Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
I see nuance is still the default in the comments… I think it’s wild that people don’t know the difference between the Gemini foundation models and Gemini, the service. Apple will be using the foundation models. They’ll run either locally or on Apple servers, and a model doesn’t feed data back anywhere by default; that would be a product decision otherwise.

Is Apple late? Yes. Did they butcher their AI rollout? Yes. Should they be trolled for it? Yes. But in the end, this is probably the best choice for users right now, provided they deliver on quality and privacy (which, again, they can if they choose to do so).

Whether we like it or not, AI is becoming a commodity, and I don’t think it will matter much which model runs in the background. The interface and the experience will matter. With this deal, Apple can take their time to do something in-house, or not. They might even have saved themselves money in the long run.
 
I see nuance is still the default in the comments… I think it’s wild that people don’t know the difference between the Gemini foundation models and Gemini, the service. Apple will be using the foundation models. They’ll run either locally or on Apple servers, and a model doesn’t feed data back anywhere by default; that would be a product decision otherwise.

Is Apple late? Yes. Did they butcher their AI rollout? Yes. Should they be trolled for it? Yes. But in the end, this is probably the best choice for users right now, provided they deliver on quality and privacy (which, again, they can if they choose to do so).

Whether we like it or not, AI is becoming a commodity, and I don’t think it will matter much which model runs in the background. The interface and the experience will matter. With this deal, Apple can take their time to do something in-house, or not. They might even have saved themselves money in the long run.
Do you really think most of the general public can understand any kind of nuance ? and it does not matter they would complain anyway
 
Same. It reminds me of when Google was the default maps app on iOS, then Apple released their own version which was trash for a long time. The order of operations is just reversed this time around.
Apple Maps is still trash in many countries. I prefer the graphical presentation in Apple Maps and it integrates well with the rest of the ecosystem of course but when it comes to the quality of its ground truth in many countries it's miles behind that of Google Maps.

I've been on holiday in South Africa and, while I always use Google Maps first, out of interest I thought I'd also try my search for craft beer places near me on Apple Maps to see if it would come up with any extra options and lo and behold it did - a place that was open only about 30 metres from where I was staying! Unfortunately it turned out that place had closed down 11 years ago. Apple Maps came up with quite a few other places within about a 2 kilometre radius that had closed down a year or more ago.

I'd be surprised if it happens, I suspect Apple is now too stubbornly wedded to keeping its Maps app an all-Apple affair, but I'm still hoping against hope that this new Google agreement might also bring Google ground truth to Apple Maps. "What time does <some shop> close?", "How long will it take me to walk to <somewhere>?" or "When is the next train to <somewhere>?" are quite common queries for me when I interact with my smart speakers (a mix of Amazon and Google/Nest devices). Google Assistant gives by far the best answers and I think that's down to a combination of Google Assistant being generally smarter than Alexa in terms of determining context, intent and intelligent phrasing for its answers and secondly Google ground truth being so much better than anyone else's. (Amazon Alexa ground truth is truly awful at least in the UK, much worse than Apple Maps. When a new supermarket opened about 100 metres from my home Google knew about it the following day. It took Alexa 6 months before it realised it was there and would stop giving me opening hours for the next closest branch about 5 kilometres away.)
 
Does MacRumors ever feel shame?
With all these Google/Gemini news, how do they feel about their „Apple‘s AI strategy will pay off“ articles?
What „strategy“?
They fumbled Apple Intelligence, then they fumbled the ChatGPT integration, and now they lean on Gemini.
This isn’t a strategy, and if it was, I’d like to see the roadmap. Quite a lot of twists and turns, no?
 
what a fumble man. They were building their cash war-chest over the last decade for the next big thing, and the next big thing flew right under their noses as they were instead chasing goggles technology and now they have to make a deal with their nemesis. What a fumble.

What they would do to rewind the clock 5 years.
Apple's next "big thing" was memojis.
 
  • Haha
Reactions: carswell
Don't even go there! I can just see it now. Every response you get from a HomePod will in future always end with ".... this response bought to you thanks to Google Gemini technology".

