What would a learning roadmap for modern AI look like?

For context, the last time I did anything in the field was some introductory Neural Network and Machine Learning courses ~10 years ago at university.

I'm not necessarily looking to move into working in the field, but would like to understand the fundamentals of Stable Diffusion, ChatGPT et al. and get to a point where I can play around with building my own primitive versions.

What would a path to get there look like, assuming I want to take as direct a route as is reasonable?