Scientists use ChatGPT to write entire paper, reflect on results: Editorial

An editorial in Nature documents the findings of two researchers who tested the extent of ChatGPT’s paper-writing abilities. Scientists Roy Kishony and Tal Ifargan (Technion – Israel Institute of Technology) produced a complete paper in less than an hour by feeding ChatGPT suggestions and prompts. While the paper was clearly written and demonstrated good analysis, the researchers reflected that its findings were unoriginal and that ChatGPT had a problematic tendency to fill in research gaps by “making things up.” Other academics explained that while ChatGPT can accelerate research processes, it can also reproduce biases and “hallucinate” data. The article concludes that for these reasons, ChatGPT’s capacity to effectively produce entire papers may not yet be a threat to the academic space.