5 ms·

A few quick cliffs notes plus added info: John "JG" Giannandrea was hired over to Apple around 2018 to head Apple's new AI/ML division, which would include the

A few quick cliffs notes plus added info: John "JG" Giannandrea was hired over to Apple around 2018 to head Apple's new AI/ML division, which would include the Siri organization that predates Giannandrea's tenure at Apple. Giannandrea was previously the executive in charge of Google's Research division. Mike Rockwell was hired into Apple from Dolby Labs around 2015 to head a group called "TDG" that would be the R&D and the product group for what we know today as Apple Vision, an AR/VR headset. The same organization had its own AI/ML applied research teams focused on computer vision problems, and had hired in folks from Microsoft's Hololens team. From what the article describes, the Siri organization is being moved from Giannandrea's scope to sit under Rockwell, and Rockwell will be moving away from the Apple Vision organization. Apple hired Giannandrea to build an organization like the one he led at Google, and appears to have found that that the AI/ML organization built at Apple struggled to manifest its work into successful product wins despite the massive financial investment. It's worth noting that at Google, Giannandrea was succeeded by Jeff Dean, and the organization was more recently reorged to become the "Google Deepmind" we know today also had the same struggles.

worstdevna 2y ago
It feels like Apple had a dream of a granting easy AI integrations for many different apps and workflows, only to discover that very few people wanted any of those integrations from apple intelligence. At least, not the current iteration of apple intelligence.
- BuyMyBitcoins 2y ago
  I suspect the most significant thing holding Apple back in the realm of AI is the fact that Apple prides itself as a company that delivers revolutionary products with ‘wow factor’ that is leagues above the competition. Apple just doesn’t have anything like that with AI. Furthermore, the challenge with AI is that even though it can deliver some shockingly impressive results, it also generates some bafflingly stupid responses, as well as answers that are seemingly correct but are never less wrong in ways that are hard to determine. Those hallucinations really shatter the illusion of quality and reliability, which ultimately undermines confidence in the functionality as a whole. Given that, Apple is stuck trying to catch up to the rest of the products in the AI space while also realizing that everyday AI is nowhere near as polished as Apple would expect it to be. So the more Apple pushes AI, the more customers are likely to pick up on the errors and inadequacies of the product. Even though other AI assistants suffer from the same fundamental problems, the ones from Apple Intelligence are going to seem more embarrassing because of Apple’s brand image.
  - ForTheKidz 2y ago
    > Given that, Apple is stuck trying to catch up to the rest of the products in the AI space Oh, of course. Like...? Nothing else has access to relevant data. That's like 95% of the pitch of siri. Voice assistants are just fundamentally embarrassing to use, imo. The only exception of which I can think is while driving.
    - LeoPanthera 2y ago
      Gemini has access to your Gmail, Calendar, Docs, Drive, Keep, and Tasks, as well as Flights, Hotels, Maps, YouTube, YouTube Music, OpenStax, and of course the entire web more generally. > Nothing else has access to relevant data. Simply not true.
      - ForTheKidz 2y ago
        [flagged]
        ethbr1 2y ago
        At the end of the day, personal context is king for useful AI. Which means Google or Apple. Or Microsoft on the business side. Nobody sane is going to share all their personal data (about everything) with a fly-by-night random AI app/company. Apple has time, in the same way it did with Maps, because despite the marketing push no one is buying phones solely because of AI features these days. They're still an "Oh, that's nice."
