Apple faces AI talent turmoil as senior Siri researcher departs

AppleInsider

A senior researcher's exit and a near-mutiny among Apple Intelligence engineers expose Apple's struggle to stay competitive in the AI arms race.

Senior Siri researcher leaves Apple

Apple is facing mounting internal fractures over its AI strategy, losing one of its top researchers while scrambling to keep key teams on board. It increasingly looks like a crisis of confidence in Cupertino.

Tom Gunter, one of Apple's most senior large language model researchers, has left the company after eight years. Colleagues say his deep expertise is tough to replace, especially as rivals like Meta and OpenAI throw around multimillion-dollar pay packages to poach talent.

The news comes in a Monday report from Bloomberg that also alleges Apple plans to integrate rival AI models more deeply into Siri.

Apple narrowly averts MLX team exodus

Apple is no longer the most desirable shop in town for machine learning. If it can't match competitors' salaries, Apple risks losing valuable talent.

Additionally, the company must provide them with compelling and meaningful projects. Otherwise, it risks its machine learning team becoming hollowed out.

Apple nearly lost the entire team behind MLX, its open-source machine learning framework optimized for Apple Silicon. Those engineers reportedly threatened to quit, forcing the company to scramble with counteroffers to keep them.

The fact that it came to that brink suggests morale is shaky and confidence in leadership is eroding.

MLX isn't a throwaway side project, it's essential for Apple's strategy to get cutting-edge AI running efficiently on its chips. Losing that team would have been a disaster.

Apple averted catastrophe this time. But paying people to stay isn't the same as keeping them motivated and aligned with the company's mission.

Apple's AI strategy shows signs of drift

These staffing dramas are a symptom of a deeper strategic confusion. Apple is debating whether to keep investing in its own foundation models or outsource core AI features like Siri to Anthropic or OpenAI.

Internally, executives reportedly see their own models as inferior. That kind of language does not inspire confidence in the teams building them.

Siri has lagged competitors for years. Outsourcing to Anthropic or OpenAI might be the only way to catch up quickly, even if it undermines Apple's reputation for vertical integration.



Read on AppleInsider

charlesn

Nice reporting. You've captured well the conundrum Apple is facing without Chicken Little hysterics. It's behind in AI development and I don't doubt that its existing internal product is inferior to what's available from third parties. So... do you wait for your internal product to get up to speed or do you seek a better, faster third party solution, thus risking mutiny by the AI team you have in place? I can't really criticize Apple's current team because I don't know what kind of sh-tshow it might have inherited. 

discountopinion

The narrative does not address the root cause for why Apple is where it is at the moment. Siri was groundbreaking when it launched and it seems like some of the execs at Apple just dont get it. My money is on Craig being one of them given how the software side has been lagging behind the hardware side in evolution.

dewme

If there is truly a lack of confidence across the rank and file of talent within Apple, that is not a good sign at all and points to leadership deficiencies. I say that with a grain of salt because we aren't privy to what is really happening internally at Apple. I've said it before but I will say it again, the media and political scrutiny that's being piled on Apple and its leadership team over the past year or so cannot be brushed aside out of hand.

Nobody can allocate more that 100% of their attention at any time. If Apple or Tim Cook is having to fend off another tantrum occurring in the White House or percolating in the cesspool known as X, he's not doing something proactive to advance Apple's position with its stakeholders, including customers. I suspect that a large percentage of Apple's leadership team's bandwidth is being consumed by defensive battles. That's a highly corrosive position to be in and something has to give and some things are going too break. The last thing you want to be dented, or much worse broken, is confidence.

The signs are there, at least in my mind, that the Apple "machine" is not running at full speed or in the capacity that it is capable of. I personally feel that this year's WWDC was a little "light" both in terms of what was shown and what was projected on the roadmap. Waving the white flag on some aspects of Apple Intelligence and Siri delivery expectations took a little bit of air out of internal and external developer's confidence. The UI updates to the major platforms are certainly appreciated, but kind of expected within the scope of incremental improvements and refreshes. If renaming everything around the year, i.e., "26" was considered a BF deal, that is really grasping at straws to find anything significant to show. I don't think I got a single text with "Woo hoo, the next version of iOS will be iOS 26, let's go party." 

