Apple Says Personalized Siri Features Shown at WWDC Last Year Were 'Real' and 'Working'

[sunapple]

It totally worked, only took 200 takes for the video, showed features that shocked some of the engineers because they were not ready, was not ready for a live demo, and now just needs an overhaul of the architecture to work like we showed in that totally working video. Really don't understand the criticism guys, we are just fine tuning like you'd expect from us. Hey look over here, we redesigned our OS!

 

[Joseph C]

I think it’s important to remember that Apple has a much higher quality bar to meet with MacOS/iOS AI integration.

When ChatGPT hallucinates a response about putting glue on pizza, people laugh at the silly AI and quickly forget about it. There is an acceptance that LLMs hallucinate sometimes. It is considered harmless because it is in a chat window that needs to be explicitly used. It is not considered personal, and is accepted as experimental.

Using a potentially hallucinating AI to manage your calendar, contacts, email, and other personal items is a disaster waiting to happen. Nobody wants a hallucinating AI touching their personal stuff (a lot like nobody wanted mandatory U2 in their music library). A clear example of this problem is the paroxysms of public outrage that occurred when Apple’s AI made some mistakes with news summaries. Nobody expresses such outrage when ChatGPT makes chat summarization mistakes. Most people don’t even notice.

Apple is being subjected to a double standard with regards to AI. Not because of animosity towards Apple. Because Apple needs to apply AI to everyone’s personal items, where people’s tolerance for errors is practically zero.

I think this is a good post.

For me I'd rather wait and get something really good than have something truly bad and possibly destructive now.

 

[Haiku_Oezu]

I don’t know what to think, it’s not that I deem Apple execs incapable of lying but I don’t see what they benefit from lying in an interview when they can just say nothing and obtain the same result
They didn’t even need to address this particular point, the lawsuit they’re facing right now is about not delivering a feature they advertised - to my knowledge it’s not illegal to do fake demos (if it was Microsoft, Google and 90% of SV would have been found guilty innumerable times)

On a related note I don’t believe for a second Apple Execs were scared to face Gruber lmao
The man gives the most safe, softball “interviews” and I’m positive the questions are screened prior to the event

I definitely believe it could be pettiness but people act like they were scared to be questioned about the AI **** and I just don’t understand why you’d genuinely think that

 

[threesixty360]

I think Gruber needs to publicly apologise a bit here. I think the concept of “vapourware” is that something literally doesn’t exist. Whether it’s an animation of what it should do, or specification, there is no actually built product. That’s kind of what he accused Apple of doing.

For anyone who has ever developed any software commercially you will know that there is often a huge difference between 80% working and 100% finished. It’s the edge cases that end up getting you. Sometimes the problems you find in the last 20% mean you have to start again and wipe the previous 80%. It happens. It happens a lot. I think Apple are saying this is what happened to them and it’s knocked them back.

Gruber’s criticism of Apple nowadays talking about a product that they are not imminently able to ship is valid. There was a credibility to how Jobs used to do things in that way. The only problem is Cook has made Apple a far bigger company that is far more important to the wider investment community and the USA economy as a whole. And that community needs a roadmap otherwise they think a business is dying. They don’t do “Steve Jobs level secrecy”. That type of secrecy is what tanked Apple stock by 40% a while back when some investors got anxious about apples future.

So Apple ended up getting into the game of promising rather than being certain. Committing to road maps that they could not always 100% guarantee.
They are now going to have to dial that back and focus. But I think it’s ludicrous to suggest that Apple didn’t have working prototypes back at the office of anything they showed.
Gruber should have known that.

 

[ifxf]

Given that it is Apple, I bet they couldn’t figure out how to keep Siri G rated given the inventiveness of the users.

 

[macbookj0e]

It doesn’t matter what the reasons were or where they were or weren’t with it. The problem was they were caught off guard with AI for reasons unknown.

They’re a tech company with almost infinite resources. That shouldn’t have happened.

 

[jimscard]

They gave themselves 9 months to figure it out for a Spring 2025 release and couldn’t close the gap. This strategy clearly worked for them in the past and they ended up on the losing side of the bet they could do it again. At least they are owning up to it, which is something a boatload of people said they wouldn’t do.

You also have to consider how far the on-device AI field came in its second year of existence. Look at software like ollama on the Mac, PocketMind on iOS and iPadOS, etc. now vs. a year ago.

 

[maxfromdenmark]

OK Apple, so you say the parallel universe exists....

 

[jimscard]

It doesn’t matter what the reasons were or where they were or weren’t with it. The problem was they were caught off guard with AI for reasons unknown.