        TwoPhonesOneKid 2y ago
        > Nobody sane is going to share all their personal data (about everything) with a fly-by-night random AI app/company. They absolutely will if the tech giants refuse to work with each others' data silos.
  - jajko 2y ago
    That 'wow factor' is a thing of the past, very firmly. When I put next to each other say Iphone 16 pro max and Samsung S25 ultra, none sticks out, none has anything 'wow' for regular users. Just different styles of working with device and small + and - all over the place. Same for tablets, notebooks, and so on. Vision pro seems like a failed product even according to Apple. Its doesn't mean that each product doesn't have something a bit special but competition caught up long time ago, sometimes went ahead (better batteries, bigger cameras etc). These days there is much more 'yawn factor', and its across whole industry.
- anon7000 2y ago
  Well, the current iteration of Apple intelligence doesn’t enable personal context or 3rd party app related features at all. (https://www.macrumors.com/2025/03/07/apple-intelligence-siri-features-delayed/ https://www.macrumors.com/2025/03/07/apple-intelligence-siri...) So who knows whether people people want it or not; it’s not even out in the world! What Apple has actually discovered is that this is a hard problem to solve. But people do want a smarter Siri; they complain about it constantly.
  - ethbr1 2y ago
    Making functional AI ('It does all things') work well across a platform is an integration problem. Which are historically the hardest problems for engineering orgs to ship: precisely because they cut across team/org boundaries. Suddenly the Maps team need to give a shit about the Calendar team, etc. And both teams tend to exaggerate and say 'Sure, our side of the integration is finished... it must be a problem on their side.' Ironically, Amazon-as-a-company probably has the most experience retooling this way -- from the Bezos API memo. Because that's what it will take.
- mirekrusin 2y ago
  The devil is in the details – you do shit clippy-like integration = nobody will use it, you do it well – everybody will, even though nobody currently does in any meaningful way beyond setting a timer.
ricknroll 2y ago
„[…]struggled to manifest its work into successful product wins […]“ 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.
rcarmo 2y ago
So they needed someone with… vision?
- _justinfunk 2y ago
  More than that, they needed a professional with vision.
caycep 2y ago
Weren't there rumors that Giannandrea doesn't have full control of the AI/ML house? There were entrenched old Siri folks fighting for their turf
- moandcompany 2y ago
  Given the size of Apple and these orgs, plausible. Siri also had a bunch of folks that came in from outside, through acquisitions, and did their own thing or tried to identify themselves and their roles as something different from Apple norms. Apple had teams involved with AI/ML in several divisions also, from AI/ML (including Siri), TDG/VPG, SPG...
potatolicious 2y ago
> "appears to have found that that the AI/ML organization built at Apple struggled to manifest its work into successful product wins despite the massive financial investment" 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.
- moandcompany 2y ago
  1) This appears to be what Apple has concluded by the action they're taking and implied by the premise of the article, versus my personal opinion. 2) The examples given may have been produced by AI/ML-based technologies developed at Apple, but may not have been the work of Apple's AI/ML Organization. Many "groups" (divisions) within Apple have their own teams using AI/Machine Learning (e.g. AI/ML Org coexisting with AI/machine-learning work and teams in TDG/VPG, SPG (special projects group: the one previously working on a car), etc.

