Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Too bad it will probably be English only (or the top 5 languages) for a year before we get in in Norwegian and other languages. We just got Norwegian Apple Intelligence in 26.1, so new Siri in 28.1 then? 😛
 
I'd be happy if Siri would be able to permanently remove any rap-crap suggestions in Apple Music. Simply saying "I don't like that song" just isn't enough.
 
  • Like
Reactions: arkitect
"AI"...

So nothing new and worthy, no fixes to the iPhone<>iPad<>Mac syncing issues, reflecting files' true sizes on a folder, complete removal of photos from devices (Zombie media), etc, etc...

But "Siri AI will be here any year now!"... Gods, AI can't implode fast enough.
 
  • Disagree
Reactions: citysnaps
Good. I won’t have to be embarrassed for the crap of software that Apple has been developing last 5 years and the need of Google to step in as I still be using iOS 18 in 2027 with almost a 99% of probability.
 
  • Like
Reactions: klasma
I can’t see this changing how I use Siri one bit: that is, not at all.

Unless it learns to understand “mute notifications from WhatsApp for one hour” when it keeps interrupting my podcasts. I might use that.
 
This article seems incredibly misleading and uninformed and a little ignorant on LLMs. Also, try to give credit to your sources. First of all, the all hands on deck meeting is irrelevant. Any expectations from that meeting were while Apple still was working on its own Siri. That was scrapped later and they'll use Gemini but don't assume prior expectations are still in place on what this Gemini siri can do in ios 26 or ios 27. Especially ios 26 when you start mentioning personal context, on screen awareness, and app integration. That was pre Gemini Siri.

Par down those expectations. Don't raise them. For ios 26 anyways, this won't be a bigger upgrade than promised. It'll be smaller. I guarantee you none of what you mentioned will be available in the Spring in terms of on screen awareness, app integration, and personal context. These so called expectations were made irrelevant when Apple scrapped its in house siri.
 
Even if they somehow pull it off it’ll probably only work on iPhone devices purchased in the last year or two, everyone else will get the dumb Siri 🙄 📱
 
What I am missing is, what was Federighi referring to in August of 2025 when he said that Apple had successfully revamped Siri? How does that align with going to use Gemini now?
 
Today, I can ask Siri: What is the volume level right now and she'll be like, sorry I can't help you with that, but I can say set the volume to 50% and she does it.

Another weird one is I'll be wearing my AirPods and I'll say, turn on Spatial Audio Fixed and she doesn't know how to do that either. So dumb.

Agreed, that is frustrating.

But that's precisely the issue. Each "task" needs to be a programmed set of instructions and it's unreasonable to think that Apple has baked in every single task into the current Siri. Volume Up. Volume Down. Current Volume. That's just three out of potentially thousands.

An LLM-based Siri will have access to "tools" that can do far more than the current task-based Siri. The new Siri will understand what "Volume" is in the real world (or at least on the device), how volume works (eg. can go up, down, has a current state), etc. and be able to respond accordingly.

Knowledge Navigator... here we come!

🤓
 
I'm curious if this change will remove the distinction between Siri on/off and Apple Intelligence on/off.

I'm curious about that, too. Will people be able to turn off Apple Intelligence but still have access to the new Siri?

Will the new Siri require an Internet connection to work, even for simple tasks like setting timers?
 
From what I can tell, the LLMs people use today are economically unfeasible and won't be around for long. They require huge data centers, huge amounts of water, tremendous amounts of electricity and cost trillions of dollars. Right now they're running these things on hopium.
This is incorrect. Inference already runs at profitable levels. It’s only the R&D and training costs that aren’t covered. There is zero question that LLMs are here to stay.
 
“Hey Siri, set a timer for 10 minutes” to which she replies, “I found this on the web”. I personally just hope that when I tell her to navigate to Paddy Caughlin’s in Fort Atkinson, Wisconsin, she doesn’t decide I’m wanting to go to a restaurant in San Francisco.

Managing timers and alarms works flawlessly for me every time. Zero issues, even when prompting her (on my phone) from the next room. Works great.

I appreciate that Apple is easing into an LLM-based Siri, not going full chatbot, but taking a more hybrid approach for a smooth consistent user experience. It would be disastrous if asking Siri to set an alarm set off the house alarm system.
 
From what I can tell, the LLMs people use today are economically unfeasible and won't be around for long. They require huge data centers, huge amounts of water, tremendous amounts of electricity and cost trillions of dollars. Right now they're running these things on hopium.

That's what is being reported in the media, yes, but is a far over-generalization of the reality. There are massive efforts being done in many facets of AI that don't fall victim to these datacenter issues. The past few months, there has been a move towards using smaller specialized models. Because model technology is getting better, and new supplementary technologies are being developed, such as Model Context Protocol, researchers are finding that smaller models can often out-perform the big commercialized models. Inference is faster and vastly cheaper.

AI is still very young in the grand scheme of things. It's just moved so quickly, we all have whiplash.
 
From the description it looks like it will be something similar to Windows Recall + Copilot. Everyone hates Recall, and not many like Copilot. I'm wondering if this is gonna be different when Apple does conceptually the same thing, but it's just not gonna be Microsoft 🙂
Copilot is a combination of bad marketing with bad ability and implementation. Something like Clawdbot is what people really want: https://www.jakequist.com/thoughts/openclaw-is-what-apple-intelligence-should-have-been
 
No rumors that the new capabilities will not work on 8 GB RAM machines yet? Or will you have to shut down everything else to run the model?
Only purpose-bound models will run on-device, like for categorizing your emails and such, and maybe recognizing current-Siri-level commands. The general-purpose conversational intelligence will be running in the cloud.
 
All I want from Siri is for it to understand my voice consistently, and to be able to use that to control my iphone. Nothing fancier. Something akin to the crew of the Enterprise conferring with the ship's computer.
I’d also like it to be able to remote-control other iDevices in my home, like having it start/stop something on the iPad over in the next room, or set an alarm/timer on another device. Even current Siri should be “smart” enough to do stuff like that, but Apple hasn’t implemented it. It wouldn’t surprise me if “new Siri” doesn’t change that, and will still be limited to the use cases Apple happens to come up with.
 
What I don’t understand is if Siri is just programmed to respond to your voice when you are asking personal questions like read my new emails or calendar entries, why is it ever saying you need to unlock your iPad first? If you are the one that is asking, and voice recognition recognizes your voice, why does that have to be unlocked first?
Because others could then perform those actions by recording and playing back your voice. Or nowadays, by training a voice AI with voice samples recorded from you and then having it say anything they want with your voice. Someone’s voice is easily reproducible, compared to Face ID and Touch ID.

On my iPad mini when I first wake up in the morning, I want to know if I have had any text messages before I even get up, but I don’t wanna have to unlock the iPad first.
At least it has Touch ID so you can do it with closed eyes. Also, you can get a HomePod mini for that use case.
 
  • Like
Reactions: johnsawyercjs
People call systems like ChatGPT and Siri “Artificial Intelligence” because it’s a convenient umbrella term, but strictly speaking they aren’t intelligent in the way humans (or sci-fi AIs) are.
That’s what AGI — artificial general intelligence — stands for. Everyone is well aware of that.
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.