AI-PreSearch: EPO’s AI-Driven Patent Search Tool
In August 2023, the European Patent Office (EPO) introduced a cutting-edge tool designed to enhance the patent examination process: AI-PreSearch. This AI-powered search tool was announced to leverage artificial intelligence to aid patent examiners in identifying relevant prior art documents. The EPO has vectorized and embedded all prior art documents within their database, creating an extensive vector space consisting of hundreds of millions of vectors. This endeavor led to the development of a patent-trained language model known as EP-RoBERTa, previously utilized for automatic Cooperative Patent Classification (CPC). The primary goal of AI-PreSearch is said to streamline the patent examination process, thereby increasing efficiency by automating routine tasks and offering advanced search capabilities.
Expectations Versus Reality
Despite the high expectations that accompanied the introduction of AI PreSearch, it is important to note that the tool may not fully live up to the initial hype. A disparity between expectations and reality is common with many AI-driven tools. In the patent context, the critical question revolves around the extent to which AI-supported tools can genuinely simplify the work of patent examiners or potentially render human activities obsolete.
The market is already saturated with tools claiming to facilitate patent searches and the formulation of patent applications. However, a closer examination of these online tools reveals that they often provide only rudimentary support, primarily easing routine tasks rather than offering comprehensive solutions.
The Limitations of AI in Patent Examination
A closer look at the official information available reveals that AI-PreSearch, like other AI tools, has significant limitations in comprehending the technical nuances of patent applications. Patent examination is a complex process involving the interpretation of intricate language, application of case law, and evaluation of unique technical features. These tasks require a profound understanding of the subject matter that AI tools currently struggle to emulate.
In particular, interpreting patent formulations and applying legal precedents are areas where human expertise remains indispensable. This necessity is even more pronounced in borderline cases, such as software inventions (computer-implemented inventions), where distinguishing between patentable and non-patentable subject matter can be exceedingly subtle. Presently, AI tools lack the sophistication required to accurately assess these complex scenarios. Ironically, AI tools often encounter the most difficulty when dealing with AI-related inventions, underscoring their current limitations in navigating intricate legal and technical landscapes.
The Future of AI-PreSearch
There is hope that AI-PreSearch will transcend being a mere stopgap measure addressing the EPO’s challenges, such as the declining number of patent examiners and the increasing volume of patent applications. The aspiration is for AI-PreSearch to genuinely enhance the capabilities of patent examiners, allowing them to concentrate on more critical aspects of their work while the AI manages routine and repetitive tasks. Given that 75% of all opposed patents in recent years have been either fully revoked or limited, there is a clear need for an improvement in the quality of the EPO's patent search and examination practices. Hopefully, AI-PreSearch can contribute to this.
Talk to us and contact us for further information at:
Phone: +49 (0)611 / 341568-0
Fax: +49 (0)611 / 341568-11
E-mail: Send email to us
PGP-Key
Author: Dr. Michael Schmid