Patent applications directed to Artificial Intelligence (AI) have fallen victim to extra scrutiny under 35 U.S.C. §101 in a post-Alice world, especially since most AI inventions involve software and mathematical-based solutions. However, in the past couple of years there has been notable signs that the USPTO wishes to protect specific areas of AI from being subject to rejections under 35 U.S.C. §101. In January 2019, the USPTO issued revised guidance directed to what constitutes patent eligible subject matter under 35 U.S.C. §101. The guidance included hypothetical examples to help examiners understand the boundaries of eligible and non-eligible subject matter. One particular example in this guidance was directed to a computer-implemented method of collecting digital images for training a neural network that is used for facial detection. The USPTO concluded that this claim was patent eligible, and it just so happens that this broad example claim covered one of the most commercially significant forms of AI in the market today – facial recognition based on deep learning.
Now Ex Parte Linden, a PTAB decision from April 2019, has been explicitly identified as “informative” by the USPTO. As a quick primer, an “informative” decision provides Board norms on recurring issues, guidance on issues of first impression to the Board, guidance on Board rules and practices, and guidance on issues that may develop through analysis of recurring issues in many cases. This is explicitly separate from “precedential” decisions designated by the USPTO which are binding.
In Linden, the claims at issue are directed to a method of using a trained neural network to transcribing inputted speech, i.e., speech recognition AI (think Alexa, Siri, or Google Assistant). Cutting to the chase, the PTAB reversed the Examiner’s rejection under 35 U.S.C. §101 and found that the claims were patent eligible.
Regarding the currently accepted “test” for determining subject matter eligibility at the USPTO, the revised guidance of 2019 revised Step 2A of the Alice-Mayo framework into a two-prong approach. First, a determination has to be made whether the subject matter is directed to an “abstract idea,” by deciding whether the subject matter falls into one of three categories – (1) mathematical concepts, (2) mental processes, and (3) certain methods of organizing human activity. If the subject matter does not fall into one of these categories, then the claim is patent eligible. Otherwise, the claim is directed to an “abstract idea” and prong two must be evaluated – whether this “abstract idea” is integrated into a “practical application.”
In Linden, the PTAB explained why the claims at issue did not fall into any of the three enumerated categories. Of note, when batting down the possibility that the claims are directed to a “mental process,” the PTAB states: “While transcription generally can be performed by a human, the claims here are directed to a specific implementation including the steps of normalizing an input file, generating a jitter set of audio files, generating a set of spectrogram frames, obtaining predicted character probabilities from a trained neural network and decoding a transcription of the input audio using the predicted character probability outputs. These are not steps that can practically be performed mentally.” [Emphasis added]. This is significant because at a high level many AI innovations are attempting to automate a mental process or human activity, but this decision clarifies that the specific technical steps of the claim are exactly what differentiates the invention from an actual mental process or human activity.
The PTAB further explained why even if it falls into one of the categories, any abstract idea in the claims (such as a mathematical concept) is integrated into a practical application by improving the technical field of speech recognition. What is significant here is the acknowledgment that automatic speech recognition itself is a technical field subject to the same treatment of the claims in Enfish and McRO.
The analysis by the PTAB in Linden is worth studying because it addresses multiple prongs of the currently accepted test for determining patent subject matter at the USPTO while characterizing the claims at a very high level. In other words, the 101 question on this case does not turn on a very nuanced feature that is unique to the claims at issue, and it is not difficult to analogize the PTAB’s findings to other fields in the area of AI. While Linden is non-binding, it may be a useful addition to the patent practitioner’s toolkit when prosecuting a challenging AI case before the USPTO.