A recent study has shown that large language models (LLMs) like ChatGPT, Gemini, Claude and others, are having a big impact on how scientific papers are written, especially in the field of biomedical research. Researchers from the University of Tübingen in Germany found that about 13% of biomedical abstracts published in 2024 were influenced by LLMs.
The study analyzed millions of abstracts from PubMed and found a noticeable rise in the use of certain words that are typically used by LLMs. These “style or stylistic” words, like “delve,” “meticulously,” and “intricate,” are examples of the more elaborate language preferred by AI models.
Researchers estimate that LLMs help write at least 150,000 biomedical papers every year. This number could be higher, as the study only counted papers using words strongly associated with LLMs.
The study also showed that LLM usage varies across different fields, countries, and journals. In fields like computer science, where researchers are more familiar with AI, LLMs are used in about 20% of papers. Non-English-speaking countries often use LLMs to help with English writing.
However, while LLMs can improve grammar and readability, the researchers warned about potential problems. AI can sometimes make mistakes, like creating fake references or spreading incorrect information. This could lead to less original and more uniform research.
The study concluded that LLMs are rapidly changing scientific writing in ways that haven’t been seen before, even surpassing the changes caused by the COVID-19 pandemic. The researchers suggest that as LLMs become more widely used, academic institutions may need to update their rules and policies around AI in research.
🖇To read full article, access preprints via https://arxiv.org/html/2406.07016v4
