I’ve been using ChatGPT for several weeks to compare and process patent-related texts. I’m impressed by its ability to turn my spoken words into structured written text. In my opinion, that is a killer feature as it significantly increases my performance as a technical assessor! I have the feeling that the data-to-machine transfer has significantly improved in comparison to the keyboard. However, there are a few challenges that could make a technically driven workflow more efficient:
-
Push-to-Talk Functionality: Instead of having an ongoing conversation with ChatGPT, I need a feature that allows me to dictate my thoughts in one go and have them directly transcribed into text for further editing. ChatGPT has the problem that it does not recognize whether I have stopped talking or am just pausing to think about my next sentence. It simply interprets the silence as a signal to respond or summarize. As a result, I have to talk very quickly and fluently to prevent it from responding too early.
-
Custom GPT Instance with Reference Documents: The option to create custom GPTs and upload documents as references is very promising. However, it currently lacks comprehensive support for larger document libraries and precise queries, such as “Check document X for specific content,” because the library is not systematically labeled by ChatGPT’s GPT creator.
-
Interface for Scientific and Technical Work: While ChatGPT shows great potential, it lacks a specialized “add-on” or extension for scientific and engineering-focused work. A feature akin to an IDE for ChatGPT conversations—one that includes filters, exclusions, and other parameters in a comprehensive user interface—would be ideal.
My question to the community: Are there any projects or extensions that already provide these functionalities or are in development? I look forward to hearing your thoughts and potential solutions.