Researchers who use AI publish more and are cited more frequently, but the diversity of topics explored and the number of collaborations are declining overall.
A team analyzed more than 41 million natural science articles using a trained language model to identify AI-enhanced work, whether machine learning or generic AI. Approximately 311,000 items were identified as such, with an F1 accuracy of 0.875 against expert annotations.
Scientists using AI publish three times as many articles and receive almost five times more citations than their peers who do not. But this same use is accompanied by a 5% decline in the diversity of the topics covered and a 22% decrease in collaborations, suggesting that AI enhances individual productivity while at the collective level narrowing the scope of the topics actually explored by research.
Read the full article on the original source
Go to the source →