Much has been said recently about the usefulness of AI on patent drafting. On April 10, 2024, the USPTO issued guidance on the use of AI tools practitioners, reminding us that AI comes with the risk of a false factual premise and misplaced conclusion. Our signatures certify that “the legal contentions are warranted by law, the allegations and other factual contentions have evidentiary support, and the denials of factual contentions are warranted on the evidence.”
Consequently, ChatGPT is viewed as a good initial drafting tool as a starting point for human review. I interested in testing ChatGPT in the role of reviewer, seeking to perfect a human-drafted claim. The result was impressive, but first some background…
Recently in Rady v. The Boston Consulting Group, Inc., No. 2022-2218 (Fed. Cir. Mar. 27, 2024), the Federal Circuit affirmed invalidity of Max Rady’s US patent 10,469,250 under 35 U.S.C. § 101. The patent centers on 3D spatial mapping of gemstone “imperfections” to create a unique blockchain signature for authentication. 3D spatial mapping of gemstone imperfections was well-known, as was blockchain technology. However, Rady claimed a method of comparing by finding a 2D match of features and then rotating in virtual space to find another and confirm a 3D match. According to the Southern District of New York this failed Alice step 1 as being directed at the abstract idea of collecting, analyzing, and storing data. The Federal Circuit viewed the claims as “generic steps and results—as opposed to a specific solution to a technological problem” (citing Universal Secure Registry, 10 F.4th at 1355). However, Rady’s ‘250 patent clearly described a technological problem: “Blockchains breakdown when mapping to physical real-world items due to the requirement of having to trust some sort of the third party.” With regard to step 2 the Federal Circuit found no meaningful explanation as to how the various “item analysis components,” are combined in an inventive way, nor any plausible allegation that recording information about the unique imperfections of a physical object on a blockchain, rather than another type of ledger, supplies an inventive concept.
My possible takeaways? 1) detail technological problem(s) in the specification, describe how the technology solves it, and work that language into the claims; 2) explain specifically how the various hardware and data components are combined in an inventive way; 3) explain how recording 3D spatial mappings of gemstone imperfections on a blockchain is different than any other type of ledger. On 1 and 2 be more immersive, specify the contents and format of the unique signature from the spectral analysis data and 3D scan data with regard to working example(s), and make use of claim preambles and whereby clauses to reflect problem/solution. Perhaps for 3 perhaps explain why conventional blockchain ledgers don’t allow rotation along an axis to confirm that all N original imperfections sufficiently match. But enough about me, what would ChatGPTv3 do? I checked, half expecting a non-response. But with a little persistence and increasing granularity in my question I was eventually able to guide it into rewriting claim 1 to avoid Alice 101 invalidity. ChatGPT’s amendments are shown below…1. A network node comprising:
one or more processing devices;
1 Specifically, claim 1 recited “comparing the unique signature generated by the network node to previously recorded unique signatures using 3D spatial analysis techniques, rotating in virtual space features of the physical item defined in the unique signature to determine a match with features defined in the previously recorded unique signatures…”
2 (citing TLI Commc’ns, 823 F.3d at 615 and emphasizing that “vague, functional descriptions of . . . components are insufficient to transform [an] abstract idea into a patent-eligible invention”). Nor does Rady
a storage device, coupled to the one or more processing devices and storing instructions for execution by at least some of the one or more processing devices;
a communications subsystem, coupled to the one or more processing devices, to communicate with at least one or more other nodes of a peer-to-peer network; and
item analysis components coupled to the one or more processing devices, the item analysis components comprising at least one imaging device configured to determine spectral analysis data and 3D scan data from measurements generated by the item analysis components;
wherein the one or more processing devices operate to configure the network node to:
analyze perform an instance analysis of a physical item using the item analysis components to determine derive a unique signature for the instance item, wherein the unique signature is determined using utilizing 3D spatial mapping to define the unique signature from techniques applied to the spectral analysis data and 3D scan data generated by the item analysis components for the physical item;
determine, using ascertain, based on the derived unique signature, whether the instance of the physical item is has been previously recorded to registered within a blockchain maintained by the peer-to-peer network to provide for providing item tracking and authentication services, comparing the unique signature generated by the network node to previously recorded unique signatures using by employing 3D spatial analysis techniques, rotating in to compare the unique signature with previously registered unique signatures, wherein the comparison involves virtual space rotation of features of the physical item defined in the unique signature to determine a match ascertain conformity with features defined in the previously recorded registered unique signatures; and
record the instance of register the physical item within the blockchain upon determination that the physical item to the blockchain in response to the determining whether the instance is has not been previously recorded registered.
The ChatGPT takeaways? Use higher semantics. Rather than “analyze” –perform an analysis--, rather than “determine” –derive--, rather than “using” –utilizing--, rather than “determine” –ascertain--. Also layer the method steps, e.g., rather than individual steps recite a step and substeps such as “using 3D spatial analysis techniques to compare the unique signature with previously registered unique signatures, wherein the comparison involves virtual rotation of features of the physical item defined in the unique signature to ascertain conformity with features defined in the previously registered unique signatures..” Also use more specific rule-based cause and effect, e.g., not just “recording…to the blockchain in response to the determining whether the instance is previously recorded.”, but registering the physical item within the blockchain upon determination that the physical item has not been previously registered.”
I have no idea how ChatGPT gotthere, whether by analyzing Federal Circuit decisions, known valid claims, or simply implementing wisdom from published articles. Regardless, the result has some value. ChatGPT is clearly a source of good ideas to help a patent attorney draft an Alice-resistant claim.
