Apple Delays “More Personalized” Siri

Here’s a suggestion that part of the reason might be related to security due to the ease of malicious prompt injection.

It’s disappointing, the one aspect of Apple Intelligence I was looking forward to was a conversational Siri I could teach my stuff to.

I do wonder how much further behind Apple might find themselves as time goes by. Getting a slight, worryingly familiar air of a dead Apple product.

I’ve been very impressed with Perplexity’s deep research, these services have really moved on. Perplexity state their focus is upon aiding research rather than generating images or suchlike. Got my money.

Alison and I have had similar experiences with Siri (or as she often calls it, “Damn Siri” ) at home. Makes a great cooking timer, especially when you have two or three things going on thanks to named timers; we just laugh when Siri names the timer something other than what we asked for. Not much use for anything else.

For some reason Siri and CarPlay are a much, much better experience. It shouldn’t be so, because commands are issued in a noisier and more stressful environment, and by definition not on one’s home network. (The AT&T wireless network in our home area is pretty fast, but can’t touch the fiber connection we have at home.)

Siri somehow manages to get my routing, messaging, and information requests correct on the first try while I’m driving. It’s not the “more personalized” experience Apple has been teasing, but it’s more like the recognition and response engines are a completely different animal from what I use outside of CarPlay.

Yes, I use Siri every day, in the kitchen, and in the car much as you outlined. But all during the day from when I’m getting dressed to when I’m cleaning the kitchen.

Not all AI needs to be deep, but a conversational Siri, back and forth, sustaining a focus would be great. And a Siri that could look at my calendar, agenda and email and make some sense of it would be great, shuffling things around etc.

[Merging discussions from a previous topic… -Adam]

Apple just told Gruber that it won’t be until 2026 that Siri sees improvement from Apple Intelligence (so forget about vastly better Siri with iOS 19 launch or even months after that).

Daring Fireball: Apple Is Delaying the ‘More Personalized Siri’ Apple Intelligence Features

So I feel like asking myself again: at which point do reasonable and patient believers become just dumb followers getting played? Is Tim Apple in it for the win (as in Steve’s “when we do it, it’s because we do it best”) or is this just marketing fluff from a bunch of execs that have lost their ability to execute?

Let’s not go overboard—no one is being “played.” No one should be purchasing a new device purely because of a lightly described promised feature.

I’m just thinking this is not about being “played” so much as being disappointed.

Amazon is rolling out their “Alexa+” voice assistant as of March 7 (or at least that’s when I got the email from them); it is going first to owners of most Echo Show devices, a group of which our household is a part. Amazon targeted last June as their rollout date, and like Apple delayed it.

Both the “personalized Siri” and Alexa+ UI experiences are deep migrations, not bolt-ons as I understand it. The Alexa engine was built on the best technology available 10 years ago, which meant many hundreds of thousands of hand-coded responses followed by many more hundreds of thousands of hand-coded corrections and tweaks. Amazon is deprecating that engine in favor of LLMs that will depend on probability to generate responses and actions. It’s highly likely that rolling it out was delayed because Alexa affects real-world systems—we can laugh when Alexa responds with nonsense, but it’s not so funny when we’re at the top of the stairs and ask Alexa to turn on the lights downstairs…and instead it turns off all the lights in the house when we’re on the second step.

Apple Intelligence may or may not be a magic wand, but I’d rather that Apple delay its appearance in Siri than put out another piece of crapola like Apple Maps 1.0.

Totally agree. I’d always prefer a delay for what then becomes a refined product over a rushed release of a crappy product.

But in this case, the product was launched already long ago and by now is already deep in junk territory to the point of uselessness — it’s the long promised fix to that product that has been delayed over and over. That said, the actual point I was getting at, is, if this is just another delay or if the long sought after Siri fix is rather the next AirPower.

I don’t think that’s even a possibility. And I confess I had to do some searching to refresh my memory about what AirPower was!

AirPower represents a rare announcement of an Apple product that was eventually cancelled. There is a much longer line of conceptual products whose descriptions and mock-ups were floated, leaked, or otherwise made their way into our consciousness. Somewhere I have a VHS tape, circa late 1990s, of Apple’s visualization of a future where desktop computing became completely untethered from the desktop and landed everywhere from our wrists to our flying cars. Most of it has emerged in some form, years later, as hard products like iPad and Watch, and soft products like Siri.

