Do the math: ChatGPT sometimes can’t, expert says
Paulo Shakarian directs Lab V-2, where challenges in the field of artificial intelligence are examined. In one recent project, Shakarian tested the new generative AI technology ChatGPT on 1,000 mathematical word problems. It did not consistently produce impressive results. In an interview, the associate professor in the School of Computing and Augmented Intelligence, part of the Fulton Schools, discusses what the results of the testing indicate about the abilities, usefulness, limitations and drawbacks of ChatGPT. Among Shakarian’s conclusions is that ChatGPT could be valuable in many practical applications in which precise accuracy is not a tantamount concern. But there are concerns about when ChatGPT might be involved in guiding decisions that would have ethical implications.