I don't really mind that they aren't on the LLM bandwagon, but Siri seems to have stagnated. The big "Apple Intelligence" capabilities of the iPhone 16 haven't exactly landed. The Vision Pro seems to be on at least a partial depreciation path.
The only real innovation I've seen in the last decade has been the M line of chips. Mind you, these are undeniably really good; but even that hasn't changed the market share that much (though it is going up and trending well).
It is objectively worse at calling people than Assistant was. If I ask you to call someone, don't come up with a scolling list of phone numbers that I have to pick from. At least Assistant called the primary designated number for someone, Gemini just froze and wouldn't take voice commands to pick the number but forced my to pick up my phone.
I turned that bullshit off a couple of days after they forced it on me without asking.
I mean not for looking up the information, but for something that alter how you use phone.
Unless you want to turn it off, which I haven't been able to figure out how to do. Every now and then my phone will randomly prompt me to "ask Gemini", which is really annoying. When I want to use the LLM, I will go to it, stop shoving it in my face over and over.
It queues up music correctly, and picks the right destination on maps in my car. 98% use case satisfied. Would I like it to be better? Don’t really care. Is it a purchasing point? Nope. Would I miss it if it disappeared tomorrow? Also nope.
- setting a timer; - sending a text; - starting a call; - adding an item to a shopping list; - playing/controlling music (very occasionally);
Siri does all of these with 95% accuracy. Occasionally it mishears "15 minutes" as "50 minutes" if I'm rushed or something.
You use the device however it works for you, but I truly cannot comprehend the use case of a voice assistant beyond these type of tasks.
More complex tasks would likely require more concentration on the response/result, at which point I shouldn't need to do it hands free.
Now, maybe it would be justifiable if there were great local AI experiences on iPhone, or an easy $5 trillion to be made elsewhere. Until then, Apple is bleeding money hand-over-fist by refusing to sign the CUDA UNIX drivers and sell the rackmount Mac as a cutting-edge TSMC inference box. The Grace superchip is absolutely eating Apple's ARM lunch right now.