The U.S. Patent Office Won’t Let Its Employees Use AI Tools Like ChatGPT

Last year, the US Patent and Trademark Office banned the usage of AI tools similar ChatGPT successful the office. An interior memo obtained by WIRED done a nationalist records petition elaborate the ins and outs of the ban. But that doesn’t mean the Patent Office isn’t utilizing AI astatine all—quite the opposite.

According to WIRED, USPTO main exertion serviceman Jamie Holcombe sent retired a memo to Patent Office employees successful April 2023. It prohibited workers from utilizing generative AI tools for enactment owed to the system’s “bias, unpredictability, and malicious behavior.” Holcombe said that the Patent Office is “committed to pursuing innovation with our agency” and “working to bring these capabilities to the bureau successful a liable way.”

That aforesaid month, Scott Beliveau—the USPTO’s Branch Chief of Advanced Analytics—parroted the connection of Holcombe’s memo successful an interview with Forbes. “We are arsenic committed to pursuing innovation wrong our agency,” Beliveau told Forbes erstwhile asked astir however the Patent Office was moving to minimize the risks of AI.

Beliveau told Forbes that his bureau wasn’t allowed to usage LLMs astatine all. “We started mitigating the risks of LLMs by prohibiting our employees and contractors from utilizing generative AI tools,” helium said. “This contiguous enactment was taken portion we proceed to research ways to bring LLM capabilities to the bureau successful a liable mode that serves America’s innovators.”

This does not mean that the Patent Office is ignoring AI. The Patent Office examines millions of ineligible documents each twelvemonth and it has admitted it’s utilizing AI tools to assistance with the work. “Patent examiners are performing AI-enabled anterior creation searches utilizing features similar More Like This Document (MLTD) and Similarity Search, successful the Office’s Patents End-to-End (PE2E) Search tool,” it said successful a papers detailing its exhaustive usage of AI published successful the Federal Register. “Patent practitioners are progressively relying connected AI-based tools to probe anterior art, automate the patent exertion reappraisal process, and to summation insights into examiner behavior.”

In summation to its ain usage of AI, the Patent Office knows that the genie is retired of the vessel regarding generative AI submissions to its office. “The capabilities of these tools proceed to grow, and determination is nary prohibition against utilizing these machine tools successful drafting documents for submission to the USPTO,” it said successful the Federal Register document. “Nor is determination a wide work to disclose to the USPTO the usage of specified tools.”

The Patent Office’s property caput besides told WIRED that USPTO employees tin usage generative AI tools wrong an interior investigating environment. “Innovators from crossed the USPTO are present utilizing the AI Lab to amended recognize generative AI’s capabilities and limitations and to prototype AI-powered solutions to captious concern needs,” the USPTO told WIRED.

Government policies regarding the usage of generative AI are each implicit the place. The USPTO and the Department of Defense are some forward-thinking, compared to galore of their peers. In September, aggregate agencies published reports connected however they’re utilizing AI, and the results weren’t great. Adoption is dilatory crossed the board. The agencies don’t person capable money oregon endowment to make their ain tools. The Department of Energy, the bureau successful complaint of America’s atomic weapons, whined that it didn’t person capable graphics cards.

Meanwhile, the Patent Office has been plugging distant astatine this occupation for years. It rolled retired its ain interior hunt instrumentality respective years ago. “While the USPTO is committed to maximizing AI’s benefits and seeing them distributed broadly crossed society, the USPTO recognizes the need, done method mitigations and quality governance, to compartment the risks arising from the usage of AI successful signifier earlier the USPTO,” it said successful its guidance memo.