AI Chatbot ChatGPT Struggles to Generate Convincing Scientific Papers

ChatGPT, an AI chatbot that has become popular since November 2022, may be good at mimicking human workers in multiple fields but not in scientific research, says a new report. Researchers have created a computer program to spot fake studies generated by the chatbot and have discovered that the AI is still capable of fooling some humans with its science writing. However, a computer model has been created to detect ChatGPT-generated fake studies that are more than 99% of the time. ChatGPT-created papers differed from human text in various ways, including paragraph complexity, sentence-level diversity in length, punctuation marks, and popular words. As a result, computer programs to differentiate between real and AI-generated papers continue to be necessary since fewer humans can differentiate the differences. The researchers have warned that this is only a proof of concept and that much more wide-scale studies are needed to create robust models that are even more reliable. Such models can be trained to specific scientific disciplines to maintain the integrity of the scientific method.