Apple shuffles AI executive ranks in bid to turn around Siri

John Browett, CEO of Dixons. Anyone from UK have said WTF at the time.

Angela Ahrendts, CEO of Burberry, the two together Apple has literally stopped expanding their Retail store despite having 10x as much customers. Rolled out many "changes" to Apple store when nearly every single one of them were walked back to Steve Jobs era. Mainly by Deirdre O’Brien who has been with Apple for 20+ years.

Changes of Direction in PR. Which leads to Katie Cotton leaving.

Forced out Scott Forstall. Tim is suppose to be the mediator.

Promote Craig Federighi, after all these years I am still not convinced he is the right person for the job. Especially after merging iOS and macOS team together.

Putting too much trust on Eddy Cue and sidelined Phil Schiller. Apple TV+, Apple Music, Apple Fitness, Apple Books, News, All these services are Eddy Cue. The committee that ultimately ruled for Apple's 30% cut? Tim Cook and Eddy Cue.

Probably a few more names missing.

So many folks will tell you “I learned so much from my previous failure and without that I never would’ve succeeded blah blah blah.” I personally believe in that sentiment a bit, but just wondering if anyone’s aware of a well-known case where that’s happened.

Well, apparently some people have at least some expectations, after Apple pushed ads touting Siri's new abilities (and Apple AI) but failed to launch them on time, and now Apple been hit with a federal lawsuit against them: https://www.axios.com/2025/03/20/apple-suit-false-advertisin...

Like you say, Siri is borderline completely useless. Just timers for me.

Even just the most basic LLM integration, without even giving that LLM any personal data would be a massive improvement. They even have this ChatGPT integration that Siri seems totally incapable of using in any meaningful way.

All things considered Apple actually delivers some great software, Apple is one of the few companies that built their own operating systems, their software UX is considered to be the gold standard. They pulled architecture transition processes that required great software engineering alongside with the hardware engineering and they did it amazingly well.

Apple's success in some areas suggests that the failure in other areas doesn't have to be about software development culture but simply wrong bets or policies.

Apple tomorrow can become the best platform for anything AI, they just have to open up their platform and let people on the cutting edge of AI to take over some functions.

IMHO Apple's greed is failing them. That revenue growth in the "Services" category on their quarterly reports is their Achilles heel. Locking down users on their platforms allows them access to a "market without competition" but there's actually a competition happening outside of this walled garden and that competition is fierce. People in the garden start thinking that maybe the grass is greener beyond the fence.

Ironically, there's a possibility EU saving them from themselves if they stop being just compliant enough to hold up in court and embrace the spirit of what DMA is trying to do. Throw a tantrum on EU and say that the lost revenue was because of the evil EU thingy and save this great brands and great company.

The concrete value-measure for me is whether I would feel any loss as a user if Siri simply disappeared. The answer is no, I would feel no loss.

But this is also true for me if the subject is Gemini, CoPilot, Alexa, &c. For all the money and energy burned, for all the advertising babble and grotesque valuation, I would not as a user be one jot less productive if the invisible Thanos hand of the market made them all disappear at the end of this sentence.

But when Apple advertises literally all their new devices as "Built for Apple Intelligence" without actually delivering on that promise, that's just false advertising.

We're on Sequoia 15.3.2 now and the keyboard shortcuts of the big new window tiling feature still don't work on non-Apple keyboards (or Apple keyboards, if you happen to have remapped the globe key to something else, like Caps Lock).

† Hyperbole, in case it wasn't obvious

His delivery is too perfect and makes him sound like a really good text to speech program. It’s off-putting. The only other person I’ve heard that has this presentation style is David Miscavige. Perhaps some analogies could be made…

I would contest this statement very vigorously. Siri has been a disappointment but nearly every single Apple product has a load-bearing reliance on an AI/ML feature.

I'm continually frustrated by the (relatively recent, since LLMs) tic where just because a ML model isn't conversing with you or otherwise making its presence impossible to ignore, that it isn't delivering a massive amount of value. It is frustrating from laypeople, it is extra frustrating from industry insiders.

Just a few examples off the top of my head:

- AirPods with ANC and spatial audio. Both headline features of these products that are 100% AI/ML.

- Watch with heart arrhythmia detection, automatic workout detection, fall detection, etc. All are headline features (some literally life-saving) and are 100% AI/ML.

- iPhones have high-profile features that are 100% AI/ML: automatic car crash detection. Others are more subtle but IMO substantial differentiators - such as automatic image enhancements out-of-camera.

Again, I know in the age of ChatGPT we seem to have twisted ideas of what ML is, but "AI/ML" is not synonymous with LLM.

working in AI/ML, is there any recommended literature one may read on how to ensure efforts result in products? Would be interested to access more experience to benefit both an engineer and a manager pathway.

