532 comments

[ 2.9 ms ] story [ 333 ms ] thread
(comment deleted)
Not surprised. No major expansion in functionality for years
Siri is mostly useless. It works about half the time for me. When I ask it to do a task like create a timer, it works, but asking it questions is almost useless. One time I asked "What is a picometer?" and it gave an answer from Wikipedia, but then I immediately asked "What is a femtometer?" and it said it couldn't answer it on the iPhone. My wife gets amused whenever I use Siri because half the time I'll start swearing in frustration at how stupid it is, and yet like Charlie Brown and Lucy, I keep trying to kick that football.
“This is what I found on the web for XYZ”… all the time!

At this point, they should just pipe the #1 search result into ChatGPT for a nice summary and have Siri read that out loud. It would be so much better.

Yes! My big complaint about Siri is that she's just the same old "here is a list of preprogrammed actions you can perform, with a fallback to just dumping it in a Google searchbox" but you don't get access to the list. So finding what she can do involves outright guessing, scouring the web, or closely watching Apple demos. She can do quite a bit, but you'll never find 90+% of the options because Apple only shows you a small handful of suggestions.

https://support.apple.com/siri

Discoverability is a huge problem. Also, Siri can't launch apps, which seems like it should be an extremely common operation. "Hey Siri, launch Words with Friends".

Try "Hey Siri, open Words with Friends"; at least on my iPhone, Siri can usually open apps for me when I use that phrasing.
Can tell they downgraded the long range microphones somewhere between iPhone 4 and 7, might be even worse in later models because it really struggles to pick up you shouting across the room at it when it used to work pretty well on the 4.

Then again the only thing I've used it for since launch is setting timers, rarely gets that right these days.

I don’t think the iPhone 4 had always on siri.
It had, as long as the phone was charging.

IIRC, "always on" Siri came with the 4s

At least Siri doesn't pester you to buy various things or to use it for additional purposes like Alexa does as far as I know.

But, yeah, all the voice assistants are pretty bad or at least bad enough that I mostly give up trying to use them for anything other than certain rote tasks. Siri may or may not be marginally worse but none of them are good enough to, say, really use hands-off in a car unless I've carefully pre-defined tasks to perform. (e.g. pick from a handful of memorized playlist names).

> At least Siri doesn't pester you to buy various things or to use it for additional purposes like Alexa does as far as I know.

I was pretty happy with Alexa until they started that crap. Retired it immediately. Now I only have a HomePod but even on this I have Siri disabled.

Amazon screwed the reputation of the voice robots for everyone

> Amazon screwed the reputation of the voice robots for everyone

I don't think amazon screwed the reputation of anyone's assistant but their own. Plenty of people I know use voice assistants at home, none use Alexa.

Voice assistants are genuinely useful for some scenarios, especially in the smart home space.

Gah, my mother has a few Echoes dotted about her house and it’s the only inanimate object I feel compelled to tell to shut the hell up on a regular basis.

Give me a direct response if warranted, otherwise a simple chime or acknowledgement. Multiple years into ownership no-one wants a “by the way…” with upsells to other Amazon services.

It’s even worse if the reason you were talking to it in the first place is to turn something down so you can actually speak to someone.

Make a routine with these commands. It has to run daily because those cowards at amazon aren't quite ready to let us off the upsell leash.

1. Set volume to 0

2. "Turn off by the way"

3. Set volume back to your preferred.

I wonder what the conversion on 'by the way' is. Like how many BTWs have played and how many customers have taken alexa up on that offer?

"By the way" is alexa's entire business model. They sell these devices at a loss and don't require a subscription for the nlp so they can "by the way" you into buying random shit. I'm surprised they let you mute it at all, especially after devices lost $10 billion last year.
I wonder why they sell these things at a loss. Wouldn't they sell at cost?
Internal cost accounting is complicated.
At 10 billion dollars the accounting doesn't have to be precise.
I am not really sure why people buy into this. These devices cost next to nothing to make, the royalties to the tax haven holding co and the private jet fuel are thrown in to the "cost" as a write-off and a tax dodge.
Then maybe it should try to make a sale instead of blathering on about how I can ask it what sound does a pig make.

