Experts from the University of Reading (UK) managed to deceive their colleagues using exam papers written by artificial intelligence (AI) , writes "Guardian".
As part of the experiment, enthusiasts created fake identities of students and submitted exam papers on their behalf, solved by the chatbot ChatGPT. The material says that the professors at the university did not notice the fraud and gave a high rating to the work.
According to the researchers, their colleagues noticed errors in only one paper out of 33. On average, the fake students' grades were higher than those of the real university students. The experts concluded that the AI was able to successfully pass the exams and remain undetected.
The material says universities around the world need to more closely scrutinize exams and research papers produced by students. It is likely that a number of educational institutions will have to abandon the practice of taking tests and exams online. The researchers note that AI has become even better at imitating humans. In particular, OpenAI's chatbot has learned to use abstract thinking.