Exobrains for research evaluation and paper writing
Résumé
The fact that scientists could be fooled by abstracts written by an artificial intelligence chatbot (Else 2023) implies that such tools will soon represent 'exobrains' to carry out tasks such as scientific writing and research evaluation (Fig. 1). Here we have tested ChatGPT for evaluating a full-length review article, and we were impressed by the results that humans often cannot deliver, such as summarizing a 7,000-word research paper into sensible, bullet-point listed findings in less than 30 seconds. Although answers from such large language models are not flawless, ChatGPT appears as an exobrain—similar to the physical exoskeletons that humans could use—to amplify their intelligence. Could we use ChatGPT to evaluate papers and measure research performance? Here the ChatGPT evaluation of one of our review articles reveals the intriguing performance of the chatbot.