They have 15 years of purchase history, prime video usage, and presumably are snooping on what podcasts I play over Echo. They should be able to suck a little less. Any random page of a 20 year old sears catalog would carry more relevance to me than anything Alexa has suggested.

I HATE that it does that, it made me try to move to Homepods, but even with Homebridge I couldn't get everything to work with Homepods. Also, it has its own drawbacks, like no screen to see a timer.
This is exactly the reason why I replaced all Amazon echo devices with HomePods. Siri makes many mistakes, doesn't hear me correctly, doesn't understand me correctly, and so many other issues. I can tolerate all these mistakes that Alexa also makes.

I just cannot tolerate Alexa talking back to me, especially when I'm trying to do something important. I've lost my train of thought too many times because of Alexa.

Same with Google Home! One too many times, I asked it what temperature it was outside, and it went on some monologue about how I should try asking it to give me a summary of my day or something.

The optimal amount of times a product should interrupt your workflow to tell you about a new feature is 1 or fewer.

Why not use Google Nest? It has Google Assistant, which seems to be better than Siri.
Just preference honestly, we have a lot of apple devices, and it's nice to be able to keep things consistent. If we both had Android Phones we would've probably used Google assistant.
I generally like Siri and rarely experience bugs. I use it to set timers, get the weather, play music via Spotify, occasionally dictate text messages, and look up trivia mid-conversation. It's been my replacement for Alexa ever since I got creeped out by Amazon's audio data retention. In this context I'm suprised to hear Apple employees are unhappy with it.
Hey Siri, “add 2 minutes to the timer”. It can’t even do something so basic as that.
if you're using an anonymous timer, then it is difficult to access it later. you might have better luck if you set an id to reference later.

timerId = setTimer(float time);

function updateTimer(timerId) {};

maybe wrap that in a class, and use that when using Siri. you just have to speak the class definition first.

Hey Siri! set a timer using myCustomClassTimer!

It does have that ability on the HomePod minis that I have around the house. If a timer is set on one of them and then I want to reset it or add minutes to it, I can do so (but only when that HomePod mini hears me).
It works on watch and HomePod only.
The inability to set multiple simultaneous timers is annoying. I compensate by setting multiple alarms. "Delete all my alarms" works when I'm done, but it's certainly a hack.
"hey siri, start a 10 minute timer for the oven"

"hey siri, set a 30 minute timer for the rice."

it works just fine for me, use it all the time.

Doesn't work for me. I get this:

>There's already a [number] minute timer. Replace? Confirm/Cancel

Running iOS 16.1, so unless they added this feature in the last few months I'm not sure why we'd see different results.

edit: Perhaps you're using HomePod? Apparently Siri on HomePod supports multiple timers, while Siri on iPhone does not. Goofy.

My Apple Watch supports multiple timers, and yet my iPhone does not (iOS 16.4.1).

I mean, I guess a watch is supposed to be for telling time or whatever, but this is just a weird feature omission…

siri definitely has multiple timers. BUT you can only have two though and it only works on Apple Watch and HomePods.

I know, right?

Also a little known hack to set a timer you don’t have to say the whole thing just press the Siri button or say hey siri ‘n minutes’ and it will set a timer for n minutes.

On Alexa I run many times simultaneously. Didn't know it was a special feature. I just name them.

Latest update(?) it quit announcing expiration by name if only one was running. Just beeps, and maybe I forgot what it was supposed to be timing. Annoying.

Especially annoying if more than one person in the house is setting timers.

