Applications (of arguable value) are limited for the current capabilities of the technology, and free alternatives like llama, vicuna and wizard are popping up. Will be interesting to see how OpenAI operates.
Check out Hugging Face. It's cool to be able to try out different models to see how they perform. RAM requirements can get steep for the larger models.
I noticed that offline LLM builds running on personal computers are now possible, but it seemed like all the solutions required the installation of dependencies, so I created a containerized solution that makes it easy to swap out the model in use: https://github.com/paolo-g/uillem