study, focused on how claims language defined the “creation” of virtual views, has interesting implications for the eligibility of AI systems, particularly generative AI (GenAI) technologies. And, at first glance, AI visualization suggests that GenAI could face uphill battles for patent protection from applicants seeking functionally powerful claims.
A key factor will likely be how the “generation” of new AI content is defined in the claims. If, as in AI visualization, the claims require generating the new content by manipulating an existing data set, the Federal Circuit may well find that the claims are impermissibly directed toward an abstract idea. The solution, however, is to cite a particular innovative technological solution for creating new content. The closer the claims come to a concrete technical solution for how AI improves computer functionality, the better the prospects for eligibility.
In his article, Datzov suggests considering AI eligibility through a stack approach, or in academic parlance, a “three-tier taxonomy”: data, software applications, and hardware system. As you can imagine, eligibility becomes important the deeper you go into the stack; although many computer hardware claims were found ineligible when they relied on conventional computer components and the real innovation lay in the production of functional data. In other words, at all three levels, the key to eligibility is making explicit claims about how the technology works, rather than just its function or purpose. I have heard patent attorneys say that they are sending invention disclosures back to inventors more than ever for further development until technological improvement becomes apparent.
The bottom line of all this is that innovative AI systems are patented, but they must provide (and claim) a technical solution or solve a technical problem. This result naturally leads to narrower progressive patent claims – but that’s how the patent system works. We don’t get 11 million patents that each constitute an archetypal constellation. However, a second reality is that the current approach means that application layer developers are less likely to make patent-eligible inventions, while those who build the base models as well as those who coordinate the linking and layering AI systems have a much better chance. Finally, we must recognize that AI continues to be a buzzword without a precise definition: according to some figures, more than 10% of US patent applications are related to AI. This means that the types of inventions can vary in ways that are at least beyond my imagination. The bottom line, then, is that creative and competent patent drafting is more important than ever.
