- AI-focused patents and subject matter eligibility under 35 U.S.C. § 101
- Texas and California criminalize election “deepfakes”
- USPTO to enlist AI in patent prosecution
- U.S. Coast Guard seeks public comment on autonomous commercial vessels
- France and China expand sensitive technology lists to AI
- Gartner analyzes legal industry trends in AI-assisted technology
- Update on Planner5D litigation
- QE in the News – IP Litigation MVP Kevin Johnson
- QE Virtual Panel Discussion on Oct. 28: “Can AI Invent?”
SPOTLIGHT – USPTO HIGHLIGHTS GROWTH IN AI PATENTS
Which AI-focused Patents Are Subject Matter Eligible Under 35 U.S.C. § 101?
Background. It has been nearly two years since the USPTO began reviewing AI-focused patent claims under the 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (Jan. 7, 2019) (“Guidance”). The Guidance assists examiners in determining if an invention is patent eligible under 35 U.S.C. § 101. It directs examiners, in the first step of the Supreme Court’s Alice/Mayo test, to assess whether the invention falls into one of three subject matter groupings of abstract ideas—mathematical concepts, methods of organizing human activity, and mental processes. The Guidance further provides, even if a claim falls into an abstract grouping, it is not directed to an abstract idea if it integrates the idea into a practical application. And, even if a claim fails both of these hurdles, in step two of Alice/Mayo an examiner must further evaluate if the claim nonetheless recites an unconventional “inventive concept” conferring patent eligibility.
PTAB Decisions. The Board has upheld the patent eligibility of several patents covering AI techniques, including related to machine learning and neural network technology. At a high level, the Board has found that, even when a claim falls into an abstract category grouping, AI provides a patentable practical application when it offers a technological improvement in the related technical field. Such improvements include, for example, enhancing the processing capabilities of a computer and giving the computer new functionalities. Conversely, the use of AI by itself is not enough to provide a practical application or inventive concept. Rejections have been upheld when generic or standard machine learning methods were simply used to automate a process without otherwise advancing the technical field.
Eligibility Upheld. In Ex Parte Adjaoute, Appeal No. 2018-007443 (P.T.A.B. Oct. 10, 2019), the patent at issue claimed a method for protecting groups of digital electronic appliances and, in particular, issuing predictions and warnings using AI classification techniques to diagnose impending failures. The application was initially rejected on the basis that the claimed limitations for reading data, assessing data, presenting data, classifying data, collecting data, and tallying data were directed toward the abstract idea of monitoring the operation of machines—a method of organizing human activity.
The Board reversed the examiner and found the claims were not directed to a fundamental economic practice of organizing human activity. First, it reasoned that the limitations of the claims, such as monitoring the operation of neural networks, logic decision trees, and smart agent profiling, cannot practically be performed in the human mind or with pen and paper. The Board further noted, even if the claims did fall into an abstract grouping, the claims integrate the AI invention into a practical application. The claims are rooted in computer technology and employ AI classification technologies to overcome a computer networking problem.
- Ex Parte Hannun, Appeal No. 2018-003323 (P.T.A.B. Apr. 1, 2019)
- Ex Parte Hueter, Appeal No. 2018-007627 (P.T.A.B. Oct. 9, 2019)
- Ex Parte Bushmitch, Appeal No. 2018-008667 (P.T.A.B. Mar. 12, 2020)
Eligibility Rejected. In Ex Parte Albert, Appeal No. 2019-000295 (P.T.A.B. Feb. 24, 2020), the patent claimed a method for using machine learning to determine whether a patient requires post-discharge care. The examiner rejected the application on the basis that the claims were directed to unpatentable abstract ideas of organizing human activity and mental processes.
The Board affirmed the examiner’s decision and found that the generic use of machine learning in the claimed risk analysis algorithm was insufficient to integrate the AI invention into a practical application. For example, the mere recitation of “using machine learning” and “computer-implemented” in the claims did not provide an improvement to a technical field, but was rather a “generic recitation of [technology] for automating the performance of the abstract idea.” The Board further noted that, under the Guidance, analyzing whether a claim recites an inventive concept in step two of Alice/Mayo is essentially the same as the Board’s step one analysis regarding integration into a practical application. Accordingly, the Board emphasized that because there was no special functionality in the machine learning algorithm improving data retrieval or network usage, the use of machine learning did not elevate the claims to an inventive concept.
