Science journalists find ChatGPT is bad at summarizing scientific papers

The article reports that science journalists have found that the language model ChatGPT struggles to accurately summarize scientific papers. According to the findings, ChatGPT "tended to sacrifice accuracy for simplicity" when tasked with writing news briefs based on research articles. The article suggests that while ChatGPT can generate text that appears coherent, it often fails to capture the nuances and details present in the original scientific papers. This raises concerns about the potential use of such language models in science journalism, where accurate representation of research findings is crucial. The article highlights the need for continued human involvement and oversight in the process of translating complex scientific information for a general audience. It suggests that while language models like ChatGPT can be useful tools, they may not be reliable enough to fully replace the expertise and critical thinking of human science communicators.