In this new age of generative AI and large language models (LLMs), we are quickly learning what tasks benefit from this new tool and what tasks are harmed by it. The legal community, and seemingly the world at large, watched with grim curiosity as New York attorneys were reprimanded for using ChatGPT to draft a brief that he submitted to the court without proofreading, or seemingly doing any kind of editing. Much to that attorney’s chagrin, ChatGPT had invented case law here, all the way from the caption to the point of law the cases stood for in the brief. This article highlights how LLMs can be invaluable yet fall short when it comes to patent research.
A Client’s Encounter:
A tech-savvy client of mine sought to leverage an LLM to enhance his inventive concepts, leading to positive outcomes in terms of brainstorming and generating patent claims. However, when assessing patentability and searching for prior art references relevant to his inventions, problems arose. The LLM returned a list of references that seemed to directly overlap with his inventions based solely on their titles, causing concern about patent protection. To verify their relevance, the patent professional checked the first reference, only to discover it had no connection to the client’s technology. Subsequent searches yielded the same fabricated results, revealing that the LLM had generated fictitious prior art references. This situation could have led the client to abandon potentially valuable intellectual property assets if he had relied solely on the LLM’s results.
The Power and Limitations of LLMs:
Large language models have undoubtedly revolutionized content generation with their impressive ability to produce coherent and sophisticated text. However, when it comes to factual inquiries, they still fall short of delivering reliable information. While ongoing efforts aim to enhance the factual accuracy of LLMs, it remains crucial for patent professionals and tech entrepreneurs to exercise caution. Relying solely on LLMs for patent research or fact-based decisions can be risky. Traditional search tools and expert human analysis continue to be indispensable for verifying the accuracy of information.
The emergence of generative AI and large language models has created exciting possibilities for patent professionals and tech entrepreneurs. However, as demonstrated by the experiences of the New York attorney and the tech-savvy client, the accuracy of information generated by LLMs remains a concern. While LLMs excel in content generation, their limitations in providing information underscore the importance of relying on traditional research tools and expert guidance for patent-related tasks. As the field of AI continues to evolve, discerning the use of these tools and staying vigilant is essential for navigating the intellectual property landscape effectively.
If you are inquiring more about AI’s involvement in the legal system, please contact Dan Pierron at dpierron@uslegalteam.com for additional assistance.
Dan Pierron’s passion for technology is reflected in the work he does for clients daily. He works diligently to ensure the protection of the intellectual assets of inventors and companies through acquiring patents, trademarks and making sure clients understand their intellectual property rights. Among copyrights, trademarks and intellectual property, Dan practices law in the areas of data security, business litigation and patents.
