Teaching physics in the age of AI: Case studies in ChatGPT and Sora

In late February of this year, OpenAI released some sample videos created by Sora, a tool aimed at stimulating the physical world. If Sora and its descendants can accurately simulate the physical world, then what can we teach students as physics educators? It seems that we are working during a transition in which new AI tools are being developed to do physics and to learn physics. Initially, I felt overwhelmed by AI technology due to its immense power. The good news is that ChatGPT still has many bugs in solving physics homework problems. Additionally, Sora has bugs that allow it to violate physical laws. For now, ChatGPT and Sora are not reliable physicists. I will describe some examples of bugs that ChatGPT and Sora exhibited when doing introductory physics problems. It is worth noting that when I showed a Sora-generated video to students, they did not identify the "inaccurate physics" portion as frequently as experienced physics teachers did. Consequently, it is still important to teach students critical thinking, fine observation skills, and healthy skepticism so that they can be alert to such bugs and inaccuracies.