Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Good to hear more details on the deal. With this, expecting Apple Intelligence to be way ahead of where it is now. Think at WWDC Apple might announce more.
 
  • Like
Reactions: mganu
"I think that the ChatGPT integration is going to die on the vine... having two large models, given the economies of scale, wouldn't make a ton of sense for Apple," he told FT.

Nonsense - it isn't "two large models".

Apple will be running a version of Gemini which will be invisible to the user and will just become "Siri".

ChatGPT will continue to be an option as an extension. ChatGPT doesn't do anything within the operating system, it never has - it's effectively just a way to use ChatGPT in Shortcuts and a couple of Apple apps - there's no reason to remove that - in fact they're likely to add more extensions, like Claude.
 
  • Like
Reactions: ipedro
It would cost Apple more than $5 Billion to make Siri capable, this is the most financially sound decision.

Correct - but Apple still has to do a lot of custom work on top.

Siri isn't just a chat bot with world knowledge. It has of course go to do things like control a smart home. You don't really want to have to invoke a server side massive LLM to process "turn the lights in the bedroom and office off" but current Siri can't do that either - you want it to be fast and not process the request server side, waiting multiple seconds for the lights to change is no acceptable.

So they need to build a very good router that can forward commands to the local on machine model for smart home control which they'll have to highly tune for that so it's both fast, but can also understand multiple chained commands and multiple ways to ask the same question. This is all going to be a lot harder than slapping a large LLM into Siri.
 
Kneecaps their AI rival in OpenAI as now Gemini will be the default model for every mobile device in the world.

That's not really relevant though as it's a faceless white label version of Gemini - and according to this OpenAI already opted out of providing a version of their models to Apple for this.

So it's not like it's the Gemini model you subscribe to and use - it's just they effectively be giving Apple and openweights version of it to fine tune and run on their own servers - there's no gemini login for it, or any of the other gemini features. It's just going to be the LLM backend basis for Siri.

For the end user they'd honestly not be able to tell the different at this stage for world knowledge between any of the frontier models. Apple will be tuning it and sending a system prompt for short answers as well anyway. What we'll basically get is a more intelligent version of the "Private Cloud Compute" model in Shortcuts.
 
Apple's newly announced partnership with Google to use Gemini models for Siri and Apple Intelligence could be worth as much as $5 billion, according to one analyst's estimate.
To Google, of course. But the question is, how would Apple recoup that money from the users?

By the way, even though it's called Apple Intelligence, Apple actually can't create that intelligence, can it?
 
Google for many years if not over a decade has been paying Apple to assign Google Search as the primary on its devices. Fast-Forward to 2026 and Apple is paying Google for its AI to be incorporated into its devices.
Still a net profit for Apple. That $20 billion oft-quoted by publications is likely outdated and I suspect Google is paying way more than that to Apple these days (a little bird estimated that it might be well over $30 billion by now).
 
With Google being Safari's default search engine, Apple is actually a very good customer of Google rather than a competitor. Android, the world's largest used phone/tablet OS, by the way, belongs to Google.
 
As long as I can turn it off I am okay with this. And hopefully the setting to disable is not buried like it is on my Pixel 8 Pro.
So my question is, can Apple let us turn off the use of Gemini trained LLMs or would that effectively disable New Siri?

I’m thinking about how today we can turnoff ChatGPT and still ask Siri to add something to the grocery list.
 
I can relate to people complaining about the decision to work with Google on A.I. However, I trust that Apple will put safeguards in place to protect our privacy. If they don’t, Apple loses the high ground on privacy in a spectacular way.
It’s a situation where my head and gut are not agreeing.

Once I understood they are training Apple models with Gemini data rather than running Google code, I’m less concerned.

But my gut just doesn’t like any interaction with Google. 🤷🏻‍♂️
 
I see why they went with Google. Google gives Apple their LLMs, and Apple gives Google a coupon for 25% off being their default search engine. Haha
 
  • Haha
Reactions: I7guy
I’m thinking about how today we can turnoff ChatGPT and still ask Siri to add something to the grocery list.
A grocery list? Such a simple task—do you really need Siri for that? A piece of paper and a pencil would do the trick, the old-fashioned way.
 
Because of bad history in the past Apple did not want to give NVIDIA any money! Now Apple can use Google's TPUs instead of NVIDIA's chips for AI!
 
That's not really relevant though as it's a faceless white label version of Gemini - and according to this OpenAI already opted out of providing a version of their models to Apple for this.