Same experience. I use it for the same purposes and it’s great for hands-free, screen-free interaction of these simple tasks.
I generally find Voice Over Utility to work better, in terms of available commands and what it can do, but it's also harder to use.
Siri, Alexa, and Google Assistant are all generally good at those things but honestly not much else. I use my Google Home every day for timers and to turn on/off lights. I don't think any of them will ever be the helpful AI that they were initially promoted as, though.
I'm shocked how much the quality of Google Assistant has degraded over time.

I have a Mini that consistently misunderstands broadcast requests and says "sorry I'm not playing anything right now". When it occasionally speech-to-text converts a broadcast word, it consistently cuts off the first word or letter, even when it's "I'll be right down" users will get "L L be right down".

It used to support simple offline requests like SMS and Navigate when data was unavailable. No more.

It used to integrate with Google Keep. No more.

No longer recognizes the word "torch" as a synonym for flashlight. Why would I be asking to turn on my phone's "porch"?

Painfully slow replies, even under ideal network environment... Just spinning forever, often until a timeout that it doesn't even have the decency to respond with a proper error message.

It's just amazing how they launched a product with a clear "this is where we are, this is our vision of where we're going" and they still sell it but instead they're going in the opposite direction.

See, at least Siri usually is ready to take your input the moment you press the button, even if it then casually discards your input because it can’t reach apple’s servers or whatever.

Google Maps: I swear, about half the time I try to activate voice search, it sits and spins before even accepting any voice input at all. Why can’t it just start reading the microphone right when I activate it, and then submit the saved audio whenever it’s done getting set up? It’s so abysmally poor that it’s usually faster to scroll through recent destinations or literally grab the phone, unlock, and put in a destination.

This is just a market begging to be disrupted. I want to see a startup combine Whisper, GPT, and a competent TTS model into a killer voice UI!

What flabbergasts me is how often the screen will display a successful speech-to-text capture, and then it poops the bed anyways. Like, you did it! You did the hard part! The part that feels like goddamned magic to me, converting the noisy messy reality of sound-waves into text. And then it drops the ball on the simple pile of "if" statements it takes to convert that text into an action.
My pet theory is that they were launched by the A team, who got replaced by the B team when the A team left for greener pastures. Or maybe they got it working, but in a bid to get another promotion they kept tweaking it, making it worse in ways that matter to us but not the promotion committee.
>I don't think any of them will ever be the helpful AI that they were initially promoted as, though.

I must disagree. ChatGPT-style LLM functionality with ElevenLabs-quality realtime voice synthesis will absolutely supercharge these products. The ability to e.g. answer kids' questions in simplified English according to parental prompt guidelines, or drill down on complex educational topics, or maintain context over many back-and-forth conversational interactions will be huge.

Strongly agree, tons of educational and entertainment value will be unlocked.
I'm really excited for LLM D&D dungeon masters.
I gave Siri a lot of time and forgiveness before writing it off as an interface to set timers and literally nothing else, and even that Siri fails often enough to annoy me. I can imagine there’s a happy medium for other people, but I’m unsurprised that with as big a workforce Apple has now, there’s a contingent there with exactly the same opinions as a lot of us out here. Siri sucks, unless it doesn’t for you, but it still sucks for the rest of us.
I use Siri to set reminders except it is completely unable to handle reminders like “remind me this afternoon to follow up with Joe about the event on June 25”. It sees the June 25 and the reminder is now for the afternoon of June 25. I manage a theater, so this is a very very common scenario and it’s super frustrating. It speaks to there being no real understanding going on, just pattern matching on things that look like dates. I don’t need GPT4, but surely we can do better than this.
You could use shorthand for date and times in the title. I tried your prompt saying "J N 2 5" instead but apparently "this afternoon" is 5PM, so that's odd.
Siri hard codes this afternoon to mean literally right after 12 PM. So if at 1pm you ask it to remind you to do something this afternoon, it will put it at 12:00pm the following day. I just double checked it right now. It’s currently 12:18 PM, and I asked Siri “remind me to test this in the afternoon“ and it set a reminder for tomorrow at 12:00 PM
Is it just me or Siri is very bad at picking up accents (mine isn't that bad having lived in the US for a while) and background noise.