I was thinking more along the lines of a little subtext in the settings menu or the like, not something so blatant and annoying. Kinda like original maps had the Google logo in the corner.
 
We cannot know that. And we know Apple marketing provides a lot of BS. So while I cannot dispute the privacy issue, neither can you support it with any proof other than Apple marketing and propaganda.
Google sells the LLM to be used (all processing done) on Apple's private servers. Google essentially only provides the software and probably some integration support/expertise. But none of the data processed will be shared with Google.
This type of partnership is not uncommon in the world of cloud services.
 
As a leadership decision this has to be terrible for morale for the developers at Apple. Tim Cook is a great ans terrible business person at the same time

It depends on what's going on inside Apple's AI R&D teams. We don't know, or at least I definitely don't - and I admit that all the high profile departures for Meta etc are not encouraging - but it is just possible that there is another way to spin this.

LLMs as implemented today are almost certainly not the end point. There is a lot of speculation that simply scaling up existing techniques is going to hit a barrier fairly soon and that some paradigm shifts might be necessary to make significant further gains with some encouraging results from some relatively recent stuff but still much more work to do.

It's just possible that at least those Apple AI developers previously working on Apple's in-house LLM are relieved to no longer have to get a solution, probably based on pretty well-explored techniques, out of the door in a product-ready state ASAP but instead can now work at least for a while on more experimental next-generation stuff while this Google/Gemini partnership saves them from having to get an actual product shipped in the immediate future.

I'm not saying I'm right, just offering up another possibility.
 
I never saw the point of Siri on Mac. I've had it disabled since whenever it was added to macOS.

Same as you, I keep AI turned off. Siri is on enabled on my phone because you can't use carplay if it's disabled.
Like Cortana on Windows that never took off.
Maybe if you have a microphone then maybe Siri could be useful, but I don;lt on this machine, well not permanently attached, I did have a muck around with typing text into it, but got bored with that.
 
kinda odd to partner with your most direct competition yet now apple is partnering with google on such core thing and of course Samsung on many parts like screens.
I heard that before. I think it was when Steve Jobs announced office for Mac and MS Explorer for Mac. And, in retrospect, it turned out not too shabbily for Apple.
 
I mapped Gemini to my Action button anyway. This deal makes sense imo, but many wont like it due to distrust of Google.
 
Health data is encrypted with the private key on your devices Secure Enclave. It’s not accessible by Apple.
Not accessible today. Who could stop Apple if they decided to change how this works? If Apple wrote code to migrate this data to their private key from your private key, how could they be stopped? (Honest question)
 
Not accessible today. Who could stop Apple if they decided to change how this works? If Apple wrote code to migrate this data to their private key from your private key, how could they be stopped? (Honest question)
If Apple wrote software to do that, the data obtained would remain encrypted to your private key. You can't just migrate data from one key to another. Encrypted data is encrypted for everyone. Apple doesn't have a universal key that can decrypt this data: that's the reason encryption is such a sticking point for certain governments around the world.
 
Google sells the LLM to be used (all processing done) on Apple's private servers. Google essentially only provides the software and probably some integration support/expertise. But none of the data processed will be shared with Google.
This type of partnership is not uncommon in the world of cloud services.
Again, how do you know that the Google Software does not share any data, even when it is running on Apple hardware, other than Apple marketing (propaganda)?
 
Is all of the technical documentation they post on this and other things (like iCloud Private Relay, Advanced Data Protection, etc) "Apple marketing and propaganda"? I expect you're the type of person that, no matter what Apple would put out there, wouldn't believe it anyhow.
As someone who has had a processional responsibility to see through agreement loopholes many, but not all things, are marketing propaganda. There is a clear difference between a marketing position and a legal agreement.

Does the Apple legal agreement with respect to the user state that no information is shared with Google. I have not been able to find that. So please forward a link. If it is not in the User Agreement then all Apple has to do is say "Oops" our bad.
 
  • Like
Reactions: carswell
Again, how do you know that the Google Software does not share any data, even when it is running on Apple hardware, other than Apple marketing (propaganda)?
Because all these environments are ringfenced, based on a Zero Trust paradigm. The entire functioning of the modern internet and cloud infrastructure depends on data security and traceability. If Google software is running on Apple servers, the only way that data could be transmitted from those servers to, say, Google, would be if Apple explicitly allowed that to happen. And Apple will not do that because security is one of its main selling points. Far too important to compromise on without informing the consumer (and that's not marketing, it's simple business sense).
 
