The forthcoming release of the Stable Diffusion image generation model in August 2022 was pivotal. I wrote how Stable Diffusion is a big deal at the time.
People could now generate images from the text on their hardware!
More importantly, developers could mess around with the guts of what was going on
Somewhat surprisingly, language models like GPT-3 that power tools like ChatGPT are a lot larger and more expensive to build and operate than image generation models
LLaMA on its own isn't much good if it's still too hard to run it on a personal laptop
