How useful is it to argue about what would happen in the extraordinarily unlikely eventuality that all LLM development will cease in 2024, wherein everyone with the proclivity to use LLMs will be stuck with exactly these same models for decades to come?