That’s a very odd take — they are a publicly traded company that has to disclose information that isn’t a trade secret. Apple also routinely posts articles describing they privacy transparency on their Security Research blog — again, for transparency — and they very specifically mention that Private Cloud Compute not only uses stateless computation (so the servers — never mind Google — have no personal data to utilize until the request comes back to your personal device) but is also has verifiable transparency — making every production build of PCC publicly available for security research and scrutiny.

Again — as a publicity traded company they can’t lie about this — and it isn’t just marketing. Marketing is “What’s on your iPhone stays on your iPhone” which is nice and vague enough to get the “idea”, but conveniently without details. But when you delineate details about transparency — then don’t — a company like Apple can’t hide from the fall out. Look no further than the usage of recordings for training they got called out on — and that didn’t even have a layer of transparency — just a bunch of users that don’t read the details when they turned on a service.

You seem to operating on the fallacy that something is certainly true even if you can’t prove it. if you can't prove something, but there are facts (as stated above) and an oversight mechanism stating it is true — the onus is on you to show why your opinion is valid. And calling it “propaganda” is wild. I don’t like or support Google in most cases — but their seems to be a legitimate wall between Google, the company, and the foundational model they are making available to Apple that isn’t connected to their services. Just like the anonymous sessions in Siri with ChatGPT.

But hey, if you know better...
All I said is that you can't prove it and you still have not. In your example of recordings for training, all that happened is they said, "Oops, our bad". But the data was still used.

Sure public pressure can do a lot, but that still does not mean it is secure and your data is not shared unless it is in the Usage Agreement. Which it won't be, because it is too much of a liability.
 
But will push gmail actually work in Apple Mail again? I need that a lot more than I need Siri to read my emails for plane flights…
 
Again, how do you know that the Google Software does not share any data, even when it is running on Apple hardware, other than Apple marketing (propaganda)?

How about if a panel of respected computer experts carefully selected from academia and industry and vetted and supervised by a panel of international judges were allowed full access to the Google code that Apple was running on its servers and could go through it line by line to confirm that no data was being leaked out to Google before the system was allowed to go live (even if that took years)? Would that prove that the code was safe? Well no because who's to say that somehow every one of those experts and everyone on the overseeing panel hadn't been bribed to say that everything was OK when it wasn't.

That's obviously a pretty silly (extreme) example but I was trying to make the point that in reality no one can 100% prove anything if you're willing to ramp up your distrust to a high enough level. Once you reach a certain level of suspicion the only solution is to stop using computing devices at all.

One possible sensible answer to at least achieve a high level of assurance would be for Apple to use monitoring software to monitor network traffic going in and out of its data centres and look for any suspicious activity that might indicate transfer of data outwards to some external server. I would be pretty confident that such perimeter monitoring is happening anyway, at least on critical Apple systems. That for instance is what alerted LastPass some years ago that it had a data breach in its password manager infrastructure when some unusual network activity alerted it to the fact that someone was doing unauthorised downloads of large quantities of data. (Luckily all the passwords that the hackers downloaded were zero-knowledge encrypted. I was a customer at the time and didn't have any of my passwords compromised.)
 
If Apple wrote software to do that, the data obtained would remain encrypted to your private key. You can't just migrate data from one key to another. Encrypted data is encrypted for everyone. Apple doesn't have a universal key that can decrypt this data: that's the reason encryption is such a sticking point for certain governments around the world.
At some point, the health app uses your private key to decrypt the data so you can work with it. Why couldn’t Apple then copy this data while it is decrypted in RAM? The OS is basically “at the keyboard” with you. I mean, they write the app, right?
 
Again, how do you know that the Google Software does not share any data, even when it is running on Apple hardware, other than Apple marketing (propaganda)?
This is THE big concern. The complexity of AI as being implemented makes it easy to obscure the truth.
 
  • Like
Reactions: nt5672
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.