How many querys are people going to be using though when they don't see the upfront costs? And how many querys would have been just as solved with a simple google search or wikipedia page that will instead/also be put through an llm and use extra power? Like if I search for the chocolate tempering temperature or the atomic number of tantalum, isn't all that energy just wasted on the companion llm results? It seems like it will add up.
