Bernstein members involved: Florian Röhrbein

For a while now, ChatGPT (Chatbot Generative Pre-trained Transformer) has been the talk of the town and is used millions of times. The speech- and text-based chatbot AI from the US company OpenAI is particularly suitable for dialog applications, as an idea generator or helper in the pre-formulation of texts. How perfect is this chatbot really? Can experts within their field distinguish a paper written by ChatGPT from papers written by humans? Sven Voss from Mitteldeutscher Rundfunk wants to know exactly. In his magazine “Voss & Team” Prof. Dr. Florian Röhrbein, holder of the professorship Neurorobotics at the Chemnitz University of Technology, is tested whether he recognizes the work of the so-called ” Fairbrechers Cornel Bunz”, which was written with ChatGPT, among anonymized term papers of his students on the topic “Bio-inspired Robotics”.
“The test already took place at the beginning of February during the shooting at our Faculty of Computer Science and in the lecture hall. How it turned out, I will of course not reveal before the broadcast of the show,” says Röhrbein, “only this much: the performance of ChatGPT is already impressive.” So it’s worth tuning in to the MDR program “Voss & Team” on March 9, 2023, at 8:15 p.m.

Translated from German by Alexander Lammers/Bernstein Network Computational Neuroscience.