It has only become worse in the past couple of years.

The inability to know that bedroom means bedroom [1], or the impossible task of adding several items to a list [2], or failing to answer what the current month is [3] have nothing to do with slow progress or UI.

[1] https://x.com/dmitriid/status/1781030511336456612

[2] https://mastodon.nu/@dmitriid/114172857097176113 and https://mastodon.nu/@dmitriid/114180042031267746

[3] https://news.ycombinator.com/item?id=43432945

https://en.m.wikipedia.org/wiki/Moravec%27s_paradox

The futuristic ads were made for the consumer to feel safe and sound.

The actual money flowing went into military applications (drones, guidance systems, etc) and commercial applications (car factories)

For the above systems, they don't have to be physically capable, just have a restricted set of actions / outcomes, for example, flying a missile. But, they need to be mentally strong, to take decisions at split seconds.

In my experience it also has more accurate lane guidance and, as you say, better delivery of audible driving directions ("go through these lights then, at the roundabout, take the first exit"). Google Maps also has an annoying habit of choosing the quickest route at all costs which, in certain parts of the country, involves going miles down single-track country lanes (which legally have a speed limit of 60 but more than 30 would put you through a hedge or into an oncoming tractor) when a dual carriageway A-road would be only 2-3 minutes slower. I can't say Apple Maps doesn't do this, but anecdotally it seems to do it less.

However the business info is still way behind Google, particularly regarding things like temporary closures.

If I search for pizza, both will give me results for restaurants that serve pizza. But if I zoom in on Google Maps, I get other, unrelated restaurants like big chains and a donut shop.

That's not to say Apple Maps is better overall for searching for businesses - they often have incorrect hours, or have closed completely.

Nonetheless, your comment convinces me that I should try Apple Maps more seriously. Maybe I’m missing out.

It has been more than a decade of them kicking the can down the road, so I don't expect Apple to provide a good user experience with Siri any time soon.

An example I ran into was looking for a specific IKEA near Oslo. There are two. The one that's closest to me is a place called Furuset.

Google: https://i.imgur.com/rHu9XOO.png

Apple: https://i.imgur.com/5HaiGny.png

Google has it as the first hit, but Apple doesn't even show it. None of the hits match, they're all either searches (which... what does that mean? If I do a search and one of the hits is "search nearby", why does it want me to another search?) or some kind of IKEA office, not the warehouse.

At the time I didn't know the address, so I couldn't infer the right one from the list. And I was in a car. I tried to zoom in near where I knew it was:

https://i.imgur.com/FrXeyow.png

...But there aren't 10 IKEAs there. Just one.

Apple Maps keeps doing this. I can trust it 95% of the time, but the last 5% is a garbage heap of wrong or missing addresses, or outright wrong directions.

Before you say this is a Norway thing, I've had the same thing elsewhere in Europe, and it once gave me some very wrong turns in New York that put me in the middle of nowhere, in the middle of night.

I was one of the defenders of Maps when it first came out, but every time I try to commit to using it, it fails in some way that drives me back to Google. Google's UX is horrific (when you've searched, selected an address, and asked for directions, why does it take five "X" taps to get back to the main map view?), but it almost never gives me the wrong thing.

Just tried this on Android, and it works (though it gives the full date)

They just need to start giving a shit. Care about bugs for once. The whole software org needs to take the time to go through their (clearly) enormous backlog of bugs, and shore up what they have.

The problem isn’t the architecture or vision or rethinking their stack, it’s that what they have is barely holding together under its own weight. They need to do the unsexy work of going through and shoring it up.

As for Siri, just give the job to OpenAI. Let them write the new thing, they’ll probably be done in a month.

Crazy, given that they are so good at marketing, that they haven't thought about retiring the name Siri, a name which now, has connotations of being an inept and unskilled assistant.

Maybe for the next one they need to do a 180, and give the AI a name seems like she would be on top of everything and runs a tight ship and conveys that old school cold, calculating, no nonsense image.

Like Frau Farbissina or Geraldine or Gertrude.

I would trust a Geraldine with my books.

Kinda like how they use Afred or Jeeves for butlers.

There is always room for improvement, but I never got the feeling they need to drastically change things or start fresh. They just need to squash some of the big bugs IMO.

Systems engineers would have responded by informing this hypothetical Steve Jobs that if he wanted to make the phone smaller by 20%, the phone would lose 10% processing power and lose 30% battery capacity, 9 months and $50m would need to be expended to retool the factories, and 12 months and $40m would need to be expended to redesign software UI for a smaller screen. And for an accurate answer, those systems engineers would have needed to have been able to consult with materials scientists, software engineers, engineers designing factory tools, and dozens of other different types of disciplines.

