The way Siri worked up to now is that you would talk into your Watch’s microphone, your audio samples are transcribed into text on device, but interpretation of that text required sending that text to Apple’s cloud of Siri servers. The watch would then have to wait for a response back before being able to enact your request.
It doesn’t matter if what you were asking was trivial or obvious, even for the simplest queries, the Watch needed to talk to a server in order to actually respond to your request. And the Watch’s network connection is slow and unreliable.
The Watch falls back on Bluetooth relay via the iPhone whenever it can to maximize power efficiency, but this isn’t exactly the fastest method of wireless networking. Out of range of the iPhone, the situation is even worse as you are then at the mercy of the Watch’s small and weak Wi-Fi and/or cellular radios.
This is where the Series 9 could be a big step up and the Siri loading spinner could become a thing of the past. Backed by the increased power of the S9 chip, Apple is running natural language machine learning models on the watch itself, so some queries can be handled entirely offline, without talking to a server at all. There is no longer a dependency on the Watch’s flakey internet connection, which should dramatically speed up common actions.
I am relatively confident that this will work as advertised as Apple already rolled out a similar change to Siri on iPhones starting in 2021, improving response times and privacy.
The one big question is what kinds of “basic” queries will the Watch will be able to handle on its own, and what will still need to fall back to server-side processing. That list hasn’t been confirmed yet. We’ll find out for sure when the Series 9 launches next week, but I’m confident that two things I really care about will be covered — starting workouts and setting timers. Apple used workouts as the Siri demo during the keynote, so that’s already a surety.