Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Working on 'LLM Siri' for 2026 Launch
- Thread starter MacRumors
- Start date
- Sort by reaction score
Apple should have wanted this 10 years ago.
Limited Learning Machine sounds about right!😎
What about having conversations with humans? You can do it now!
Bet it will lean strongly in one direction.
This new Siri will require the iPhone 17 Pro Max Ultra Plus so everyone will have to upgrade again. But the feature won’t be released until right before the iPhone 18 Pro Max Ultra Plus
If it's still an LLM technology, most people might disagree with me, but I'll say it : I don't think it matters. It will be a HUGE improvement over the current Siri.
If the goal is to be able to ask day-to-day tasks to Siri (turn ON/OFF the lights for example), it's going to be hard to tell the difference with a good LLM and a bad LLM, because the technology itself performs very well. Even with the GPT 3.5 engine from 2 years ago, it's perfect for this.
If the goal is to be able to ask anything and not get hallucination (i.e. math questions, history etc.), then OK, yes, it matters a lot. GPT 3.5 had way too many problems with these questions. But Apple doesn't seem to have this intention. At least not yet. They're just using whoever is best at these and connecting to it (ChatGPT).
Apple should just admit they now suck at writing code and just quit doing it. Maybe ask Google or Tesla to write a new Siri in their spare time. I'm sure they'd be glad to help.
But on another note, does anyone actually use Siri, like in the commercials? I doubt it.
But on another note, does anyone actually use Siri, like in the commercials? I doubt it.
I use Copilot at work. It’s quite good. Better than anything I’ve seen with Apple Intelligence.
What exactly has John G. done since joining Apple ???? From Wikipedia: "In December 2018, it was announced that Giannandrea had been appointed Senior Vice President of Machine Learning and Artificial Intelligence Strategy at Apple..." Less verbosely, he was hired to develop Siri. How many RSU/Options has he collected since then ??? Man o' man, Apple really, really, really needs fresh blood and ideas at the C-Level. This group of middle-aged, white guys has been at it too long, accomplished too little, and appears to damn "clubby". (Yeah I know how happy shareholders have been, but they've run out of the runway Steve gave them. Hence Warren dumping all his stock... piper needs to be paid, and too little too late apparently as tech shifts madly and John G. et. al. got caught counting their money w/o thinking about AI). Sorry to be so negative, but this guy and what he has done (and Craig driving OS quality into the ground)... WTF John G.??? Again sorry. Flame away. NSC
I like when Siri commands I use every day totally break like asking for the weather forecast. "you don't have any items about weather". Thanks Siri.
A bit behind, maybe, but Apple has had the institutional will to move forward in such cases. Remember when Apple Maps was first released. So many people were parroting the received wisdom that “Google won; Apple should give up.” Indeed, Apple directions were laughably bad at the start. Now, they are stellar and since it is part of the Apple ecosystem, it works seamlessly across devices without having to let third-party servers document your travels.
It is actually a good thing for Apple to lag a bit. They can see which of the efforts by others pay off and cherry pick the ideas that are winners for the Apple ecosystem. Not everything implemented in ChatGPT, Gemini, et al., is necessarily a fit for Apple’s system, especially when running on device and emphasizing privacy.
Given that it won’t be available before 2026, “all” seems unlikely.
You mean, available in 2026, but not in the EU because Vesteger has determined that users should be able to choose whether they want Siri, Alexa, Cortana, Google Assistant, or some mythical European to-be-determined-but-never-to-be-produced product as their default AI agent on iPhone.
Apple already has an LLM that's used in things like the writing tools and summaries. Building one out for Siri is an obvious next move, but also one that requires much more polish and attention to detail.
Without LLMs, every single use case has to be separately integrated by an Apple developer. It’s not much more than sentence pattern matching currently.
Even with LLMs, it won’t be easy to make all sensible use cases work. I would expect many things to still not work when it launches in 2026, and to be added bit by bit over time.
Last edited:
13 years later and Siri is still dumb as rocks. It’s one step above basic voice control.
And that is the main issue what many fans here do not see. I use also several models (mainly 4o and Claude 3.5 Sonnet), privately and partly for work, and the performance is incredible.
Some people here do not have the slightest clue what these AI models are already capable of, if you also prompt correctly and force reasoning. Complex calculations in the fields of mathematics and IT. Summarisations. Recommendations etc.