> No, all they need to do is use the client device to perform speech-to-text On an iPhone, Apple has the full run of all device capabilities, and can even ship

> No, all they need to do is use the client device to perform speech-to-text On an iPhone, Apple has the full run of all device capabilities, and can even ship custom chips, but they still stream audio back to a server to process what people say to Siri. Speech to text is very hard even when the speaker is purposefully speaking to the device. It would be way harder to try to extract meaning out of ambient audio.