One of the more impressive things about using the Google's voice assistant was, it did very well in noisy environments. Whereas with Siri, that is quite a bit of struggle. This is only about speech-to-text, not text-to-whatever.

It deals with my oven fan and similar regular kitchen noise pretty well. I hold it close to my mouth at parties, so I can't really compare it to e.g. Google Home or Alexa's far-field capability.
The Google Home on my kitchen counter hears very poorly if I talk behind it instead of in front of it.
I have Alexa and I hate it as well as hating Siri. At this point when bing (and maybe Google IDK) can handle general knowledge questions with AI answers (and a disclaimer) I basically consider Alexa / Siri unmaintained.
Siri is also quite terrible at setting Homekit Scene. I totally gave up on it. For example I have a scene named "Play Music Everywhere" and even if I tell it explicitely with "Hey Siri set scene Play Music Everywhere" it fails miserably. Also simple task like: "Hey Siri play music on all HomePods" ends up with "Ok playing music Everywhere". When did I say everywhere??? I only want you to play on Homepods!
I worked for several years in IoT and as a consequence avoid smart home stuff like the plague, so this hasn't been a problem for me.
Yesterday: “Hey Siri, watch a movie in the dark” “Ok setting scene watch a movie in the dark”

Today: “Hey Siri, watch a movie in the dark” “I can’t find a movie called In the Dark.”

I also had a very frustrating issue randomly start happening because one of my lights had the word “lamp” or “light” in it. Thankfully googling the symptoms found others with the same problem and a solution, but it was baffling as there hadn’t been any changes made by me in over a year prior.

Well it's clearly better to use non-ambiguous keyword to solve that. Like my scene name equivalent to "watch a movie in the dark" was called "cinema". Another was called "sunset" for instance.
It worked for over a year before it became erratic. If it is problematic Apple should warn me in the UI. It worked again shortly after - making me think it is not deterministic.
I’ve a problem with setting scenes with a timer that it often asks who I am. Which is odd because it will identify me just fine normally.

It usually goes like this:

Me: Hey Siri, “bedroom on” in 15 minutes.

Siri: who is speaking?

Me: $Name

Siri: I don’t recognise your voice… setup personal requests bla bla.

Me: Hey Siri, who am I?

Siri: You’re $Name!

Me: Hey Siri “Bedroom on” in 15 minutes.

Siri: Who’s Speaking?

Me: @%#+#!!!

Siri then goes into a loop asking who I am and then ends with “Hi!”

Five minutes later,

Me: Hey Siri, “Bedroom on” in 10 minutes.

Siri: okay $Name I’ve set bedroom on for 12:08.

So it’s not the voice recognition, it’s whatever’s going on after that step.

That’s exactly how Siri is supposed to be used.

If it can do something — well, that’s the usecase. It just works.

If it can’t do - then why would even one be trying to do that! That’s not how it’s done.

That’s the Apple way. And that’s exactly how Siri is perfect!

You do know that Apple was also caught retaining raw request data from Siri as well, right?

I’m also surprised you’re surprised?

> I generally like Siri and rarely experience bugs

This is mind-blowing to me because I'd guess that ~50% of my interactions with Siri are buggy. I mostly use it through CarPlay.

It even fails to call the correct contact in my phone, which should be the easiest thing it does.

I'd argue that microphones in the car are laughably bad, which probably hurts recognition. But it should be better.
A) This happens whether in the car or not, and in the car I'm often using AirPods to get better noise cancellation. There's no excuse for this. I could talk to my Pixel 5 while driving from across the car.

B) It's not always a recognition thing. Siri often knows what I said and has a completely baffling response to it.

How do you get Siri to actually work with Spotify?