AirPower may have started out rooted in Steve Jobs’s hatred of cables and wires, but by the time it emerged as an announced product, it was clear that Apple was a day late and a dollar short. I am still using a 2-way charging pad and Watch charger that I purchased during that era. It does not get warm in use, and lets me know when I’ve laid down my iPhone incorrectly over the coils.

By contrast, the best Apple’s engineers were able to come up with resembled in function an induction stovetop (literally, because of the overheating) than it did a svelte and nifty charging surface that suited Apple’s other products. I believe they had no choice but to consign it to the ash heap, swallowing their corporate marketing pride in the process.

Siri, on the other hand, has been out in the wild for nearly 15 years. In that time it has been extended, but the UI and underlying logic are stuck in that past. Unlike AirPower, which was after all a minor accessory, Apple has wound its tech identity deeply around Siri and voice assistance. They can afford to delay a rollout, and risk eroding the goodwill of people like you and me. But they can’t afford to abandon the improved version, because as others have stated their home and mobile devices are almost completely dependent on Siri.

And in the meantime (to maintain relevancy!), Siri can get worse, and it will get worse, until that improved version finally rolls out.

Of course, YMMV, but I actually have found Siri to be improving, personally. Certainly not getting worse. Tiny improvements, to be sure. Like for example, it can now successfully add items to the “Trader Joe’s” list, whereas it used to tell me “I can’t find the Trader Joe’s list. Would you like to create one?” And just yesterday, it correctly spelled my cardiologist’s name when I told it “At 10 a.m., remind me to call Dr. (non-English surname here)”. It seems very unlikely Siri just guessed right, but instead found a similar-sounding name in my Contacts and (correctly) used that.

Room for improvement? Sure. A lot. No argument there.

I gather you still believe they absolutely will deliver. I hope you’re right. I’m just no longer so sure about it. Now I get that you’d perhaps counter with something along the lines of “well they have to” because…

But does that really compel Apple to improve Siri? After all, all those folks have already bought into these devices and iOS. We all already use Siri (to a degree, or not) in spite of all its deficiencies. If you allow me to play devil’s advocate here just one more time: how many people do you know who say that if Siri doesn’t improve soon they’ll dump their iPhones for Androids? I don’t know any such people. Those that gave up on Siri just gave up on Siri, not on iPhone.

Classic Alexa (as I’m going to call it here) is just a scripting language. Just like (I assume) classic Siri is.

I’ve skimmed through the Alexa developer guides, and when a developer defines an Alexa “skill” it is just a set of keywords and responses. You define specific words, with placeholders for parameters and a service in the AWS cloud (typically a Lambda function) to run in response.

The Alexa system performs the speech-to-text conversion, identifies which skill should process your command (based on which skills you’ve added to your Amazon account), and hands it off to the cloud service, which does the action, possibly sending a command back to Alexa to speak a response. The result is a really great voice-driven UI to cloud services, without any machine-learning or artificial intelligence whatsoever. And it proved to be an incredibly useful system.

Will a voice-driven chatbot based on various large-language models (presumably still able to access Alexa skills already developed) work better than the current system? I guess we’ll be finding out really soon.

Of course, because I think it brings up an important consideration. Apple has seldom delivered or given up on any given product because people wanted it or gave up on it. No one wanted the first Mac—no one, including the executives at Apple Computer who were aware of the project and kept trying to kill it—until they saw it for the first time.

It doesn’t matter how many users give up on Siri. Apple Inc bases a lot of its marketing identity on that voice and the little pulsing rainbow circle. They will keep at it until they have the Siri that no one knew they wanted—like Apple Maps 2.0.

My opinion is that they could not care less how many people are indifferent to Siri 1 on their devices. It may even make their job of rolling out AI Siri that much easier.

Yes, I agree. And technical articles I’ve read over the past year have described the huge code base and necessity of coding for new cases that Alexa had not yet encountered. For fun (!) I sometimes participated in the Alexa crowd-sourced correction and enhancement process. It was illuminating to realize how many inquiries we think of as routine had fallen into holes in the Alexa framework.

