The Role of Functional Language in AI Patent Claims – IPWatchdog.com

This post was originally published on this site.

As artificial intelligence (AI) continues to grow in various industries, securing patents for AI technologies is becoming more important. However, protecting inventions utilizing machine learning, neural networks, and other AI techniques can be challenging, and potentially implicates claim construction issues under 35 U.S.C. § 112(f), which addresses means-plus-function claim limitations and can impact how AI inventions are interpreted during the patent process and enforcement. It’s helpful for patent practitioners to understand how § 112(f) applies to AI inventions and to develop strategies for navigating these complexities. This article provides an overview of recent Patent Trial and Appeal Board (PTAB) decisions that highlight how § 112(f) has been applied to AI-related inventions.

Background of 35 U.S.C. § 112(f)

The U.S. Patent and Trademark Office (USPTO) has underscored the importance of connecting functional claim language to specific structures, especially in computer-implemented inventions. According to the 2019 guidelines, when a claim includes a functional limitation like “means for processing data,” the specification must clearly identify the structure responsible for performing that function—whether it’s an algorithm, processor, or hardware component.

35 U.S.C. § 112(f) states: “An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.”

When determining whether § 112(f) applies, Examiners use a three-prong test (see MPEP 2181):

(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;

(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and

(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.

Recent Cases

1. Ex Parte Joon Woo Son, Appeal No. 2023-000761, 2024 WL 3175795 (PTAB March 19, 2024). The application is titled “Apparatus for Autonomous Driving Algorithm Development Using Daily Driving Data and Method for Using the Same” and purports to describe an improvement to autonomous driving. Claim 1 recites:

1. 1. 1. An apparatus for autonomous driving algorithm development comprising: one or more processors configured to:

[a] receive daily driving data including a front photographed image, GPS information, acceleration sensor data, and vehicle driving data of a first vehicle from a terminal mounted inside the first vehicle;

[b] perform preliminary machine learning using the daily driving data to evaluate accuracy of the daily driving data, and store the daily driving data in a connected database if the accuracy above a threshold is calculated;

[c] pre-process the daily driving data by matching view angles of a road image of the front photographed image in the daily driving data stored in the database with an image conversion technique, and converting the acceleration sensor data into acceleration components;

[d] learn the pre-processed daily driving data by applying it to an autonomous driving algorithm through a preset learning engine; machine

[e] reconfigure, using a learned autonomous driving algorithm, the autonomous driving algorithm according to a route to be currently driven; and

[f] provide a command for controlling autonomous driving, which corresponds to a reconfigured autonomous driving algorithm, to a second vehicle to be currently driven, the one or more processors are configured to: [g] when a learned model of the machine learning corresponding to the route to be driven exists, select an autonomous driving algorithm of the learned model as the reconfigured autonomous driving algorithm; and

[h] when a learned model of the machine learning corresponding to the route to be driven does not exist, subdivide the route to be driven into subdivided routes, and combine individual autonomous driving algorithms corresponding to the subdivided routes respectively so as to reconfigure the autonomous driving algorithm customized to the route to be driven.

In this case, the issue was whether the term “learned autonomous driving algorithm” should invoke § 112(f). The PTAB found that the term “algorithm,” when used without additional details, connoted no more structure than the word “means.” It reasoned that “algorithm” is akin to generic terms like “device” or “mechanism,” which do not impart any specific structure. Additionally, the phrase “learned autonomous driving algorithm” did not inform skilled artisans on how the algorithm performed the claimed function of reconfiguring autonomous driving algorithms based on a route. Due to the lack of structural disclosure, the PTAB concluded that § 112(f) applied, and the claims were indefinite under § 112(b).

Takeaway: This case highlights the importance of providing adequate structural support for functional claims, especially when using terms like “algorithm” in AI patents. Simply reciting functional language or generic algorithmic terms is insufficient unless accompanied by detailed explanations or specific algorithmic structures in the specification. When drafting AI-related claims, be sure to define underlying structures, such as specific algorithms or processing steps, to avoid invoking § 112(f) and potential indefiniteness rejections.

