I'd argue that, in the case of stable diffusion (and AI in general), the pivotal year was 2017, not 2022. 2017 is when that team of Google researchers invented

I'd argue that, in the case of stable diffusion (and AI in general), the pivotal year was 2017, not 2022. 2017 is when that team of Google researchers invented transformers, and AI (including stable diffusion) has been zooming at light speed ever since. I don't think I've seen a single cutting-edge generative AI since 2018-ish that didn't include a transformer somewhere in its architecture. (Except, I guess, NeRF techniques.) They're even being used to create, effectively, robots that think and plan and communicate those thoughts and plans in natural language.