Comments by "Mikko Rantalainen" (@MikkoRantalainen) on "How Does ChatGPT Do on a College Level Astrophysics Exam?" video.

  1. ChatGPT actually did better than I expected. I have done zero reading for this kind of exam and I would have got maybe 55–60% total based on the questions shown here. I'm interested in natural sciences in general as a hobby and I would have expected ChatGPT to be closer to my level based on the answers I've got from it. And I really really appreciate the open book exam! I think that's the only way to really test understanding instead of memory items. In long run, you have to design exams and homework to test understanding instead of something that can be answered by ChatGPT. And yes, creating such exams and homework assignments is hard but that's the only way forward. And maybe simply accept the fact that AGI is going to be invented near year 2030 and the work those students are doing in the future is going to be in collaboration with AI.
    2