I think that’s the motivation behind implementing a large language model system. It extends Alexa beyond what programmers and volunteers have taught it, and what it can easily look up on the internet because it has been shown where to find it.

It will be interesting, as you say, to see how it works out. It looks like the current Alexa system will work in parallel, and a user can easily opt back into it. Non Prime subscribers will only have that system, and some older Echo devices can’t use Alexa+. That’s much like the system constraints for Apple AI.

I had a typical day with Siri yesterday - about 40% of my commands were correctly interpreted. These are always very simple commands (e.g. “Siri, Dining Room lights on”), because with that batting average I don’t even dream of doing anything more sophisticated. I know the commands work, because issuing the exact same command after a failure (and some swearing), it will often work. Maybe I should swear before I give the command?

My bank once had a promotion where they gave you $5 any time something went wrong. ATM is out of money? Go into the branch and claim your $5. I got about $30 out of them before they ended the promotion (probably prematurely).

Imagine if Apple did something like this. You’d have to give up some privacy, presumably, like allowing recordings to be accessible to evaluate the command, and the result. I’d be signing up in a heartbeat.

I know they’re trying to keep up with the Joneses, but putting time and money into some kind of personalized, advanced version of Siri before basic Siri even gets a passing grade is a waste.

I’m in the “or not” camp. I try to turn all Siri options off.

But evidently I failed. On my new MBA (macOS Sequoia 15.1), in System Settings > Apple Intelligence & Siri, I have the Siri slider set to Off (and all the other options in the Siri Requests box are gray), and there is a button under Apple Intelligence & Siri labeled Turn on Apple Intelligence, which makes me think that Apple Intelligence is off. Nonetheless, Calendar had an event from Mail and the corresponding message in Mail had some banner about Siri. (I clicked the x in the banner, so I don’t know what it said. But the event in Calendar did vanish.) Is there some switch that I’ve missed that is allowing Siri to send information from Mail to Calendar?

My grocery store (actually five different chains, in two widely spaced locations) used to offer one item free if the scanned price did not match the shelf price. (In other words, if I was buying two of the same item, and the price mismatched, I would get one free.) That offer died years ago, and the mismatches have mounted since then. I suppose there’s no longer the same motivation to get the prices correct.

This Apple support article seems to be somewhat out-of-date, in that if you have a relatively recent Mac, these settings are now in the Apple Intelligence & Siri category, not Siri & Spotlight. Nevertheless, this appears to work for me:

• In System Settings, chose Apple Intelligence & Siri from the sidebar.
• Click the About Siri, Dictation, and Privacy… button. It should be clickable even if you have all the sliders turned off.
• In the dialog that appears, choose Mail, Calendar, etc. from the sidebar, and turn off the Show Siri Suggestions in application option.

Screenshot 2025-03-10 at 11.59.111312×820 101 KB

(And yes, that’s another button that says “About Siri, Dictation, and Privacy…”, inside the very dialog that the first button opened!)

John Gruber had a very interesting note on the potential security implications of “personalized”
Siri: SIMON WILLISON ON THE PRIVACY/SECURITY RISKS OF PERSONALIZED SIRI

I hadn’t really thought about this, and now I understand why they may be delaying the updated Siri to deal with these very complex issues.

Dave

Yep, I noted Simon’s point earlier too. It’s a little hard to wrap our heads around the fact that security with AI often revolves around how they’re prompted.

Gruber just had some choice words related to Apple’s fumbling of Apple Intelligence and Siri improvements. He’s not so much angered (his choice of word) by the delay as by what he claims was effectively fraud.

The fiasco here is not that Apple is late on AI. It’s also not that they had to announce an embarrassing delay on promised features last week. Those are problems, not fiascos…
The fiasco is that Apple pitched a story that wasn’t true, one that some people within the company surely understood wasn’t true, and they set a course based on that.

What Apple showed regarding the upcoming “personalized Siri” at WWDC was not a demo. It was a concept video. Concept videos are bullshit, and a sign of a company in disarray, if not crisis.
…
You can stretch the truth and maintain credibility, but you can’t maintain credibility with bullshit. And the “more personalized Siri” features, it turns out, were bullshit.

