It's also worth noting that Apple traditionally is not a first mover and looks for "inspiration" from smaller competitors. In this case, there is no comp to reference. There is no startup mobile OS innovating in integrated AI. That, and the supposedly rushed timetable, probably explains a lot.
The lengths Apple went to build a secure and private system will make it stand out and help it hold up to regulatory scrutiny. Doing this now is better than doing it later.
In, let's say, 3 years, the features are more legitimately useful, the gatekeeping on new hardware will be a non-issue. In 3 years the majority of Apple's users will have an iPhone 15 Pro/iPhone 16 or newer. They are probably already mostly on M1 or newer Macs.
On the other hand, I totally agree that it's pretty useless as it stands right now. I also think that if their launch strategy is to have an amazing WWDC and then deliver 5-10% of the features in September, that's going to turn Apple into just another company that promises the moon and delivers gimmicks.
We have seen that play out with all the previous AI chips (mostly Google).
The software and its requirements are moving faster than devices can be shipped and accumulate significant marketshare.
This is not to say thay Google and Apple haven't or won't be able to ship some minor models such as voice recognition or translation, but for frontier level AI the local chips just won't suffice.