Note: First published in The Intellectual Property Strategist and

This article is Part One of a Three-Part Article Series

Artificial intelligence is changing industry and society, and metrics at the US Patent and Trademark Office (USPTO) reflect its impact. In a recent publication, the USPTO indicated that from 2002 to 2018 the share of all patent applications relating to artificial intelligence grew from 9% to approximately 16%. See “Inventing AI, Tracing the diffusion of artificial intelligence with U.S. patents,” Office of the Chief Economist, IP Data Highlights (October 2020). For the foreseeable future, patent applications involving artificial intelligence technologies, including machine learning, will increase with the continued proliferation of such technologies. However, subject matter eligibility can be a significant challenge in securing patents on artificial intelligence and machine learning.

This three-part article series explores USPTO handling of Alice issues involving artificial intelligence and machine learning through a sampling of recent Patent Trial and Appeal Board (PTAB) decisions. See Alice Corp. v. CLS Bank Int’l, 134 S. Ct. 2347 (2014). Some decisions dutifully applied USPTO guidelines on subject matter eligibility, including Example 39 thereof, to resolve appeal issues brought to the PTAB. In one case, the PTAB sua sponte offered eligibility guidance even with no Alice appeal issue before it. These decisions inform strategies to optimize patent drafting and prosecution for artificial intelligence and machine learning related inventions.

Part One

“Generic Machine Learning Algorithm”

In Ex parte Hussain, Appeal No. 2020-005406 (PTAB Feb. 18, 2021), the PTAB considered the subject matter eligibility of claims reciting a “machine learning algorithm” in relation to mitigation of risk of consumer default on an online transaction. Representative claim 1 recited as follows:

1.         A computer-implemented method, comprising:
under the control of one or more computer systems that
execute instructions,
          providing executable instructions to a client
computing device associated with a user that, as a result of
being executed by the client computing device, causes the client
computing device to:
                        collect client data that includes:
                                    personally identifiable information
                        about the user;
                                    an identifier associated with the client
                        computing device; and
                                    a measurement captured by the client
                        computing device associated the user interacting
                        with the client computing device, the measurement
                                               an action performed by the user
                                    to an object displayed in a user interface of
                                    the client computing device;
                                               an identity of the object; and
                                               a time value indicating a time at
                                    which the action was performed to the
                                    object; and
                        provide the client data to the one or more
                        computer systems;
               obtaining stored transactional data associated with
one or more previous transactions involving the user;
               obtaining verification data verifying that the
personally identifiable information is accurate;
               transforming the stored transactional data, the verification data, and the client
data that includes the measurement into a set of variable values usable as input into a
machine learning algorithm that is trained to infer characteristics about the user from
the set of variable values;
               obtaining, as a result of inputting the set of variable values into the machine
learning algorithm, a fidelity score output by the machine learning algorithm; and
               based at least in part on the fidelity score and
               without obtaining additional information about the user from a
               third party:
                              determining a payment or credit option to
               display in the user interface for a current transaction; and
                              updating, contents of the user interface to
               provide the payment or credit option.


Id. at 2-3 (emphasis added). To assess subject matter eligibility of the representative claim, the PTAB applied USPTO guidelines mandating the familiar two step analytical framework. See USPTO, 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (Jan. 7, 2019); USPTO, October 2019 Update: Subject Matter Eligibility, 84 Fed. Reg. 55942 (Oct. 17, 2019).

As to the first prong of Step 2A in the analytical framework, the PTAB indicated that the representative claim used only a “generic machine learning algorithm” to output a fidelity score “in some unspecified manner.” The PTAB also addressed Example 39 of the USPTO guidelines, the example reciting machine learning in a hypothetical claim deemed eligible. In particular, the PTAB contrasted relevant detail in the claim of Example 39 versus the relative absence of such detail in the representative claim. The PTAB acknowledged that the representative claim expressly recited that the machine learning algorithm was trained to infer characteristics about a user from variable values generated from specific types of data. Nonetheless, the PTAB reiterated that the machine learning algorithm as claimed was trained to make inferences in an “unspecified way without any technical details.” The PTAB gave little consideration to the recited transformation of the specific types of data into the variable values, which were specifically claimed as inputs to the machine learning algorithm. Depending on the facts, the claimed inputs to the machine learning algorithm could have been deemed suggestive of data to train the machine learning algorithm. For that reason, the claimed inputs might have been argued to potentially resemble or parallel the recitation of training data details supporting eligibility in Example 39. However, no such arguments were raised.

The PTAB found that a machine learning algorithm “as such” was not described in the specification – despite the acknowledged references in the specification to a logistic regression, random forest, supervised learning algorithm, neural network, vector machine, and other classification algorithm. According to the PTAB, the description of these “other concepts” without technical details confirmed the abstract nature of the claimed machine learning algorithm. In particular, the PTAB noted that the specification described algorithms to generate fidelity scores “without details of training them” to infer characteristics about users. Refusing to also consider the machine learning claim limitations under the second prong because they recited the abstract idea under the first prong, the PTAB ultimately determined that the representative claim was ineligible after finding no inventive concept.

Accordingly, not just any claimed specifics about an artificial intelligence related invention will satisfy the PTAB about eligibility. Although the representative claim in Hussain recited the inference specifically generated by the machine learning algorithm, the PTAB indicated that the claim still did not specify enough. In view of the PTAB’s observation that both the specification and representative claim lacked technical detail, expressly claiming training data and identifying it as such – and of course beforehand drafting the patent application in support thereof – might have secured a different outcome.

Part Two Preview

Part Two of this article series will further analyze recent PTAB decision making regarding artificial intelligence and subject matter eligibility.