When I try to use it I get consistently laughably bad results. `Play songs by Albert King on Spotify` yields something like "Playing songs by R Kelly on Spotify". No, thanks.

Even when I ask for songs, albums, or artists I have saved in my library, Siri invariably finds something else entirely unrelated to play.

What's really frustrating is that this used to work reasonably well for me.

Same for Google. I don't understand how it's got worse not better, given advances in machine learning.
>I use it to set timers, get the weather, play music via Spotify, occasionally dictate text messages, and look up trivia mid-conversation.

I use my Google Home similarly. But understand that these use-cases are utterly trivial, and Siri (by most accounts) and Google (from experience) still manage to get them wrong too often. It's unbelievable.

I have Google Home products and a YouTube Premium subscription. By 1 of 10 times when I ask for music it'll default to Spotify, which I have never used. How? Why? One of my speakers now reacts to every request with an error when it's not the primary speaker being spoken to.

It's a novelty, nothing more. I wouldn't rely on this stuff for anything. In fact, it's a bug or two away from being permanently removed.

> I use it to set timers

Yeah, literally all I use it for is to set a timer when I've put warm beer in the freezer.

> The report today explains that Google wanted these three engineers “badly enough that its CEO, Sundar Pichai, personally wooed the group.”

How do they conduct the interviews, in that situation?

Not the usual tech megacorp brogrammer "technical interview" hazing?

“Thank you for coming. As you know, I am a busy man so let’s get started: iterating numbers 1 to 100, every third number print Fizz…”
Hopefully they skip that step but it is weird how often a recruiting process goes from having a recruiter cold contact you with platitudes singing your praises and how valuable you would be to their company and then the next step is to have a group of interviewers demanding to know why you're worthy of being in their presence while expecting you operate like a broken clone of Google and also recite CS theory that has never been even slightly applicable to the open position. But at least the 'inside-out condom' brain teaser has finally gone out of favor.
Sigh, I've been on the receiving end of that multiple times. :)

In some companies (the ones that aren't just cargo-culting or on autopilot) I'd bet that someone in HR/compcommittee consciously intends it to be hazing rather than evaluation. Maybe theory like that it makes the company psychologically seem more attractive (the brain thinks, if you're jumping through hoops for them, there must be a reason), and to take candidate's ego down a notch so they're less demanding in compensation negotiation.

Generally, you sit down to a meal or coffee with the executive in question; they answer your questions and try to convince you to join.

That conversation might span a few sit-downs or extend into e-mails.

If you eventually say “yes”, you’ll be funneled into the hiring pipeline midstream, skipping the entire front-end process.

It’s largely a formality — HR is told to hire you unless there’s a glaring red flag.

When I’ve been in that position, I wasn’t even asked for my resume until after we’d already negotiated my compensation, solely for inclusion in my HR file.

I love my iPhone, it's a great device.

However, quite literally the only two things I use Siri for are asking "what song is this?" and "set a timer for X minutes."

I'm sure there are ways I could be using it for any number of things that might improve some routine process in my life, but I never found them.

The timer, sometimes, and occasionally I try to get Siri to play something in the car so I don't have to look at my phone. It normally fails -- it will play something random from apple music before it plays anything from my library.
This drives me nuts. I get it, sometimes albums and songs have similar names. It seems like Siri should be able to deduce which I'm asking for given I've listened to one album over a hundred times and never listened to the random thing with the same name it decides to pull up.
The most impressive thing I can get Siri to do is "Hey Siri, play my <playlist name> on Spotify", and it works consistently. "Hey Siri, reply." is another decent one, but it seems to be working poorly lately.
> The company’s senior leaders haven’t shown much stomach for the kinds of headline-grabbing gaffes ChatGPT and similar services have stumbled into over the last several months.

Risk aversion from immature tech is a part of Apple’s DNA. It should be no surprise that Apple declined to push Siri beyond its limited capabilities.

LLMs are quite new and Apple has plenty of cash to hire star devs to catch things up. Writing off Apple’s AI future is premature.

