I am a graduate student in a STEM field. I am interested in using LLMs to automate the academic writing process, though I have found current models (GPT-4, Claude 2) to be generally poor academic writers. As LLMs improve, I will continue to explore whether they can be used to write academic manuscripts.
This market resolves YES if by 2030 I am first-author on published peer-reviewed journal article in which >40% of the text was written by an LLM. I plan to publish about 1 (or slightly fewer) journal article per year over this period.
I will update this market with any progress I make.
@cece I'm only going to counts the parts (or words) of the paper that are copied verbatim, since otherwise the percent attributable to the LLM is pretty subjective.