I really thought on-board voice would be a game changer. Not having to do the round-trip to the cloud to determine intent, and providing a conversational interf

I really thought on-board voice would be a game changer. Not having to do the round-trip to the cloud to determine intent, and providing a conversational interface (i.e. "now send this to Pam") I thought would generate that buzz. It sure seemed to when Google announced that they had reduced this ML model to ~500mb so it could run locally. In contrast, Apple's A13 chip's contribution to extend battery power is incredibly impressive to me.