See also, Ex Parte Contreras, Appeal No. 2019-002601 (P.T.A.B. Dec. 20, 2019).
OTHER RECENT DEVELOPMENTS
GOVERNMENT & LEGISLATION
- A “deepfake” is a popular image synthesis technique often rooted in AI. Texas and California have enacted laws to criminalize the use of “deepfakes” in connection with elections. Several similar bills are pending in Congress (e.g., H.R.6088). Concerns remain regarding the practical enforceability of this legislation—for example, how to enforce against state actors located overseas.
- A recent USPTO presentation disclosed that a prototype AI is under development to better classify patent applications, select the best examiner for a particular application, and conduct prior art searches.
- The U.S. Coast Guard requested public comments on the use of autonomous commercial vessels in maritime transportation systems. Questions posted include:
- What are the benefits and cost-savings of automated and autonomous commercial vessels and vessel technologies, if any?
- What impact to the maritime workforce do you anticipate would occur with the introduction of autonomous commercial vessels?
- What threats do autonomous commercial vessels present to cybersecurity or privacy?
- Similar to CFIUS review of foreign investments and technology exports, France recently expanded the list of “sensitive” activities subject to review and prior authorization under the French Monetary and Financial Code (FMFC) Article R.151-3, which now includes Artificial Intelligence and Cybersecurity. See FMFC Article 6. Likewise, in August 2020, China added 23 new technologies to its export control list, including “certain artificial intelligence interface technologies.” See Adjustment to China’s Catalog of Technologies the Export of Which is Prohibited or Restricted (Mandarin).
AI AND LEGAL PRACTICE
- Gartner analyzed current trends in legal technology and categorized products into five stages of development: innovation trigger, peak of inflated expectations, trough of disillusionment, slope of enlightenment, and plateau of productivity.
- This Law360 article summarizes several legal tech products in each stage of development. For example, the use of legal chatbots is on the rise. Legal chatbots use neural networks to understand and respond to human inquiries. In some instances, legal chatbots make legal services more accessible in routine, standardized situations such as challenging parking tickets. However, due to ethics rules, service providers must be mindful of pitfalls such as unauthorized practice of law.
- Another use case is leveraging AI translation technologies along with existing certified patent linguists to help IP departments streamline the translation process and ensure quality.
UPDATE ON THE PLANNER5D LITIGATION
- In the April 2020 edition of the AI Bulletin, we reported on Planner5D’s copyright infringement and trade secret misappropriation action against Facebook and Princeton University, relating to 2D and 3D models of virtual objects. Shortly before that update, Judge Orrick in the Northern District of California granted the defendants’ motion to dismiss both claims, with leave to amend.
- On July 24, Judge Orrick issued an order rejecting the defendants’ motion to dismiss the amended trade secret misappropriation claims. In the same order, Judge Orrick granted the motion to dismiss the amended copyright claims, noting that the Copyright Office issued registration certificates indicating the year of completion of the objects was 2019, while the year in which Princeton allegedly scraped the objects from Planner5D’s website was 2016. However, Judge Orrick again granted leave for Planner5D to file registration applications for all its work up to 2016 so it could include them in this lawsuit. Planner5D will have until October 23 to amend its copyright claims.
QUINN EMANUEL IN THE NEWS
- Kevin Johnson, a partner in QE’s Silicon Valley office, was named one of Law360’s 2020 Intellectual Property MVPs. Kevin was co-lead counsel in a two-week trial representing the California Institute of Technology against Apple and Broadcom, and obtained over $1.1 billion in damages for infringement of three of the university’s patents directed to error correction code used in Wi-Fi chips. On the same day as the Caltech verdict, Kevin also obtained a complete defense verdict as lead trial counsel for C3.ai—an AI enterprise software company—and its CEO Thomas Siebel.
- Jordan Jaffe, a partner in QE’s San Francisco office, will moderate a virtual panel discussion on Wednesday, October 28 on the question of “Can AI Invent?” Panelists will include Wayne Stacy (Silicon Valley regional director of the USPTO), Dennis Crouch (professor at the University of Missouri School of Law and proprietor of the Patently-O blog), and Ryan Abbott (author of DABUS, the Artificial Inventor project).