They’re a tech company with almost infinite resources. That shouldn’t have happened.

They weren’t caught off guard at all. Why do so many people confuse a very conscious and strategic decision to build an innovative on-device AI system that’s not just build another giant, datacenter-based cloud chatbot with “getting caught off guard?”

I keep reading so many people talking about Apple being “years behind” — when on-device inference has only been possible for just over two years, during which time, the field went from focusing on model training, to fine-tuning and adapters, to tools, to RAG and agentic AI. Unlike other companies that are still focused on chatbots, Apple is building - and with this week’s announcements, enabling developers to build - AI based systems that are actually useful.

 

[stevet]

we can all put the “Is Apple pissed at Gruber talk”. That relationship is over with this regime it seems.

 

[smaffei]

The "Intelligent Siri Backpaddle Tour 2025" sponsored by Apple Legal.

 

[Compote]

But she goes to a different school.

 

[hanser]

I am still waiting for the obligatory "Steve Jobs would never have" comment, so that I can reply with: https://nymag.com/intelligencer/201...ked-his-way-through-unveiling-the-iphone.html

 

[maj71303]

Isn't there an ongoing class action lawsuit about this AI stuff?

Exactly this was to not say anything that will make the lawsuit worst. But if they got proof this was no working the lawsuit will succeed. But company ethics for any company not Japanese, will always be to deny and never admit anything.

 

[jpcyr]

I think Gruber needs to publicly apologise a bit here. I think the concept of “vapourware” is that something literally doesn’t exist. Whether it’s an animation of what it should do, or specification, there is no actually built product. That’s kind of what he accused Apple of doing.

For anyone who has ever developed any software commercially you will know that there is often a huge difference between 80% working and 100% finished. It’s the edge cases that end up getting you. Sometimes the problems you find in the last 20% mean you have to start again and wipe the previous 80%. It happens. It happens a lot. I think Apple are saying this is what happened to them and it’s knocked them back.

Gruber’s criticism of Apple nowadays talking about a product that they are not imminently able to ship is valid. There was a credibility to how Jobs used to do things in that way. The only problem is Cook has made Apple a far bigger company that is far more important to the wider investment community and the USA economy as a whole. And that community needs a roadmap otherwise they think a business is dying. They don’t do “Steve Jobs level secrecy”. That type of secrecy is what tanked Apple stock by 40% a while back when some investors got anxious about apples future.

So Apple ended up getting into the game of promising rather than being certain. Committing to road maps that they could not always 100% guarantee.
They are now going to have to dial that back and focus. But I think it’s ludicrous to suggest that Apple didn’t have working prototypes back at the office of anything they showed.
Gruber should have known that.

You need to read again the post of Gruber. He's really step toeing into this and telling you everything how it ended up to this conclusion. I never seen him, in 20+ years, making that of a bolt statement. He never had, and I think he's really touching something there. After his post, there was a big shake up internally at Apple (without saying it just because of him). It's not because the VPs in interview say it was a working demo that it is. They most certainly had begun working on it, but it was certainly not close to a demo (stage 1 as he refers it in his post). So the marketing department filled the gap. Remember there is a lawsuit regarding those marketing promises, so they probably are just told not say otherwise.

 

[b17777]

Isn't there an ongoing class action lawsuit about this AI stuff?

Sue over what exactly ?
Thats like suing your parents who said they'd get you pony for Christmas and it never came.

 

[michaeljk]

I was relieved to watch this interview. Glad to hear that they had a "real" working version of what they showed last year, but not good enough to release. Makes complete sense. And relieved that there is this whole other architecture they were working on separately and, rather than releasing a crappy Siri version now and a better version later, they are waiting until the better version is ready. Yes, they should never had tried to shoehorn an LLM into an already super crappy Siri and get it out the door asap. And also, I'm glad they scrapped it entirely. Gives me (some) confidence that they've learned their lesson about how dumb Siri is, and they can't afford to continue to pretend Siri is good enough, when it never has been and just plain isn't, good enough.

 

[michaeljk]

Craig and Joz say it in the interview. It's delayed until 2026 because they tried build an LLM Siri on top of a bad foundation of original Siri and realized post-WWDC 2024 that wasn't going to work and had to start from scratch with a fully rebuilt LLM Siri. I actually think they're being uncharacteristically transparent about this given how Apple usually operates.

They actually said that, at the time they were working on the crappy Siri version for near-immediate release, they were simultaneously working on a completely different (not version) but totally new architecture for Siri/LLM and then realized that the crappy version was never going to be good enough, so they would just wait until the second architecture was ready (which it sounds like they knew was not going to be ready until some time later, as in 2026).

