Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Kill Siri in about a year, and call the next upgaded version Steve

IMG_1225.jpeg


Steve.
 
  • Like
Reactions: jozeppy26
Of course it is.

Whilst OpenAI are content to burn venture capital just to keep their lights on, Apple doesn’t do anything unless it can underwrite it with a future profit.

The big fruit has more than enough liquid R&D to build its own LLM now the pressure from shareholders to just have anything at all is off.

In two years time we’ll look back on the empty promise of 2024 as a small blip.
 
Apple either unconsciously recognised that and shied away from it, or they knew full well what it would mean and sought to hold back the tide.
This, according the Bloomberg article,

"Rockwell reports to [Craig] Federighi, who’s taken on more responsibility for Apple’s AI software-related product road map…

Federighi, Apple’s software chief, remained reluctant to make large investments in AI — he didn’t see it as a core capability for personal computers or mobile devices and didn’t want to siphon resources away from developing annual updates to the iPhone, Mac and iPad operating systems, according to several colleagues. “Craig is just not the kind of guy who says, ‘Hey, we need to do this big thing that will require big budgets and a ton more people,’” a longtime Apple executive says…

Giannandrea’s product managers have been moved under Federighi’s oversight, while Rockwell has revamped the Siri management team by putting his top lieutenants from the headset project in charge. Walker, who’d been running Siri under Giannandrea, lost most of his engineers and was assigned to a new project.
"

 
The current version of Siri already supports "extensions", which is how ChatGPT integration works, at least on macOS. So they've already enabled your idea to an extent.
Those extensions don't allow any assistant other than Siri to be integrated as tightly and broadly with any Apple OS as Siri was supposed to be able to do, getting data from and controlling apps and settings, looking at the contents of the screen ("contextual awareness") to give you feedback on it, etc. The integration of ChatGPT into Apple OSs is merely enough to allow Siri to hand off questions it can't handle, which is most of them.
 
Last edited:
  • Like
Reactions: arkitect
We’re calling it Apple Intelligencer, and the iPhone 17 was built for it.

*Features available in beta.
**Coming late 2026 via software update.
 
A lot of the "Apple cannot make a modem" and "Apple cannot make a GPU" vibe here...
It's easy to make these sorts of strong claims when you know nothing about what's involved in making an LLM (or modem, or GPU) and nothing about what's going on inside Apple.

Meanwhile in the real world we have things like this
which has led to things like Mamba and similar linear or sub-linear attention schemes or attention replacements.


Apple looks like it's behind for the same reason they looked like they were behind in modem – they have somewhat different goals from the mainstream, and they won't release until those goals are met.
In the case of the modem, the goals were not just to implement the spec (like any other modem) but to hit certain power levels.
In the case of the LLM, the additional goals appear to include
- lower power (always ...)
- the usual security/privacy stuff
- a deep set of APIs that both expose the LLM and give the LLM agentic power

You can see this if you track the Apple papers. It's not that they are behind others, it's that they're continually looking at different types of things from others. For example, if you quantize an LLM, sure performance goes down a little acrossa range of tasks; that's expected. But exactly WHERE does the performance go down? What's the TYPE of functionality that's hurt most by quantization? That's the sort of question Apple is investigating and that I don't see anyone else investigating.
Well, even as an Apple user for the last 35 years and proud owner of a M3 Pro equipped MacBook, to be honest Apple couldn’t or didn’t want to make either a modem or a GPU from scratch. That’s why they paid 1 billion dollars for Intel’s modem division and also still to this day use GPU tech designed for PowerVR GPUs from the Sega Dreamcast era, now owned by Imagination Technologies.

Now that said…I think there is a kernel of truth to Apple taking the cautious and pragmatic route to AI and making sure that Apple AI actually enhances existing Apple tech, software and user experience. This is a big reason why they are objectively behind other firms like OpenAI, Anthropic and DeepSeek. They got burned bad with their rollout trying to play catch up and the world knows enough now about LLMs and see too much chaos and hallucinations to trust existing AI. If Large Language models suck…it’s doubly difficult to make Small Language Models that can fit locally in memory on a phone that doesn’t suck worse. That’s a big haul even for Apple and it shows.
 
  • Like
Reactions: 01cowherd
Do we even actually want a Siri that is like ChatGPT?

What I want is a better Siri, not another blithering chat bot.
 
That a company would aim for its next-gen product to be just "on par" with what the competition already offers is quite telling.
I can't believe this shameful saga would have happened under Steve Jobs.
 
CTFU. On par with... That's great. The reality distortion field in Cupertino is wild. I don't think in 6 months Apple has managed to leap ahead of companies that have been developing their AI for 20 times longer.
Yes, and connecting Siri online would make great strides in making her not totally stupid. Hell, maybe she can learn grammar and dictated a short sentence without 20 errors then too.
 
The race is on and Apple is still thinking
Maybe Apple needs to buy ChatGpt
Bringing Siri into the competition will take years, if it can be done

Apple buying ChatGPT is hilarious. Their dedication to "privacy" would never work there. If they applied it, ChatGPT would become dumb, like Siri.

It's the privacy thing that kills Apple and why their falling further and further behind in everything they do. You can't do things on device with current tech and get good results. This hits everything from bad song recommendations in Apple Music to Siri not able to transcribe well. Everyone else is cloud processing, storing and learning voice in the cloud, etc. and sending back encrypted info to the device. Apple doesn't feel encrypted cloud connections are.... private?
 
  • Wow
Reactions: gusmula
"As good as" is not supposed to cut it at Apple. Apple is traditionally late to the party but with a better version or implementation of a technology.

It was like Disc Burning. They were the last major manufacturer to bring it to their products, but when they did it was cleanly integrated into iTunes and the Finder, and then iDVD for video. They took a technology that could be very complicated and empowered everyday people to be able to integrate it into their everyday life in a way that just works without a lot of effort.

Same with the iPod and digital music.
Same with the iPhone and iPad.
Same with the Watch.

Apple should be well positioned to figure out AI for the everyman, but as I am sure they are finding out, the current state of AI is that it you can't fundamentally rely on it to "Just Work" the same way all the time. LLMs output varies so much depending on the prompt. Whenever someone complains about the results they get from AI, the hype club response is: “you’re just using it wrong.”

The last time Apple tried a line like that, it got a nice slice of humble pie. You can’t tell people that something new is going to revolutionize, simplify and overall make their experience with technology better and then have the default response be “it’s not us, it’s you.” (AKA “You’re just holding it wrong.”)
 
Well let’s see how long it takes to arrive. Maybe the name Siri needs to go and a new name is producted to move on from the Siri history
 
An executives interpretation of a products quality is useless in the real world. People at that level are functionally braindead.
I’ll believe it when I see it, in at least 6 years time.
 
Well, even as an Apple user for the last 35 years and proud owner of a M3 Pro equipped MacBook, to be honest Apple couldn’t or didn’t want to make either a modem or a GPU from scratch. That’s why they paid 1 billion dollars for Intel’s modem division and also still to this day use GPU tech designed for PowerVR GPUs from the Sega Dreamcast era, now owned by Imagination Technologies.
No one can build a completely new modem from scratch due to existing foundational patents. Thats why they bought Intel’s modem division. Regarding the PowerGR GPU tech, so what; AS’ GPUs are modern, powerful, and efficient.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.