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
Whatever it takes, I guess? If they’re marketing an AI that can do a lot then seemingly simple requests like that should be possible. Otherwise- it sucks. I’m being harsh but really, it can either do those “seemingly” simple things or people lose confidence in it.
Apple has to start working on making iPhone 16 Pro Max last longer than 2 hours.
Now you'll need 16GB RAM and >18 trillion op/s NPU to set a timer.
I was excited to see apple intelligence being added to to macOS 15.1, but I found it pretty useless without the 15.2 GPT integration.
I thought siri was going to be able to provide me with some ideas, or helping develop a topic, but it was only able to do some simple automations (such as adding reminders and whatnot).
I think the gpt integration will make siri at least a little useful.
I don't want to tout Microsoft's horn (a company I'm very critical of), but I found the free version of copilot that comes to microsoft edge much more useful, especially with teaching me how to do things or developing creative ideas, so much that I installed edge on my mac just to use copilot.
The thing is that apple has such a great potential into making a clean OS with some useful ai into it.
Windows 11 is super bloated, and the operative system feels like a big commercial for microsoft apps.
I also find the implementation of copilot to be confusing, as there's a free tier, a paid tier, a special tier for adding some pf copilot functionality to Office, an other tier with copilot studio.
Apple's implementation of AI within the system seems much cleaner and more elegant; what they'd need is just an in house competitive LLM.
I'd even pay for a premium version of the tool, if a discount would be made accessible via apple one.
I thought siri was going to be able to provide me with some ideas, or helping develop a topic, but it was only able to do some simple automations (such as adding reminders and whatnot).
I think the gpt integration will make siri at least a little useful.
I don't want to tout Microsoft's horn (a company I'm very critical of), but I found the free version of copilot that comes to microsoft edge much more useful, especially with teaching me how to do things or developing creative ideas, so much that I installed edge on my mac just to use copilot.
The thing is that apple has such a great potential into making a clean OS with some useful ai into it.
Windows 11 is super bloated, and the operative system feels like a big commercial for microsoft apps.
I also find the implementation of copilot to be confusing, as there's a free tier, a paid tier, a special tier for adding some pf copilot functionality to Office, an other tier with copilot studio.
Apple's implementation of AI within the system seems much cleaner and more elegant; what they'd need is just an in house competitive LLM.
I'd even pay for a premium version of the tool, if a discount would be made accessible via apple one.
Last edited:
My iphone 13 is more than 3 years old, but I noticed quite the reduction in battery since I installed iOS 18.
Not long ago, I could leave my phone on the night stand with 30% of battery and find it with 20% the next morning.
These days I find the phone almost dead each morning, with the same average battery starting point.
And I have been very gentle with my charging, charging 45% of the time with a slow 7.5w Qi charger, 45% of the time with a super slow 5w cable and only the remaining 10% of the time with a 20w cable.
Going out for any reasonable amount of time, with no powebank, has become impossible.
I'm quite sure the new siri will be quite good, but the 2026 release date clearly demonstrates what happened.
Apple was no doubt caught off-guard by the absolute insanity GPT 3 was at launch.
Microsoft (and other companies) decided to build around Open AI, while Apple clearly decided to use the ghat gpt integration as a temporary measure.
And that's a smart idea, has having no real AI in their OS by late 2025 would have been hell for the company.
I grow ever so frustrated in seeing apple expanding into questionable new ventures (such as TVs and smart home hubs).
I'd invest all the budget allocated for these things into ipdating siri, otherwise the company will be destined to play forever catch up with open ai.
Since apple has the $$$ I expect Siri in 2026 to be like the iPhone in 2007.
Here's Siri today, on my Macbook Pro M1 running macOS 15.2 beta 4, while Chrome was the frontmost app, Apple Intelligence was theoretically active, and Siri was linked to my paid ChatGPT account:
I mean, as if it had to repeat it twice
I mean, as if it had to repeat it twice
Last edited:
Maybe Apple should just buy OpenAI?
We'll get smart Siri sooner...and one that is keeping pace with competitors.
We'll get smart Siri sooner...and one that is keeping pace with competitors.
Not sure that's currently possible, since Microsoft now has a pretty big stake in OpenAI.
To be fair, if you enable the chatGPT integration in the current 18.2 beta, “I found this on the web” is basically gone.
Yes and no.Even hardcore fans have to admit that the 14-year pioneering role with Siri was a bit of a waste
Yes, Apple was the first company to create (purchase) a modern voice assistant, then spent years dropping the ball.
On the other hand, I don’t think any other company has really been doing much impressive.
And I don’t necessarily count ChatGPT in that, that’s a totally different animal.
But I more meant the things Siri has actually been advertised to do. The basic household tasks like timers and alarms, home control, Music.
When it comes to all of these tasks, from my experience, Siri, Alexa, and the various incarnations of Google Assistant/Google Now are all still and always have been hit or miss.
Alexa has a lot of advantages over the other two, but even it has the trade-off of being riddled with advertisements after every couple queries.
But none of them are great, they all are just passable.
At least they have “concepts of a plan”.
Two years behind is being generous. ChatGPT came out 2 years ago
Two years behind is being generous. ChatGPT came out 2 years ago
i wish Apple would change the voice command to “Hey Apple”. Would be cooler than saying “hey siri”
Hope for better Siri keeps me alive
The battery life, after upgrading to iOS 18.1 is fracking terrible, it's a joke, on the 13(16) pro max. It isn't the phone, it's iOS' issue.
Even more - long term this idea of on-device processing is pretty dumb. Is the network latency really that high that we need local llm’s?
Generative AI (like an LLM) is just a text/image/sound generation and transformation machine. Connecting it to actual functionality (like if it should control all kinds of iOS and app functions) in a robust way is not straightforward at all. It doesn’t have any concept of agency, or of time passing. Similarly for maintaining state/memory for it, because it doesn’t remember anything by itself. Everything it might need to know from prior contexts has to be re-fed to it internally for every single interaction. This is why most present-day AI functionality is oriented around generating or transforming some text or media, or around search.
There’s a lot of work to be done to make it perform functions well that would seem straightforward to a human.
We are also assuming that the competition will still be around two years from today. A lot of things can happen in that time. I recall Microsoft made quite a splash with their flashy AI announcements and investment of openAI, and today, we see more problems crop up with the company and controversy surrounding Copilot.
As the saying goes - he who laughs last, laughs best.
As the saying goes - he who laughs last, laughs best.
On-device processing will become even more desirable in the long term, especially with increases in the amount of data crunched and the number of times we use AI in a day. There are several reasons, among them being:
• Network latency can be pretty minimal in most uses, but there are and will continue to be situations where latency is worse, including spotty or no Internet connection. And while we often get almost instant responses when we type or speak at an AI when we have a good Internet connection, often the response time can still sometimes be annoyingly slow for whatever reasons, and on-device processing is supposed to reduce this issue. You wouldn't want extra, annoying lag times while using FaceID, language translation, augmented reality, etc.
• Many people are concerned about privacy issues in which personal data they reveal about themselves is sent to the cloud while "conversing" with an AI. On-device processing can minimize this issue.