Thank for you this, but in an ideal world it wouldn't be necessary!
If there's one thing that "AI" could bring to client PCs/smartphones it would be a browser providing instant convenient translation of any language (which it would infer from context) to the language of one's choice. It need not be perfect, and wouldn't be expected to properly translate idioms and so forth, but being able to hover your mouse over the part of an image containing those characters and see "multi core" (or even a possibly somewhat wrong translation like "many piece" would be good enough, with the non-A "I" between our ears filling in the rest for us) pop up would be wonderful.
You can feed a URL into a translator and it'll translate the text of the web site all right, but not the graphics - and that's often the most interesting content when it comes to stuff like we discuss at AT. Now sure someone will point out that OCR software exists that could do this, then you could feed the result into a translator, but the more steps required the less it is worth the time it takes.
So c'mon Apple, let's get that NPU you've included in every iPhone since 2017 working on this when you run Safari! C'mon Mozilla, Google, Microsoft - if the latter company believes in its "AI PC" hype here's a perfect opportunity to show a real use case, instead of handwaving about how great building "AI" into every PC will be without any concrete examples of useful things that can be via that local processing capability.
Here's a browser update that would actually help the end user, rather than the 99% of browser updates that help web site designers by supporting expansions of HTML, etc.