The most recent article in The Economist highlights that at least 10% of research may already be co-authored by generative AI. This figure is considered a lower bound since, in some fields like computer science, over 20% of research abstracts are estimated to contain text generated by large language models (LLMs). The article mentions that researchers from the University of Tübingen studied vocabulary changes in 14 million PubMed abstracts, demonstrating how the emergence of LLMs led to a sudden increase in the frequency of certain stylistic words. Notably, the word "delves" saw a very significant rise (refer to the image above). You can find the original article here https://arxiv.org/pdf/2406.07016
I conducted a quick search on Scopus for the terms "Delve," "Delves," or "Delving" within abstracts and found 42,000 publications, reflecting an 800% increase in the past two years. The author with the most publications in the sample is an engineering professor from an Asian country. When I compared his h-index on Scopus with his h-index on Web of Science, I discovered a striking 300% discrepancy.
PS - It's curious that the aforementioned study by researchers from the University of Tübingen did not reference the paper by Sikander et al (2023) https://19-pacheco-torgal-19.blogspot.com/2023/12/chatgpt-4-outperform-human-researchers.html
