Why would they want to run LLM queries every time someone asks Siri for the normal shit people ask their voice assistants: weather, timers, controlling music; on a scale of every iPhone user out there.
It'd just be waisting compute money with no obvious way of recouping the cost. Doubt many people would be interested in paying for Siri.
They already run compute for siri for voice to text. So they could in theory still provide an LLM backed siri service for free for their customers/users.
It'd just be waisting compute money with no obvious way of recouping the cost. Doubt many people would be interested in paying for Siri.