Obviously it didn't quite work out that good (although I guess in my personal use cases, it has). But the point of telling me the story anyways, which I think matches here, is he had a way of pattern matching the average persons expectations based in reality, and applying it to their technology, and then aligning people towards ambitious goals that were on some level based in reality (i.e. people use TVs, they are similar technology, people don't expect them to break). I don't know how much credit is him or others and all. But I know from these kinds of stories, and from my own experience elsewhere, that some people just regularly and instantly "get it", and Steve Jobs seemed to be one of those people, and to also have the station to force others to go along.

Here I think its pretty obvious Apple is both the worst performer in AI, and also the best positioned to capitalize on it. In that sense their current state is clearly demonstrative of some major failure. Its hard to imagine someone like Jobs being in the same position in today's rate. LLM's are phenomenal technology and their use cases are vast. And we've all got these devices ready to take advantage of them, ready to listen, to respond, to coordinate, etc. And here i am asking Siri to turn down the volume and it either turns it up, gets confused, or misinterprets and goes on some tangent. There's a wildly large gap in what Siri does now, and what we 100% know it could do based on the existing tech many of us are using every day.

This is my imaginary scenario -- if Steve Jobs were to tell me in a meeting, "I see air bubbles coming out of the phone", I (as an engineer) would reply to him, "Dude, that air is carrying excess heat out of the phone via convection. If you don't leave any space in between hot electronic components, your phone will be a toaster or a clothes iron".

"If Steve Jobs still ran Apple"

The engineers know that the phone isn't literally a solid brick of electronics and "can't be made smaller" is normal human speak for "can't be made smaller without time and money we don't have to do it." The smart thing would have been just saying he wanted the iPhone to be smaller, money and resources is no object and we're willing to dedicate time/people to this project to get proposals put together.

Jobs was very good at his… uhh job but stuff like this isn't an example of it.

Starting a specific album at a specific song, for example... well Siri handles it, to a point, but it has to interface with Spotify, and there you hit a wall of hot garbage, along with the sequencing and process-management/integrity issues of integration software. Garbage meets a money fire, in other words. High cost, brittle, constantly changing, glue code that is mostly depending on third parties also requires session management? Not a winning product formula.

Sprinkle in some inhuman HCI nightmares like not pausing speech appropriately when reading an album title and you've got a recipe for an expensive turd that annoys your customers, oh my, by a lot.

I'm not convinced it was a technical limitation, even stuff like voice parsing was not improved or basic functionality expanded horizontally.

Summarise a meeting here. Summarise the report there. Summarise the email from marketing. Schedule a meeting with Jim. Turn the lights on in the kitchen for the housekeeper.

While yes, there is some usefulness in having something summarise things for you, most people in their day to day lives don't handle a "100 emails from everyone" and "30 meetings with 40 different department heads" or whatever these people do.

Quite a big gap from the market they are targeting with Siri.

"Well, um, sir, we've realigned our strategy around generative AI and..."

"No no no, tell me what you've DONE. In the past six months. It shouldn't be hard."

Now consider the Billions they've spent not only integrating, licensing, but also in opportunity cost towards other projects they could be working on.

I wonder if it will meaningly hurt Apple's reputation to the general public... they were doing so good with the m1 chip lines and to me it is fairly tarnished after chasing leprechauns for two years.

Many of the most interesting features (like LLM Siri) were ultimately cut from ios 18. Maybe next year.

For technical people, it's a privacy nightmare because it doesn't really work as advertised on the tin. The only way they can get it semi working is by pushing the request to the mothership and do processing there rather than locally on device which means privacy is an option. Also, be technically minded, some people developing these systems make more money on data harvesting than any other service, so local only prevents them from that precious data. So the local processing isn't just a technical problem, but one that interferes with their core business. Of course, none of that is advertised on the tin

I'm excited for it, and like Apple's high level privacy protecting narrative with the secure cluster stuff. We'll see if they're ever able to deliver.

Has the Pixel Gemini assistant gotten to that level? The Google Assistant was always better, and I know the switch to Gemini was a little rocky at first, but in theory it should be able to do that stuff, without all the privacy verbiage.

Anything more ingrained or trying to be too clever or invasive, not at all. Microsoft has a funny example here, as what I describe is the consumer Copilot app. The Microsoft 365 Copilot experience seems the worst extreme, muddling the product, injecting itself across domains for dubious benefit. The rating in the app store of both of these 2 apps reflects this divergence.

On mobile, I could see the potential as a substitute for search engines, with pros and cons for it's style of information retrieval. Beyond that it's stretching the use case of LLMS

> Hey Siri, please play Emerald Rush, by Jon Hopkins on Youtube Music.

> Sorry, you'll need to continue in Youtube Music.

