Solutions to homework problems found in textbooks are widely available on line. This limits the utility of such problems, or any other problems that are available on the internet. On the other hand, creating new homework problems aligned with student learning objectives (SLOs) is a tedious task. This is particularly true for more specialized, advanced courses like Modern Physics, which is generally taken by STEM majors in their Junior year, and which covers in an introductory fashion topics like Special Relativity, Quantum Mechanics, Particle Physics, and Astrophysics and Cosmology. In this project, we used Generative AI this issue in a Modern Physics course. After defining the SLOs, AI was used to generate homework and practice exam questions aligned with the SLOs and validated for accuracy and clarity by the Professor. This proved to be a quick and efficient method for generating a unique set of problems. In this study we report on the general accuracy of AI in generating usable problems and the ability of AI or Google to given correct answers to the problems. This latter set of statistics provides a measure of what students would find if they tried to obtain the answers from an internet search as opposed to actually solving the problems.

Keywords: Technology, Science education.