So it's not like it's the Gemini model you subscribe to and use - it's just they effectively be giving Apple and openweights version of it to fine tune and run on their own servers - there's no gemini login for it, or any of the other gemini features. It's just going to be the LLM backend basis for Siri.

For the end user they'd honestly not be able to tell the different at this stage for world knowledge between any of the frontier models. Apple will be tuning it and sending a system prompt for short answers as well anyway. What we'll basically get is a more intelligent version of the "Private Cloud Compute" model in Shortcuts.
It is relevant because it’s keeps OpenAI from partnering with Apple further. That’s my point.
 
What world are you living in where Meta doesn't exist? ;)
Fair comment. But I think Google is pervasive on the web in a way that Meta isn’t. I’ve made comments on third-party, open-to-crawlers forums using Safari on an iPad which Google within a few days associated with me, and which popped up in my YouTube feed using the exact same, very distinctive keywords.

If you were to ask me, would I prefer Google AI which worked on my Apple devices over Apple AI which was flaky and never got used? I think I would say Apple should make this deal. Considering they cannot afford to let Android win the race to being first with a working AI assistant mode, I think it’s sensible. What we will end up with is parity between iOS and Android on this front.

Considering the choice between Google and OpenAI, I think Google are in the better position. Their AI team at Deepmind has a long background in using AI for decision making, which they have been training on playing games. I think the learning from that will give them a significant advantage in creating AI which doesn’t screw up so often.
 
That's not really relevant though as it's a faceless white label version of Gemini - and according to this OpenAI already opted out of providing a version of their models to Apple for this.

So it's not like it's the Gemini model you subscribe to and use - it's just they effectively be giving Apple and openweights version of it to fine tune and run on their own servers - there's no gemini login for it, or any of the other gemini features. It's just going to be the LLM backend basis for Siri.

It is probably more than just the open weights and 'good luck'. OpenAI passed on doing the porting work. That extremely likely means that Google had to contribute to the porting work for their model also. It would be dubious path if Apple held bake-off between Google , Claude , and perhaps others on a solution and had to run around doing 100% of all the fine tuning from scratch themselves. Especially, if trying to get it all done on a reasonably short timeline. They would need access to expertise on the model's specific tendencies and not spend tons of time 'rediscovering the wheel'. That said I doubt Google is doing 100% of porting work, but it probably isn't 0% either.

Would make sense for Google ( and Claude , etc) to get expertise on making models run on Apple Silicon also for their own independent apps could leverage local processing when possible also. OpenAI seems mostly intent on building the most colossal cloud thing possible. In addition to broad ecosystem around that cloud first brain also. Their set up is more like a black hole that is out to suck up everything. Not be someone's good partner.
 
It is relevant because it’s keeps OpenAI from partnering with Apple further. That’s my point.


OpenAI doing extensive personnel raids on Apple wasn't going to help that situation either.
Besides that I think there are big philosophical differences between where Apple is going ( keep it local if you can) and OpenAI ( most colossal cloud brain possible ). I don't think Apple wants PCC to be the biggest collection of hardware on the planet with the most power consumption possible.
 
Apple's statement explicitly says it runs on "Google Cloud"
The statement referred to their cloud technology which is apparently part of the initial training of the models, but the models themselves will then be customized for Apple and will still run on Apple's own silicon servers (for PCC).
 
So my question is, can Apple let us turn off the use of Gemini trained LLMs or would that effectively disable New Siri?

I’m thinking about how today we can turnoff ChatGPT and still ask Siri to add something to the grocery list.
Pretty sure that would disable anything under the "Apple Intelligence" umbrella, so maybe a basic version of Siri would still be available but not the "new" Siri. Depends on how they roll it out I suppose.
 
  • Like
Reactions: fatTribble
A grocery list? Such a simple task—do you really need Siri for that? A piece of paper and a pencil would do the trick, the old-fashioned way.
I don’t think there is a right and a wrong. I can rattle off my list to Siri much faster and easier than writing it. If I want to check off items in the store I don’t need to carry a pen.

But not the type of thing I have strong feelings about.
 
I don’t think there is a right and a wrong. I can rattle off my list to Siri much faster and easier than writing it. If I want to check off items in the store I don’t need to carry a pen.

But not the type of thing I have strong feelings about.
I like that I can still read the list off my iPhone even if I accidentally spill coffee on it.
 
  • Haha
Reactions: fatTribble
Apple is a competitor to Google in a way, with Android and iOS in direct competition, but Apple actually generates revenue for Google, not the other way around. Apple is a customer to Google, Samsung, LG, and others, as well as China, India, and Vietnam. When the Apple-flock buys Apple devices, they also indirectly contribute a portion of their money to them.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.