Efficiency and AI

Efficiency has become much of the default of the ideal of contemporary world, sometimes at the expenses of quality. However, if someone is super-efficient in producing something with (not best but) passable/acceptable quality, we still generally go for it. That seems to be what LLM (or AI more broadly) is giving us (yes, it has 'unexplainable' hallucination, but proportional-wide, overall accuracy is really not bad at all). What does this imply?