[rueynshard]

Wouldn't the user's speech data need to be sent to Apple to convert to text, or identify the intent first?

[peckrob]

We were doing speech recognition 20 years ago without any kind of networking. I dictated part of a term paper in 1995. Dragon Dictate I think is what it was called. You could even navigate the word processor menus and UI with it, or say things like “make that bold” and it usually worked! Just a bit less often often than Siri works actually.

Sure it was more of a novelty, and had to be trained on your voice. But that was 20 years ago.

[walterbell]

Not only is Dragon still a product, it is owned by Nuance, which helped to develop Siri and drew on SRI research, https://www.forbes.com/sites/rogerkay/2014/03/24/behind-appl...

[Angostura]

Hence Siri’s name.

[systoll]

Also, a purely local 'Voice Control' was standard in iOS 3-4, and it's still there in iOS12... just, buried.

[MiddleEndian]

Even Apple had this working a long long time ago in classic mac os. I used to launch all my applications via voice til I got bored of it.

[Razengan]

macOS still has non-Siri voice-control under Accessibility Preferences.

[Angostura]

And iOS has very good non-Siri dictation

[saagarjha]

I believe that still requires the use of Apple’s servers.

[Angostura]

Having just double-checked, yes you are correct.