2. Ex Parte Jing Lan, Appeal No. 2023-003812, 2024 WL 2798289 (PTAB, May 31, 2024)

The application is a system and method for a machine learning based approach for classification of encrypted network traffic data. Claim 1 recites:

1. 1. 1. A method comprising:

            receiving, by a processor of a network security device, a stream of packets representing a network flow;

determining, by the processor, metadata relating to the stream of packets;

matching, by the processor or by a pattern matching and regular expression matching module of a hardware acceleration subsystem of the network security device, application layer payload data of one or more packets of the stream of packets against string patterns and regular expression patterns;

collecting, by the processor or by the hardware acceleration sub-system, statistics relating to the application layer payload data; and

classifying, by the processor, the network flow as being associated with a particular network service of a plurality of network services by applying a machine-learning model to the metadata, results of said matching, and the collected statistics to yield a classification.

The dispute in this case centered around the term “processing resource.” The Examiner argued that the term invoked § 112(f) because “resource” was a generic placeholder lacking specific structural meaning. The Examiner contended that functional language following the term required structural support in the specification.

The PTAB disagreed with the Examiner, finding that “processing resource” did not invoke § 112(f). The Board reasoned that although “resource” is a generic term, the claim context and specification indicated that “processing resource” referred to specific electronic circuitry. The claim recited sufficient structure, and the specification supported this interpretation, overcoming the presumption of § 112(f).

Takeaway: Not all functional terms automatically invoke § 112(f). The claim context and disclosure in the specification can rebut the presumption of § 112(f). In this case, “processing resource” was sufficiently tied to a known structure—electronic circuitry—to avoid being considered purely functional. When using potentially functional placeholders like “resource” or “module,” ensure that the claim or specification links the term to a known structure.

3. Ex Parte Melvin Lopez and Jessie Rincon-Paz Appeal No. 2022-003718, 2024 WL 244337 (PTAB, January 18, 2024)

The application relates to systems and methods for prioritizing and tracking incidents and changes that occur in an information technology infrastructure. Claim 1 recites:

1. 1. 1. A system for intelligent incident and change management, the system comprising:

a support resource displaying a user interface;

an active machine learning module configured to monitor an information asset to detect an error in the functioning thereof and generate an incident report based on the detected error;

 a ticketing system communicating with the active machine learning module and configured to

receive the incident report of the detected error transmitted from the active machine learning module,

transmit the incident report to the support resource for display on the user interface, and maintain a status of the incident report; a configuration and orchestration engine configured to repair the information asset with the detected error in response to initiation thereof by the active machine learning module,

communicate with the ticketing system to cause the ticketing system to change the status of the incident report from an open status to an acknowledge status when the repair is initiated,

verify repair of the information asset with the detected error reported in the incident report, and

communicate with the ticketing system to cause the ticketing system to change the status of the incident report to a resolved status when repair of the information asset with the detected error in the incident report has been successful, and

wherein the active machine learning module is further configured to:

reassign the incident report and attempt-failure data associated therewith for investigation by the support resource when the

status of the incident report does not reach the resolved status,

receive application data;

receive information asset data; and

determine and assign priority to the detected error based on the application data and information asset data;

wherein the generated incident report is based on the detected error and the assigned priority.

The issue was whether several terms in the claims, including “an active module” and “a configuration and orchestration engine,” invoked § 112(f). The Examiner determined that these terms acted as generic placeholders paired with functional language without reciting sufficient structure to perform the stated functions, thus invoking § 112(f).

The PTAB agreed with the Examiner, noting that the term “module” is a known “nonce” word that acts as a substitute for “means” under § 112(f). The Board explained that the terms were described only in functional terms—e.g., “monitor an information asset,” “reassign the incident report,” and “determine and assign priority.” These functions lacked accompanying structural details in the claims, leading the PTAB to conclude that § 112(f) applied. However, the Board found that the disclosed structure in the specification was insufficient to support these functional limitations. The specification merely described a “processor in communication with a memory,” which the PTAB deemed as a general-purpose computer. The Board emphasized that for computer-implemented inventions, disclosing a general-purpose computer without an algorithm does not satisfy the requirement of a corresponding structure under § 112(f). Without a disclosed algorithm to perform the claimed functions, the specification failed to provide adequate structural support.