Who said “Sure, let’s promise this” and then “Sure, let’s advertise it”? And who said “Are you crazy, this isn’t ready, this doesn’t work, we can’t promote this now?” And most important, who made the call which side to listen to? Presumably, that person was Tim Cook.

Sure looks to me like Gruber is saying Apple played us.

Or the well-briefed and deep insider Gruber is sympathetic to certain voices within the company, it has the feel of internal strife, perhaps a push on Giannandrea. If there was some unity I think the messages he would be getting would focus on the challenges a deep, personalized Siri faces with issues of confidentiality, privacy etc, pace of development change, ie issues emerging as testing unfolded. Which I am sure there are anyway.

In terms of the ‘state of the company’ which Gruber uses this to open things up, whether it’s a disconnected Cook or a complacency emerging with all the success, he has a point. The company has not been design/product led for a long time and marketing and corporate appear to be the ones in the driver’s seat, revealed here with this crisis.

If they weren’t ready and they still put it out there it was coming, that was panic at the top.

Pitchbook said that 60% of VC capital in the US last year went into AI ventures. Unreal. And Apple clearly behind the field on various fronts, they must be feeling the heat. I can’t believe the level of quality in Perplexity Pro Deep Research, and that’s a consumer level app. I wonder what the top researchers in Apple are showing the management team as to what’s possible.

The stakes are high and a flub, a Siri that’s a joke in comparison to the competition, is exactly what will kill the credibility stock Gruber talks about.

Gruber has more about the all-hands Siri team meeting leaked to Bloomberg.

It turns out my hunch about incompetence and lack of focus inside Apple being implicated in our miserable “Apple Intelligence” experience was even understated:

Sadly, ossification and risk aversion taking over previously innovative teams and companies is not a new problem. But Apple is a special case because it is the perfect breeding ground for incrementalism and not-invented-here syndrome due to its massive size, current senior mangement, dependence on a single product for most of its revenue, and extremely secretive and insular culture.

The history of the composition of the Dow Jones Industrial Average is a great illustration of how “nothing is forever” in business:

Let’s stay on the topic of Siri and not devolve into general Apple whining.

Thanks for sharing and what a disaster.

Full disclosure: for years I’ve thought about converting my frustration into a blog designed to humiliate Siri over its ridiculous performance. Haven’t gotten around to doing so yet.

Also worth noting that Apple convinced tons of us to buy iPhone 16 in order to get AI and now people are pissed about the bait and switch

I’m waiting for somebody to finally sue Apple for running an ad with Bella Ramsey.

Having seen the fruits of AI over the past 2 years, most of which resemble what would result if you asked AI to draw a cross among three fruits, I was underwhelmed by Apple’s introduction of AI in iPhone and did not see it as a reason to upgrade from an iPhone 11.

An apporanamelon. (AI by ChatGPT)1024×1024 42.5 KB

The current trade climate, on the other hand, did persuade me to replace our iPhones over last weekend. It was a nice bonus that Apple offered US$130 each for the old ones, which was a much more generous offer than our wireless carrier. (AT&T wouldn’t even consider any model below iPhone 14 Pro.)

I do understand why someone might feel misled by the ads, but I believe it was a really weak selling point to begin with, and I haven’t seen anything to change my mind. The new iPhones feel almost identical to the old ones (which is good!!), and they seem to respond a little faster.

That’s all we want from them. I don’t need my phone to make my day artificially better.

To be honest, I think that the vast majority of iPhone buyers didn’t care at all, if even know at all, about these features when they bought these phones. I think the pissed people are the smaller number of techno-nerds and they didn’t buy iPhone 16 models because of Apple Intelligence - they would have bought one anyway, because they always buy the new ones - and love the schadenfreude of criticizing Apple (or any other company) over any shortcoming.

I do think people are a lot more aware of the well-publicized shortcomings of Siri these days, but don’t think of it as AI, and experience it equally on older iPhones that don’t have Apple intelligence.

Since we’re being honest, let’s try this. Apple pitched a story that wasn’t true. They showed people “future functionality” indicating it was a demo when actually it was all staged bullshit.

They took all that trust they had built up over the years and threw it out the window all just because overzealous marketers wanted to scream “us too” and not a single exec had the stones to say stop. For that they deserve all the legal fallout that’s coming their way. Every bit of it.