All I'm saying is that if Apple is tripping over its own feet trying to play offense and defense at the same time they need to bring in someone to share the burden and allow the most influential leaders to stay focused on keeping the teams happy and motivated to perform at their very best, while enjoying every minute of it, of course.

loquitur

There are other forces pushing/pulling Apple's approaches to AI rather than just an internal/external "make vs. buy"
decision, one side favoring Apple's "walled garden" using in-house IP, the other side ceding some control in the arena.

That is, AI is regarded as so recognizably important by the world, Apple will be forced to at least entertain an open
framework by rest-of-world.  E.g. just as Apple (and Microsoft) allow for multiple browsers and search engines, partly for
antitrust reasons, they will be required to do so for LLMs.

The EU can easily mandate room for multiple LLM providers, just as they are now pressuring for app stores, together
with how others do so for multiple payment-processing methods. China likely has preferred state-sanctioned providers,
and Apple will have to roll with that.

Personally, although I'm fine with Google Search morphing into Google AI for Q&A search, I use a paid ChatGPT
account for software development experiments, but I'm not allergic to others such as Anthropic's Claude
or open-source LLMs.

Yes, there's an immediate vacuum for better Siri functionality, but I don't care where it might come from for
English speakers such as I. Plus, there might be a special version just for France, so Apple doesn't shouldn't have
to be a sole provider of everything.

Lastly, the media makes out everything as a "winner-take-all" battle, but it's not.  Within the current frontier of AI, there
is room for everyone.

edited June 30

danielgomes

These cracks began showing a while ago. It’s been a gradual decline over there and Tim seems to be floundering. 

You need a strong visionary leader at a company like Apple. One who is able to make the right decisions. Ignoring Siri for almost a decade was a huge failure in leadership. Under Forstall or Jobs Siri would have outstripped Ai  before it began. 

I think Apple needs new leadership. The problem is finding a strong visionary leader. I don’t think Apple has any. 

Maybe entice Forstall back as CEO? I’m up for that!

discountopinion

danielgomes said:

These cracks began showing a while ago. It’s been a gradual decline over there and Tim seems to be floundering. 

You need a strong visionary leader at a company like Apple. One who is able to make the right decisions. Ignoring Siri for almost a decade was a huge failure in leadership. Under Forstall or Jobs Siri would have outstripped Ai  before it began. 

I think Apple needs new leadership. The problem is finding a strong visionary leader. I don’t think Apple has any. 

Maybe entice Forstall back as CEO? I’m up for that!

I have to defend Tim a bit as he has made Apple into a behemoth and premium brand with massive share gains. Apple may be a victim of less ideal recruitment over the last decade . Elon is the master of talent attraction and Apple can learn a lot from his companies to reignite the Steve spark.