worstdevna 2y ago

It feels like Apple had a dream of a granting easy AI integrations for many different apps and workflows, only to discover that very few people wanted any of those integrations from apple intelligence. At least, not the current iteration of apple intelligence.

BuyMyBitcoins 2y ago
I suspect the most significant thing holding Apple back in the realm of AI is the fact that Apple prides itself as a company that delivers revolutionary products with ‘wow factor’ that is leagues above the competition. Apple just doesn’t have anything like that with AI. Furthermore, the challenge with AI is that even though it can deliver some shockingly impressive results, it also generates some bafflingly stupid responses, as well as answers that are seemingly correct but are never less wrong in ways that are hard to determine. Those hallucinations really shatter the illusion of quality and reliability, which ultimately undermines confidence in the functionality as a whole. Given that, Apple is stuck trying to catch up to the rest of the products in the AI space while also realizing that everyday AI is nowhere near as polished as Apple would expect it to be. So the more Apple pushes AI, the more customers are likely to pick up on the errors and inadequacies of the product. Even though other AI assistants suffer from the same fundamental problems, the ones from Apple Intelligence are going to seem more embarrassing because of Apple’s brand image.
- ForTheKidz 2y ago
  > Given that, Apple is stuck trying to catch up to the rest of the products in the AI space Oh, of course. Like...? Nothing else has access to relevant data. That's like 95% of the pitch of siri. Voice assistants are just fundamentally embarrassing to use, imo. The only exception of which I can think is while driving.
  - LeoPanthera 2y ago
    Gemini has access to your Gmail, Calendar, Docs, Drive, Keep, and Tasks, as well as Flights, Hotels, Maps, YouTube, YouTube Music, OpenStax, and of course the entire web more generally. > Nothing else has access to relevant data. Simply not true.
    - ForTheKidz 2y ago
      [flagged]
      - ethbr1 2y ago
        At the end of the day, personal context is king for useful AI. Which means Google or Apple. Or Microsoft on the business side. Nobody sane is going to share all their personal data (about everything) with a fly-by-night random AI app/company. Apple has time, in the same way it did with Maps, because despite the marketing push no one is buying phones solely because of AI features these days. They're still an "Oh, that's nice."
        TwoPhonesOneKid 2y ago
        > Nobody sane is going to share all their personal data (about everything) with a fly-by-night random AI app/company. They absolutely will if the tech giants refuse to work with each others' data silos.
- jajko 2y ago
  That 'wow factor' is a thing of the past, very firmly. When I put next to each other say Iphone 16 pro max and Samsung S25 ultra, none sticks out, none has anything 'wow' for regular users. Just different styles of working with device and small + and - all over the place. Same for tablets, notebooks, and so on. Vision pro seems like a failed product even according to Apple. Its doesn't mean that each product doesn't have something a bit special but competition caught up long time ago, sometimes went ahead (better batteries, bigger cameras etc). These days there is much more 'yawn factor', and its across whole industry.
anon7000 2y ago
Well, the current iteration of Apple intelligence doesn’t enable personal context or 3rd party app related features at all. (https://www.macrumors.com/2025/03/07/apple-intelligence-siri-features-delayed/ https://www.macrumors.com/2025/03/07/apple-intelligence-siri...) So who knows whether people people want it or not; it’s not even out in the world! What Apple has actually discovered is that this is a hard problem to solve. But people do want a smarter Siri; they complain about it constantly.
- ethbr1 2y ago
  Making functional AI ('It does all things') work well across a platform is an integration problem. Which are historically the hardest problems for engineering orgs to ship: precisely because they cut across team/org boundaries. Suddenly the Maps team need to give a shit about the Calendar team, etc. And both teams tend to exaggerate and say 'Sure, our side of the integration is finished... it must be a problem on their side.' Ironically, Amazon-as-a-company probably has the most experience retooling this way -- from the Bezos API memo. Because that's what it will take.
mirekrusin 2y ago
The devil is in the details – you do shit clippy-like integration = nobody will use it, you do it well – everybody will, even though nobody currently does in any meaningful way beyond setting a timer.

ricknroll 2y ago

„[…]struggled to manifest its work into successful product wins […]“ 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.

rcarmo 2y ago

So they needed someone with… vision?

_justinfunk 2y ago
More than that, they needed a professional with vision.

caycep 2y ago

Weren't there rumors that Giannandrea doesn't have full control of the AI/ML house? There were entrenched old Siri folks fighting for their turf

moandcompany 2y ago
Given the size of Apple and these orgs, plausible. Siri also had a bunch of folks that came in from outside, through acquisitions, and did their own thing or tried to identify themselves and their roles as something different from Apple norms. Apple had teams involved with AI/ML in several divisions also, from AI/ML (including Siri), TDG/VPG, SPG...

potatolicious 2y ago

> "appears to have found that that the AI/ML organization built at Apple struggled to manifest its work into successful product wins despite the massive financial investment" 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.

moandcompany 2y ago
1) This appears to be what Apple has concluded by the action they're taking and implied by the premise of the article, versus my personal opinion. 2) The examples given may have been produced by AI/ML-based technologies developed at Apple, but may not have been the work of Apple's AI/ML Organization. Many "groups" (divisions) within Apple have their own teams using AI/Machine Learning (e.g. AI/ML Org coexisting with AI/machine-learning work and teams in TDG/VPG, SPG (special projects group: the one previously working on a car), etc.