It will take years for them to rebuild the trust they so casually sacrificed for that ‘event’. Delivering on each and every one of those lofty promises promptly and with excellent quality will be just a start.

I would have bought one either way, but I’m not a typical consumer. But Apple spent millions on the AI ads, and millions more developing the technology, so you might want to drop an email to their CFO that that was a vast waste of money :-)

Including me. I couldn’t care less about AI. Or the latest camera. Or all of the cool iCloud services or other gee-whiz things Apple is always advertising.

I want my phone to:

• Make and receive calls and texts
• Support Apple Pay with my various payment cards
• Maps/navigation
• Play a dozen or so games (mostly puzzle-type games)

And do it without crawling to a halt or draining my battery.

Which means I don’t go looking for a new phone until my favorite apps get “upgraded” to something that is no longer compatible (or no longer works well) on the phone I’ve got. Historically, this means it gets replaced in 5-7 years (and with 1-2 battery changes over that time). I’ll get the new model because it will be able to run my apps, and nothing else really enters into the equation at all.

Maybe Apple did lie in their ad. But I don’t care, because I don’t care about the feature they were hyping. Just like I didn’t care about any of the other features they were hyping in the past.

On one hand, yes. On the other hand, Apple’s behind in providing hallucinatory Skynet AI? Can I buy extra for them to stay behind?

When you wrote it that way, the connection of AI security and how it is prompted clicked immediately. As I understand it, you need to personalize speech recognition to the primary user for best results. When I use Otter, the App seems to translate my voice differently than that of whomever I interview, especially if they have much of an accent. I recall running into a security system somewhere that is supposed to be able to identify my voice. Whether or not that’s a good idea is another question, and time may tell on that. Personally, I don’t use Siri’s features at all; after decades of earning my living by the written word, I prefer to ask questions through the keyboard.

Heh, heh! You just don’t like to talk on the record . . .

As the voice recognition gets better and better I find I’m using Siri more but not for what seems to be what most people on this forum want. Text messaging when out and about. (Yes, I have to edit everything it says but I don’t have to sit down and laboriously type out my cunning paragraph in the middle of the sidewalk.) What temperature is it outside? (Yes, I have a watch and a phone and a workstation but I don’t want to touch any of those as I stumble out the front door.) What’s the word for disgusted in French? (Dégoûté, if you must know, and, no, I won’t tell you why I needed it.) What’s 7% of 2447? (171.29) Ask ChatGPT what the most life-saving drugs of the last hundred years have been? (Antibiotics and vaccines by a landslide.)

Simple questions instantly answered with no fiddling to find the appropriate app. That’s useful.

And what’s with this “I’m turning this new Apple AI off right now!” How do you know if it’s useful if you won’t try it? “Well, it takes 6G of space!” and how many of you have 2TB drives with 200G used? Try it! And if you don’t like it or parts of it, turn it off (email categories—what a disaster).

I do not want a personal assistant or avatar or agent. I most certainly don’t want machines to correct how I express myself (spelling excepted but even there the machines are kinda stupid). Everything I’ve seen about the attempts to do these high-level things are a horror show. If I must have an assistant because my time is so valuable, well then maybe I should hire one of those odd humans that, properly vetted, are genuinely useful and a pleasure to be around but are not for free.

Dave

I turned Apple AI on when it first arrived on my MacBook Pro. I then turned off Mail categories because they weren’t helpful, but I had an open mind about what other benefits Siri-AI might offer. The only place I saw its effect was also in Mail, providing short summaries of my emails in place of just quoting the first two lines (I don’t use Siri for much, I guess). The problem was that the summaries were less useful than the first two lines. In some cases they were actually misleading, giving me the impression that the email said the exact opposite of what it actually said. After a few weeks, I turned AI off. Siri still offers me canned responses to texts on my watch, but I never use my watch (or Siri) for texting. I would turn that off as well if I knew how.

Well, I’m a fan of using AI to research and look things up with guardrails. I typically ask Siri to launch Perplexity for those research queries. Not a fan of using it to generate anything beyond that. Image playground, no. Text clean up, no.

I do continue to use Siri in the usual way that has proven useful, eg set a timer, read latest messages, what’s the weather like. All I wish is that it could be conversational and hold a line of thought from query to query.