acydtrip

It is an interesting peak into human psychology to watch criticism of the most profitable technology company in the world. Evidently Tim Cook doesn't know anything about running a company, Eddie Cue doesn't know anything about business, Craig Federighi doesn't know anything about software and John Giannandreas knows nothing about AI.
In the real world Microsoft is losing $20 for every AI subscription. Chat GPT operating cost is about $7 billion per year. its subscriber base generates ~ $3 billion.  Open AI survives on "funding " from Microsoft ($10 billion 2021), SoftBank ($30 billion 2025) and private equity ($10 billion 2025). Which is all good for Nvidia. (First bitcoin mining, now brute force large language model tensor crunching. The luck involved but that is for another day...)
If you ask chatgpt a question 10 times how many different answers do you get?  
Open up a tab and come back, I will wait...
Did it hallucinate? well lucky you. 
No LMM voice assistant is ready for prime time. Why? because voice assistants are used by normal people. Set an alarm. Set a timer. Send a text to "" I am running late. Be there in 20 minutes."
If you reading this blog, you are a tech fan. You live in a bubble. Most of the world doesn't care how the sausage is made. Just that it tastes good. 
Apple is held to a different standard. If google chatbot goes off the rail or generates a racist tirade, it might make the tech section of The NY Times but it will not be on the evening news. The press craves storylines and Googles Beta program hits a bump isn't news. Even "Google Product Fails' isn't news. (Google+, Google Reader, Google Glass, Stadia, Hangouts) the Google Graveyard is profound, a lot of throw it at the wall and see what sticks. Apple putting out GPT-4 level LLM would be crucified for the 15% error rate in the beta. It has go be as good but on superior guardrails. Guardrails so profound as to make it useless as a chatbot.  Case in point, I give you Image Playgrounds. Image generation aimed at children.
Could Apple put out a tool to compete with Midjourney?  Probably not. But If an Apple tool was used to generate something foul would there be a storyline? Apple failed. Look at what Apple intelligence generated. Lawsuits to follow.. It would be safer to point the user to a third party and let them take the heat.  
(TLDR: HERE'S THE IMPORTANT PART)
So I don't judge the absence of a Chatbot or Image Generation Software as an indication of the state of LLMs at Apple. If I were in the LLM group at Apple and felt my models were as good or better than others available but management would use them for liability reasons, I would be upset. If I also believed that the guard railing  management desired was impossible because of the nature of the technology in use, I would argue no other team is as handcuffed and probably quit and let Meta or X hire me a use my models to prove how good they are. You know, like the guy who quit to make a crypto wallet.
In the real world, LLMs are very good at summarization and patern recognition. Google notebookLM is amazing. I feed it a 800 page manual and ask it questions and it summarizes answers and gives annotation. It's great for research. I use it on my Mac.
But of course, Apple is doomed.

applezulu

Whatever the actual degree of internal strife at Apple, it’s important to remember the broader context. The AI that the rest of the industry is churning out at breakneck speed is also hugely flawed and not at all what the hype makes it out to be. Pay package bidding wars are an indication that everyone is floundering, not just Apple. 

In the end, we need Apple to produce technology that is good, not first. My biggest worry would be that the hype and personal greed has infected too many egos and is undermining Apple’s purposeful, deliberative pursuit of something better. Let’s hope that’s not the case. 

charlesn

loquitur said:

That is, AI is regarded as so recognizably important by the world, Apple will be forced to at least entertain an open
framework by rest-of-world.  E.g. just as Apple (and Microsoft) allow for multiple browsers and search engines, partly for
antitrust reasons, they will be required to do so for LLMs.

Apple allows for this already, no "requirement" necessary. But--to use your browser example--much in the way Apple markets its Safari browser for its privacy protection vs other browsers you could use with the various Apple OSes, I assume it will continue to market consumer privacy protection as a key differentiator of"Apple Intelligence" and whatever LLM it chooses to employ. 

anthogag

Top AI researchers are likely being poached left and right; they have to beat recruitment firms down with a stick. 

kellie

Apple has always excelled at and focused on hardware design.  Soft has always come second to hardware.  Look at most of the annual product announcements, starting with the Steve Jobs product reveals.  There’s always been a focus on the artistry of the design.  Cue Jonny I’ve.  Look at the adjectives that have been used over the years - thinnest, lightest, colorful, minimalist, etc.  Frequently the design focus even overran functionality.  A one button mouse, minimal ports and buttons, etc.  The software always played second fiddle to the physical design.  Look at Siri, an interesting first to market capability, but one that has had problems for years and whose shortcomings have not been significantly addressed.  Apple leadership has been too focused n living in and protecting their walled garden which has had significant revenue and profitability benefits.  But AI is a totally externally driven force.  And Apple has been too busy being proprietary and protective of their platforms.  Now that these massive, industry shifting AI forces are accelerating dramatically, Apple is stuck.  Things they should have been paying attention to years ago, were ignored or minimized.  Catching up is a bear.  Even if they acquired an AI platform, integrating it into the Apple ecosystem will take years.  So it’s a combination of culture, focus, leadership, proprietary systems, software capability and externalities that have put Apple in a very difficult spot. It’s gotten to the point where excellence in hardware design is not going to be the main driver of sales.  Software design, functionality and capability are going to be the factors consumers focus on before making a purchasing decision.  This period of time could very well turn out to be a major business school case study a few years down the road.