"Unmasking AI with a Single Question"

CAPTCHA, which stands for "Completely Automated Public Turing test to tell Computers and Humans Apart," does what its name suggests: distinguishing between humans and robots. However, Artificial Intelligence (AI) technology is advancing rapidly, with modern AI systems, such as ChatGPT, making headlines for their capability to convincingly mimic human-generated content. It becomes increasingly difficult to design questions that are simple for humans to answer but challenging for machines as AI becomes more sophisticated. Researchers at the University of California, Santa Barbara, are participating in this cybersecurity arms race by formulating questions designed to unmask AI. They are exploiting the differences between how humans and machines process data. In a recent study, the researchers asked humans and AIs various types of questions, scoring both groups based on the accuracy of their answers. One question type, noise injection, had a 100 percent success rate for humans and a 0 percent success rate for AIs, including ChatGPT. This article continues to discuss the study "Bot or Human? Detecting ChatGPT Imposters with A Single Question."

Harvard University reports "Unmasking AI with a Single Question"