Then yes - I want AI on my phone

I could be wrong of course, but my intuition is that only that tiny fraction of the population already following LLMs would potentially not want AI Stuff on their phones.

If Apple made a voice mode as advance as ChatGPT, and no one used it, then you can throw away Siri all together (and i'd assume Apple has the data that shows people actually use the Siri)

Whether the fault lies with individuals or the structure of the organisation is a more interesting question. I wonder if American corporations would take a more individualistic approach to this compared to their European and Asian peers.

Voice assistance was the hyped thing sometime before block chain and LLMs. Everyone needed one or was working on one. So Siri got stuck into the iPhone, like Apple Intelligence is now, because "the market" needed Apple to do so to keep stock prices going up. Then the hype died out Apple sort of needed to keep Siri around, because it is one of the interface to the HomePod and CarPlay. There's no reason do develop Siri beyond that, because very few people use it, not having a better version isn't going to sell less iPhone and they can't mine the data, because they sold everyone on Apple being privacy friendly.

There's absolutely no business case in having Siri do anything beyond "Call my wife", "Next song", "Play U2".

In the age of LLMs I also increasingly view my experience with "Hey Google" and "Alexa... " with similar contempt. Their output feels so bad compared to what ChatGPT's voice gets me. All of the big tech companies seem to be struggling immensely to deploy the latest in AI advancements.

- Their promise isn't as ambitious as Apples. It's basically Gemini + pre-existing Assistant features as plugins. Not trying to be the privacy-touching agent at least at the first cut. (Or they promised big, but they have trained the audience not take it as a fade value haha.)

- They ship the Gemini version as an opt-in for the Assistant replacement so that they can learn "iterate" on the adventurous enthusiast while leaving larger user base as is.

The main issue to Apple is the different mindset that takes to deliver hardware vs. services. Delivering hardware you have one shot to "wow" your customers, and Apple keep nailing this process in a way that the whole industry can't keep up with anymore. On software, that doesn't work that well. You can totally be disrupted by a startup that delivers an MVP faster with 30% of your perfect product that would take 3 years to deliver. In this way, Google's assistant strategy in the short term was spot-on compared to Apple's. You take the hits of bad feedback in the short term in the hope that being in the game can create enough momentum to get you to a solid position of features and know-how in a couple of years.

Me: Hey Sir, set a timer for my management meeting in 10 minutes.

Siri: I've deleted your minutes from the last 10 management meetings.

Kind of like how Meta keeps the disaster that is Horizon Worlds going, but much worse because Apple has a monopoly on AI integrations in iOS, and nobody else is allowed to compete with their terrible implementation.

I think many inside and outside Apple hoped that the ways that they scoped their features in the Apple Intelligence announcement would help them pull off something duly reliable and practical, but it's not that surprising that even those ambitions might have bought too deeply into the hype.

Depends what you mean. Building an LLM seems to be easy enough that several companies have built one, and even Twitter was able to magic one out of the ether from pretty much a standing start given the resources.

Getting an LLM to do the right thing for you 99.999% of the time? Has anyone managed that yet? Frontier models still hallucinate the most basic shit, and consistency is very, very difficult.

Meta, OpenAI, and others out-compensate Apple for ML talent by a factor of 2-4x.

Are you serious? Did you miss the headlines around their news and text message summary products?

I don't know how it's implemented, but Apple Intelligence's proxying to ChatGPT ends up being far worse than using ChatGPT directly. You could ask the same question to ChatGPT and Siri powered by Apple Intelligence proxied by ChatGPT, and the former gives the right answer while the latter gives the wrong answer.

> Regular people don’t expect new this and new that from a device

But Apple advertised iPhone 16 specifically along with Apple Intelligence. They didn't market the new generation of iPhone as iterative improvements; they marketed it as specifically for Apple Intelligence.

Can anybody square that circle?

It would take very little effort to put an LLM frontend on top of this, and yet they've not only abandoned applescript (or the underlying apple events) at a time when they could expose it to the masses, but have gone in the exact opposite direction with "Shortcuts".

Oh and the icing on the cake is that apple events can be sent over the network as well, and this infrastructure has existed since the early days of OSX.

They have an advantage when it comes to running them locally, but it feels like they're trying to push it onto consumer hardware before consumer hardware is at the point of actually being able to run useful LLMs.

It surely is just your opinion. Nvidia is king and Apple has found a way to market an integrated GPU and CPU RAM as something magnificent, rather than something that has existed since the dawn of computing.

There is a reason Nvidia is king. There is a reason corporations buy Nvidia and not Apple for their LLM uses.

And Siri will continue to be dumb regardless of whatever "AI" monicker they slap on it.

That's a page right out of the Windows Update playbook