Head of Apple's AI Search Project Leaves to Join Meta

[Populus]

But you cannot do that, it takes LOTS of memory and power to run these AI models, LOTS. As I understand. It would be impossible to run them on a phone at present or even iPad to an extent. The tech to do that on a device that fits on your pocket just isn’t there.

I agree, it’s not yet there for smartphones or tablets. On the Mac however, if they make 24GB of RAM the minimum, maybe they could launch that local, on device Siri LLM.

And before you reply saying that 24GB is not enough, I know. A year ago I was researching about local LLMs for the M4 Mac mini, and the bare minimum was 48GB of that shared fast memory… better if it was 64GB or even the 128GB from the Mac Studio.

Yes, that’s an insane amount of unified memory. However, I still hope that Apple may find a way, maybe through some revolutionary quantization, maybe through some other type of trick, that allows us to have a much more powerful assistant running locally on an M5 Mac.

If Apple wants to differentiate itself from the competition, they must take a different, unexplored approach.

 

[Art Mark]

This does remind me of the race 25-30 years ago to be the dominant search engine. Google won that race.
Apple's not even in the running in this AI race. They started off level with the rest (leading, you could argue), but are now too far behind to catch up. They should give up at this point and just do what they historically do best: wait for a dominant leader to emerge then buy them.

What race exactly? The race to generate porn? The race to generate PowerPoint slides? How about Ai that actually keeps you private and does you know, actually useful stuff? Not just bad art and 3D movies no one cares about. You have bought into a hype cloud. There is no 'race' to anything since there's no finish line. AI companies, ALL OF THEM, lose billions of dollars. Not a ONE makes a profit. So what exactly is the race for?

 

[Art Mark]

There is no 'race'. Since there is no 'finish line'. unless it's which AI company loses the most money. Don't buy the hype from the tech. bros. Few of them ever make anything anyone wants or uses enough to pay what it actually costs.

 

[JaydenA]

How does Meta even make money?

Finding new ways to learn everything about you to maximise advertising dollars.

 

[trssho]

Apple bleeds talent. I don't understand if they are cheap, or not a joy to work for....

 

[urnotl33t]

Blame Tim Cook. Siri would likely be the AI industry leader, or at least one of the AI industry leaders, if Cook had not fired Apple’s most Steve Jobs-like visionary, Scott Forstall, in 2012. Forstall introduced Siri in 2011.

Yep. Scott would be SCREAMING every day that Siri was less than #1, but he’d also make sure it got done. He was no Steve, but he carried that same drive and passion. Firing him over Maps, which wasn’t his fault, yet this garbage for Siri is allowed to continue… focus isn’t where it needs to be.
Billions wasted on Project Titan for nothing, untold amounts on Vision Pro that was no one asked for.. yet Siri languished. A real shipping product left to rot.
Scott has to be so sad about that today. Heck, I’d be surprised if he even still uses iPhone and iPad anymore.

 

[urnotl33t]

It makes me think there must be some problem with management to have such a high turnover. Or they are too locked into some architecture or approach which is inferior. And the employees feel like it’s pointless to continue when the strategy is bad at the root. You need one person or a small group of people with a clear intelligent vision of the way forward….

But it sounds like they just keep flip-flopping around and don’t know what they’re doing.

I would also agree with another who said that Apple is moving too slow slowly on something like AI. Every time you turn around openAI has some new feature ready for the public! Specifically for AI it needs to be a fast and aggressive game.

I wager it’s the architecture. Original Siri was rebuilt shortly after its debut onto an Apache Mesos distributed infrastructure then. No doubt it’s been rebuilt twice or more since then.
Now they’re just stuck having to handle old school clients on iOS 12 or iOS 14-ish era devices while figuring out how to rebuild for new clients….
They likely can’t just “cut off” a certain swath of older client devices (well.. they could.. just like that FaceTime fiasco in iOS 6 or 7 era, blaming it on a web certificate they allegedly couldn’t renew…lying jerks).

So yeah.. technical debt of the sizes we couldn’t even comprehend.

 

[urnotl33t]

absolutely. if this predicted bubble bursts then Apple might actually benefit most.

the current situation is that if you don't have AI then you're behind. Yes, Apple has AI but it sucks and doesn't match the likes of Google/Microsoft.