Takeaway: This case illustrated the importance of providing more than just a general-purpose computer as a corresponding structure in means-plus-function claims involving computer-implemented inventions. To avoid indefiniteness under § 112(f), practitioners must disclose specific algorithms or specialized structures in the specification that perform the claimed functions. When using nonce words like “module” or “engine,” ensure the specification clearly links these functional terms to a detailed structure, such as an algorithm or specific hardware configuration.

4. Crusoe Energy Systems LLC v. Upstream Data Inc. PGR2023-00052, 2024 WL 1343443 (PTAB March 29, 2024)

The patent is titled “Blockchain Mine at Oil or Gas Facility”. Claim 1 recites:

1. 1. 1. A system comprising:

a source of combustible gas produced from a facility selected from a group consisting of a hydrocarbon production, storage, or processing facility;

a generator connected to the source of combustible gas to receive a continuous flow of combustible gas to power the generator; and

blockchain mining devices connected to the generator;

in which:

the blockchain mining devices each have a mining processor and are connected to a network interface;

the network interface is connected to receive and transmit data through the internet to a network that stores or has access to a blockchain database;

the mining processors are connected to the network interface and adapted to mine transactions associated with the blockchain database and to communicate with the blockchain database;

the network is a peer-to-peer network;

the blockchain database is a distributed database stored on plural nodes in the peer-to-peer network; and

the blockchain database stores transactional information for a digital currency.

This case involved the § 112(f) analysis of the term “blockchain mining devices.” The petitioner argued that the term was indefinite because it could refer to either mining servers or entire data centers, leading to ambiguity in the claim scope.

The PTAB found that the term “blockchain mining devices” was sufficiently definite in the context of the claim and the specification. The Board determined that the specification provided adequate guidance on what constituted “blockchain mining devices,” including descriptions of the components, such as mining processors and network interfaces. As a result, the term did not invoke § 112(f), and the claims were not indefinite.

Takeaway: This case illustrates how ambiguity in the scope of functional terms can lead to disputes over indefiniteness. However, when the specification provides clear descriptions of the components or structure of the claimed devices, the claim can avoid being deemed indefinite under § 112(f). Ensure that your specification clearly describes the components and structure of claimed devices, particularly when using terms that could be interpreted in different ways.

Best Practices for Navigating § 112(f) in AI Patents

Using these cases as a guide, patent practitioners should consider the following best practices when drafting AI claims that might trigger § 112(f) analysis:

• Provide detailed structural support: Functional claims should be backed by sufficient structure in the specification. Avoid black-box descriptions, and instead describe the specific components or algorithms that perform the claimed function. Failing to disclose such details, as seen in Ex Parte Joon Woo Son, can result in indefiniteness.
• Avoid overly broad functional language: Using broad functional language without clear structural support is a common pitfall. In Ex Parte Jing Lan, the term “processing resource” avoided § 112(f) because it was tied to electronic circuitry. Make sure functional terms are anchored to specific structures.
• Use recognized structural terms: Use terms that are recognized in the relevant field as structural, such as “controller” or “processing unit.” Terms commonly understood by skilled artisans as referring to known structures, as demonstrated in Ex Parte Jing Lan and Crusoe Energy Systems, are less likely to invoke § 112(f).
• Be clear on claim scope: Ambiguity in functional terms can lead to disputes over claim scope. Provide clear guidance on the meaning of functional terms in the specification, as seen in Crusoe Energy Systems where the specification’s detailed description helped avoid an indefiniteness finding.
• Tailor functional claims to the specification: Ensure that the structure performing the function is clearly identified and described in detail in the specification. This can help avoid rejections under § 112(f) or § 112(b) for indefiniteness.

The Recipe for Success

When drafting or prosecuting patents under § 112(f), the key to success is to ensure that the specification adequately describes how the claimed functions are performed, especially in software and AI-related inventions. Lack of clear structure or algorithmic detail can lead to indefiniteness rejections, while thorough disclosures can help broaden the scope of functional claims and overcome these challenges.

Image source: Deposit Photos
Author: buchachon_photo
Image ID: 13043324