Can generative AI and ChatGPT outperform humans on cognitive-demanding problem-solving tasks in science?