 

[KurtWilde]

do you actually have data/information that makes you convinced they are lying?

That no one outside of Apple has seen any version of that Siri or set of Apple Intelligence features points to it.

The fact that they ‘demo’d’ something that was in such a poor state it’ll be close to two years before the public sees it.

 

[orpheus1120]

I do not believe this.

do you actually have data/information that makes you convinced they are lying?

There is no need for any data/information needed for personal opinions. We can believe what we want without justification. That's freedom of speech and beliefs.

Conversely, do you actually have data/information that makes you convinced Apple is telling the truth, and John Gruber was lying about Apple in his article Something Is Rotten in the State of Cupertino which MR reported and perhaps that was what tipped people not to believe what Apple said?

 

[ThailandToo]

Yeah, maybe you should stick to this in the Politics forum, because the idea that anything Gruber says is intense enough to be 'hate speech' does a disservice to actual hate speech. Maybe you just don't agree with him and it makes you cranky?

That’s not it. I go to a tech site to read about tech not politics. I believe all politicians in America are liars, greed focused, and the media follows suit. The thing is I have noticed on both sides that hatred is what is happening in the US. It should be let’s figure out what we agree on but everyone wants to write off everyone who doesn’t agree with them.

If I want to read about politics, I don’t want to hear any biased opinions or drama as it’s not entertainment like the media companies want you to believe. Not getting it from social media either.

Jon Gruber was an Apple tech blogger but go look at any time his last 50 posts and you will see at least 25 of them are politics not tech. And of those, it’s all hatred and calling people names, accusing them of things that are opinions. If Jon studied political history and politics in college and wants to write on it, it might be interesting. But never once has he sold me on anything he says that’s not about tech. I don’t care if he’s right or wrong in assessment, to me it’s hateful and spiteful and not why I want to go read his blog. It has caused me to check it once every two weeks or so when he might have written something about tech. But he bragged and just linked to someone else’s MBA M4 as he didn’t want to write the review even though Apple loaned him one for review. I am not saying he should have to write a review, but don’t hate on Apple, hate on Google, and hate on all politicians that don’t lean your way. Don’t hate. Find common ground.

 

[Rogifan]

This is just a reminder of what John Gruber wrote in is blog post back in March.



Re-reading that post it seems Gruber came to this conclusion based on the fact that no one at Apple did a live demo of the personalized Siri for members of the press. I can understand why Apple execs would be p*ssed about that.

 

[Jakeycov]

I think it’s important to remember that Apple has a much higher quality bar to meet with MacOS/iOS AI integration.

When ChatGPT hallucinates a response about putting glue on pizza, people laugh at the silly AI and quickly forget about it. There is an acceptance that LLMs hallucinate sometimes. It is considered harmless because it is in a chat window that needs to be explicitly used. It is not considered personal, and is accepted as experimental.

Using a potentially hallucinating AI to manage your calendar, contacts, email, and other personal items is a disaster waiting to happen. Nobody wants a hallucinating AI touching their personal stuff (a lot like nobody wanted mandatory U2 in their music library). A clear example of this problem is the paroxysms of public outrage that occurred when Apple’s AI made some mistakes with news summaries. Nobody expresses such outrage when ChatGPT makes chat summarization mistakes. Most people don’t even notice.

Apple is being subjected to a double standard with regards to AI. Not because of animosity towards Apple. Because Apple needs to apply AI to everyone’s personal items, where people’s tolerance for errors is practically zero.

THIS! I've been using Gemini quite a lot on my other phone and it makes mistakes all the time. Especially when dealing with email, files, notes etc. Or anything basically! We should be angry at Google for launching half baked beta software onto the masses more than at Apple for holding back when they couldn't get it right. Gemini is a hot mess. I showed my mum some of the Gemini live camera stuff which I thought was impressive, but it got all the plants wrong in my mum's lounge - queue laughter from her! You simply cannot trust ANYTHING it says or does. Apple couldn't run with that.

 

[turbineseaplane]

This is just a reminder of what John Gruber wrote in is blog post back in March.

View attachment 2518872

Re-reading that post it seems Gruber came to this conclusion based on the fact that no one at Apple did a live demo of the personalized Siri for members of the press. I can understand why Apple execs would be p*ssed about that.

I don't know -- I'm no Gruber defender of late, but I appreciated him roasting Apple.

They never should have gone as public as they did with ADs and saying things were coming to the degree they did.

They gambled and lost and really shouldn't even be "gambling" in this way anyhow.

 

[OGdrknight]

So, will this release before iOS 27?