You can't be apologetic about Apple by using some possible future scenario. Siri sucks, and just about everybody agrees.
If you are betting that Siri will be the same 10 years from now, I will not join you.
Perhaps, but who cares now?
> Risk aversion from immature tech

Boy, that Simpsons episode where they make fun of the Newton really burned them bad.

(comment deleted)
I used to have Google Home and moved to Homepod Minis. I should have ran them side by side, but that would have gotten very confusing for other people in the family.

Siri is fine at controlling smart home devices, giving weather, timers, etc...but if you ask it questions more times that not it wants you to use your phone. Google Home handled that so much better.

We have Apple Music and have Homepods around the house so we use it for that.

<kitchen homepod> "Hey siri set a timer for 5 minutes"

  "hey siri what timers are running" -> "it's 10:45am"
  "hey siri how long left on the timer" -> "there's one timer with 5 minutes remaining"

  <bedroom homepod>
    "hey siri list all timers" -> "there are no timers set"
    "hey siri how long left on the timer" -> "it's 10:45am"
I'd have expected speech recognition software to be good enough that you could have direct speech -> gpt type services. It almost feels like traveling to the past when asking things of Siri when I otherwise get very useful responses from chatgpt.
Speech to text is already excellent with an iDevice. Turn off WIFI and put it in airplane mode... then open up notes, hit the microphone and start some transcription.

If you say "Buy a two by four" the transcription goes from

    Buy
    Buy a
    Buy a two
    Buy a two by
    Buy a 2 x 4
"Four inches by three inches"

   For
   Four inches
   Four inches by
   4" by 3
   4" by 3"
And that's all on device.

The issue is that when going to GPT, that's going to cost someone a few pennies each time a request is made. When that's scaled up to the installed base of all Macs and iDevices that gets expensive fast.

I rarely use Siri, but my technophobe elderly parents have embraced it as a fairly intuitive way to navigate their phone settings, contacts list, etc.
I don't use Siri all day, every day. It's certainly not perfect. But did you know you can say things like...

"Turn off the Apple TV."

"Turn off low power mode."

"Stop navigating."

"Open [name of app I don't want to spend time finding in my iOS folders]"

"What song is this?" (Listens and usually is able to discover the name/artist of the song in the background.)

So when people say things like, "Siri is mostly useless," I think – wow, that is a very hyperbolic statement.

It's useless beyond manipulating iDevices, navigation tasks and simple questions. Even then, if you don't ask in the exact right way it often doesn't work.
Is this any different than Alexa or Google, though? I guess Google is probably better at looking up facts and Alexa at ordering stuff (if buying things without even seeing a picture is your thing), but neither is that useful. Is this not just a case of voice interfaces being much harder to do right than people originally hoped?

(I say this as someone who has a Google Home that I use regularly, but mostly as a kitchen timer and music player.)

I mentioned an example in another comment - basic manipulation of timers don’t work well - like, adding time to an existing timer. When I’m working in the kitchen this is really useful. There are also major performance issues - “working on it” and “sorry can you say that again” are probably two of it’s most common utterances. Random bus like “there’s nothing playing” and the wrong siri taking over the request from a million miles away and you just can’t even hear is extra annoying.
Google Home is pretty bad at adding time to existing timers too. Sometimes it works, and sometimes it creates a second timer with a difficult to access name.

Personally, I'm become convinced a good talking/listening clock is actually a useful thing, but Google isn't providing it, and it sounds like Apple isn't either.

At this point it will be a race see which virtual assistant can become universally useful first, and from what the article is implying Siri is not anywhere close to where they need it to be.
My favorite is on the Apple TV, "search for [some video] on youtube" will actually open the app and search for it.

I think the big issue is that a lot of these aren't discoverable, and because Siri isn't very smart at natural language you won't discover them by asking for something similar like you could with Google Assistant. An LLM for parsing requests could be super interesting here.

