Their ability to reason is on par with humans’ reasoning powers. That’s a claim made about OpenAI’s new CPT-5’s reasoning and problem-solving capabilities. But in a new research paper, some experts in artificial intelligence claim that assertion is erroneous. Chengshuai Zhao, a computer science student in the School of Computing and Augmented Intelligence, part of the Fulton Schools, and Subbarao Kambhampati, a professor in the school, present their case for why CPT-5 does not quite qualify as a master of sophisticated reasoning. Kambhampati describes CPT-5 as a producer of “jagged intelligence,” referring to its ability to answer complex problems if those problems are in its training data, but if not then it fails to accurately answer extremely simple questions.