If i could predict the future I'd be a billionaire.

You just might be on point here. This may be their “exit strategy”. Bubble pops, firesale prices for many of them.. just gotta move fast before Facebook does.

 

[pratikindia]

AI is good for searches and helping to write a document. However, AI will really take off when it has been given personal context. There is no future scenario where I will provide Meta access to any of my personal information. Meta can develop all the gadgets in the world, but I would never use them.

May be you won’t. But billions of people in the world will. That’s a huge win for Meta. In many areas such as FB, IG and WhatsApp there is no competition for Meta and it’s sad.

 

[urnotl33t]

How does Meta even make money?

You.

 

[urnotl33t]

Unpopular opinion: Apple is ahead in AI. They are on the right path to integrating AI with user-friendly applications and delivering real user value. Best, not first. They have a deep AI research bench as shown by their significant body of published research combined with a design-led culture to actually make the tech useful. Unrivaled privacy stance essential to earn trust with sensitive health and personal data. Meta has no vision and has time and again recruited top technical talent to deliver nothing of substance. No one else has shipped anything like the personalized Siri assistant that was delayed, and Siri is falsely equated with Apple Intelligence as a whole. This is yet another version of "Apple can't do cloud"/"Apple can't do services"/"Apple can't compete with Android" narratives and a bunch of other narratives created by those envious of Apple's success and press desperate to report the end of Apple's winning streak. Gurman has conflicts of interest and relentlessly pushes rumors about Apple's alleged AI failures, while meanwhile Apple Intelligence is gathering steam. Just like Maps. Just like iCloud. In 5 years, the critics will be silent and Apple will remain a leader in tech with smartly integrated AI features leveraging the personal data, on-device computing power and leading spatial computing capabilities that only they have. Disagree all you want - I will be proven right.

I like your version of the future. I really really hope you’re proven right. Right now, that future is mighty dim.

 

[rp2011]

Mak Gurman says beyond Zuck hating Apple, they are willing to pay far more.

 

[haplessdreamer]

This looks like another example of what happens when the people doing the real technical work — often Asian engineers and researchers — realize their value and stop settling for being under-recognized and underpaid.


For years, Apple’s AI division (and much of Silicon Valley) has relied on brilliant immigrant and Asian talent to build the foundations of their products, while senior credit and paychecks often went to white executives who managed from above. Now that those same engineers are being recruited aggressively by Meta, OpenAI, and other AI-first companies that value their skills, Apple’s leadership is left scrambling.


Siri has been stagnating for years not because of a lack of intelligence on the engineering floor, but because of misplaced priorities and top-heavy management. You can’t keep losing your top builders and expect innovation to survive.


Talent always finds its level. The global AI race is showing that real technical expertise — not office politics — drives progress. Apple may soon learn that lesson the hard way.

 

[haplessdreamer]

Odd you’d want to poach folks who clearly aren’t up to the task. Plus since when do senior execs matter.

Senior execs claim all the credit, but do none of the actual work. These brilliant Asian and Russian ML researchers know their worth.

 

[haplessdreamer]

Exactly right. Some of the comments here are clearly misdirected. Siri has been around for 14 years now. That it is still so routinely useless and failure prone is a reflection of the priorities and corporate culture set by those at the top, especially the CEO. I know if I was given a task to fix long-standing issues but found I would not have the actual authority to fix it, I would start looking for another job. Responsibility without authority = scapegoat.

Yup, their entire white management team is bloated. ALl the hard working Asians doing the real work leaving where they're appreciated.

 

[JoeG4]

I know I know this isn't great business-wise, but hear me out here:

I'd rather they leave the general LLM stuff to companies that specialize in it. If Apple's teams think up good applied reasons for using that tech, great! Make those happen!

I'd much much rather Apple focused on making great hardware and the future of computing standards. Apple doesn't need to be a one stop shop for everything. There are things others do well enough, like image editing and Wi-Fi routers.

 

[svish]

Looks like the entire team behind Apple Intelligence is changing. Don't know whether this will lead to delay in Apple's plans.

 

[SpanishAppleNerd]

That might just lead to key Perplexity employees leaving to greener pastures.

Hahahahah true. Probably better for Perplexity to remain independent

 

[iMac The Knife]

With the current state of Apple AI, I would be jumping ship too!