I haven't delved into the subject myself at all yet, but Ars has an intro to the basics:
https://arstechnica.com/information-...-at-wwdc-2024/
And how they're implementing them into the on-device functionality:
https://arstechnica.com/gadgets/2024...os-and-mac-os/
As well as not-very-concrete assurances of privacy and security:
https://arstechnica.com/ai/2024/06/a...ible-by-apple/
I'm a little worried, personally, that I'll now have a personal assistant that does what Siri was initially promised to do, except blathering complete bullshit at me in between actual info, making it useless and/or dangerous.
I'm curious how Apple can jump on the bandwagon while radically vetting results to avoid the dystopian sinkhole/hellhole that is any non-expert usage of ChatGPT.
(I use it to teach myself Python, but that comes with the analysis and reality-testing that reveals the bullshit it offers up occasionally in between all the Truly Useful.)