Does ChatGPT pose a risk to the validity of scientific research? - Digital Journal

ChatGPT (Generative Pre-trained Transformer), the conversational artificial intelligence development, can write convincing fake scientific abstracts capable of fooling reviewers. This is according to a new study. In the trials, human reviewers could only detect fake abstracts, generated by the OpenAI developed software, 68 percent of the time. For the research, blinded human reviewers were given a mix of real and falsely generated abstracts and asked to sort which was which. Each reviewer was given 25 abstracts that were a mixture of the generated and original abstracts and asked to give a binary score of what they thought the abstract was.