Apple brings OpenAI's GPT-5 to iOS and macOS
arstechnica.com
arstechnica.com
I don't understand why their best strategy isn't to just stand and watch, and then ruthlessly integrate the winner with their product line.
This is the same reason Apple ensures that apps from its App Store work better than websites. They need to keep their own platform essential.
This is exactly why the "Apple Intelligence" branding was so crucial. The plan was to make the AI feel like a unique Apple feature, justifying the hardware purchase. While the launch may have tarnished that brand, expect Apple to lean heavily on it going forward.
The challenge is that any partnership with ChatGPT or Anthropic comes with a branding conflict. OpenAI will want credit, but a "Powered by ChatGPT" label is the last thing Apple wants. How Apple navigates this dilemma will be fascinating, as they are truly in a bind.
There's a lot of reason to have and use a model, OpenAI or otherwise.
In terms of humanity, Claude is the best and probably the best fit for Apple audiences. It’s the emotionally friend who is level headed. The NYTimes did an article on Claude which captures this sentiment.
ChatGPT is like the exuberant nerd who wants to talk shop.
Gemini is a the brainy but less emotionally intelligent friend.
It would be hard to imagine your Messages history showing up in Google search results, but ChatGPT just went through that same scenario with their user chats. Also, it used to be common (though I'm not sure how much it has happened recently) to sign into ChatGPT and see other people's chat history in your session. IMHO ChatGPT has not earned any trust.
iOS, the iPhone operating system, which allows
users to make certain queries via GPT-4o
I have not seen a glimpse of it yet (I think it asked me once to enable it and I denied?). And when I ask friends about it, they also do not use any AI provided by Apple.
If anyone here is using it - how and what for? Do you like it?
This tends to give far better answers than Siri
They might, but not actually know. Does iOS (poorly) summarising notifications on the screen count as using AI? I can't remember what the default setting is in iOS 16, but in the new iOS this year you get prompted for this during OOB setup.
On device speech to text.
On device language translation.
And then realized I was not getting that space back :-D
I have used the built in editor in their photos app a few times to remove an object from a photo (which works very well) which they label as "AI" but I swear existed before any of this.
As for their other stuff, I use Apple's email inbox summary feature constantly, and I occasionally use the writing tools AI stuff just to proofread long text I've written to make sure there aren't any glaring mistakes I'm missing.
In addition, i know a lot of apple device owners dislike AI integration. For example, my non tech friends turned off a lot of the apple intelligence stuff in the recent big iOS update.
How / where do you make these queries?
https://support.apple.com/guide/iphone/use-chatgpt-with-appl...
It's optional, and the user's permission to send a request to ChatGPT is required every time.
> In most cases, LLM-related features built into iOS and macOS use Apple's own models, which live under the Apple Intelligence branding umbrella. But it gives users the choice of referring a prompt to ChatGPT on a case-by-case basis when the prompt is outside the scope of what Apple's models are designed for.
I find when opening a new voice chat it talks entirely too much.
Any tips or are you just far more patient than me?
This strategy makes a lot of sense if you believe that progress will plateau at the current level with small incremental improvements in reliability and UX. Apple really only pays a price if failing to develop their own foundation model results in their phones not selling, which doesn't seem to be the case, regardless of how much of an embarrassment Siri is. Right now, Microsoft, OpenAI, Meta, Google, and Grok are all bludgeoning each other with trillions of dollars and creating a perfect situation for Apple to swoop in in a couple years when reliability is sorted out and GPU compute has crashed to rock-bottom prices.
"Providing tokens" is the bottom layer of the value chain. If you can't build a moat around it, it's a race to the bottom because you can only compete on price.
Apple doesn't compete on the bottom of the value chain.
Neither of these seem like a shame so much as leveraging SOTA.
That doesn’t mean that Apple shouldn’t be following the space closely but I think it’s a mistake to think that normal people are changing their buying decisions based on any of this yet. One trade off here is that Google is spending a ton of money but as of yet is massively in the red on that investment, which means they’ll stick with their own LLM while Apple is free to switch to whoever the market leader is without a large sunk cost factoring into the decision.
I am very very far from a google fanboi.
My perception is that a huge percentage of the mass market just like OpenAI because they were the first to market and still have the most name recognition. Even my coworker who works in DevOps says “Gemini sucks, Claude sucks” even though he has never once tried either of them and has never looked at a single benchmark comparison.
Google must keep up with others because it is ad company. And oh boy the potential of influence with AI models.
Until you can't!
Disclaimer: I am speculating here.
Apple may not be the worst offender, but they do from time to time, introduce things that have to be pulled back later.
Will the "Use Apple AI" switch go away? We just don't know, and there are aspects of MacOS that I detest, like search always starting in "This Mac" instead of the working directory, that I have to change almost every time, and search beginning as soon as you type in a search term, (enter the beach ball) rather than waiting a few letters, while slowing down your typing. These may never go away.
And there are lots of other examples on mac forums.
What or who determines what goes away and what stays, begging a fix from some app vendor?
I doubt very much apple had any say over the personality of GPT-5. And if it did, it’d be in the prompt it sends over to ChatGPT - not in the training and reinforcement part.
Yes, Apple is using another model right now, but, they desperately need an AI win. Thus, they likely want a “latest and greatest model” and they have the leverage to influence what goes into the model. Due to Apple’s previous AI hiccups, performance, safety and guardrails are prioritized, thus a Siri flavored GPT-5.