ChatGPT Writes Fake Scientific Abstracts That Fool Reviewers

January 17, 2023 | Automation
ChatGPT

A research team at Northwestern carried out a study to see whether OpenAI’s ChatGPT could fool reviewers of scientific papers with fabricated abstracts.

The reviewers were shown 25 abstracts, mixing real and fake ones together, and they had to give a score based on what they had read. They were fooled by the AI-generated ones 32% of the time.

https://news.northwestern.edu/stories/2023/01/chatgpt-writes-convincing-fake-scientific-abstracts-that-fool-reviewers-in-study/
https://www.biorxiv.org/content/10.1101/2022.12.23.521610v1
https://openai.com/blog/chatgpt/