Very chill semester.
Interesting class. Some topics covered include the rasterization pipeline, halfedge meshes, pathtracing, spatial data structures, and animation. Assignments are pretty fun (A3 was my favorite) and have very few moving parts, shielding you from the pains of working with C++. Course infrastructure could be improved, as feedback for assignments came very slowly and much of it wasn’t very insightful. Code infrastructure was great, considering what past students have told me about past iterations of the codebase. However, there are some oddities to it (why is the build script in javascript?).
Course infrastructure wasn’t too great considering it was the first iteration of this class. Aside from the common troubles associated with being a new class, I thought it was a limited overview of the current landscape of generative AI. The course mainly focuses on autoregressive models (think GPT) and diffusion models (think Stable Diffusion), and associated techniques (fine-tuning, in-context learning, distributed training, etc.). I hope they add more to the course in the future.
I did not care much for this course. My overall impression was that goal of this course is for you to become a manual circuit solver. It was great at teaching that, but I wish there was more of a design/problem solving aspect to it.
Good class, because it was a gened I found to be interesting.