Siri mostly runs on device. Having a LLM needs a much more powerful device or Apple would need to drastically scale up its cloud buys from AWS, GCP, and Azure (which would in turn make the service more expensive).
"Fast-forward 2 minutes" to skip commercials made watching on the Apple TV way better than watching on my laptop, especially for sports.
Siri's functionality is fine for me, it just seems to really struggle to understand me compared to Google
My favorite is “[Let’s] go home”, which is what I use to navigate back to my house via Apple Maps.

There’s also sending messages to people - you can send voice messages (“send a voice message to X”), which works decently well and avoids transcription failures.

This is even better, I usually use "navigate home". Thanks for sharing the voice message bit!
Ok, maybe I’m an outlier here, but I love Siri on my HomePods. It does what I want in enough cases that it’s useful, multiple timers while cooking, setting them and canceling them when I am putting things in or taking it out of the oven.

Granted, I have very few things I use it for, but I like it. The kids like them in their rooms too. I’m happy with Siri, but maybe I don’t ask it to do much.

That's the trick with Siri. I use it on my Apple Watch when I'm riding my bike with an AirPod in: "Siri, play this podcast," or "Siri, call so-and-so."

I think the secret is low expectations.

I send text messages with it while on my bike as well. It usually works fine and gets the point across.
Every company has engineers who don't like some products or features. We aren't mindless corporates drones. I don't personally use any of my employer's applications.

Some article like this comes up every so often and it's so non-newsworthy. Anyone who's been at a project planning meeting knows there's a small contingent of engineers in the corner muttering that it's a terrible idea and it will never work.

I’m frankly surprised that Siri hasn’t moved to on-device voice recognition. It’s by far my biggest complaint with Siri: random, unpredictable delays or outright failures to recognize simple messages because the network is unstable or unavailable.

There’s other trash too: the other day I tried asking Siri to play a song from my library, but it misinterpreted the song title and proceeded to activate a seven day trial of Apple Music Voice to play some random track on Apple Music. I didn’t ask for that subscription!

To be fair, for the tiny subset of functionality I do use - navigate by voice, set timers, get terrible jokes to amuse passengers - it works fine.

Yes, incredibly frustrating that to ask Siri to play a song from your library on your device you need a network connection.
Siri does use on-device voice recognition. Active airplane mode and you can still use it.

Edit: for some commands only it seems. I agree with OP, this is odd.

Seems like it does - cool! - but why does it sometimes spin for several seconds when faced with a seemingly simple request? I'm guessing it has some path where it tries to use the network if possible, but if the network is spotty then the requests timeout and it reports a failure?
It probably takes more time to run the analysis locally on a phone chip.
When Google brought the Assistant locally to the Pixel, one of the talking points was how much faster it was.
> because the network is unstable or unavailable

All of Apple’s services by and large handle spotty networks horribly.

Take Apple Music for example: if you want to play music you’ve downloaded to your device, it’s often better to switch into airplane mode, because with a present-but-poor signal it will still try to load album listings etc over the network, and sit there on a spinner.

Spotify is just as bad in this regard! It used to handle spotty connections flawlessly a few years ago, but at some point it was altered to keep trying to pull album art and track listings from the network even for downloaded albums.
Is there not an option on Spotify to only stream on wifi?
I have an issue with the Spotify app on the Apple Watch. I’m not sure if it’s an apple or a Spotify problem, but I have loads of trouble when I’m using the watch app to play music when I’m at home, near my phone and near my Wi-Fi network. However if I leave the house and walk away without my phone, leaving the watch to use this 4G connection only, it works flawlessly.

If I come back from my run Spotify will often stop playing as soon as it gets home in range of the phone.

Not that this excuses it, but I've been able to get this to work by going into the "Downloaded" tab in the Music app. That seems to play directly off the device.
This is a common developer curse. I just ran into an issue at my job where customers were having an issue in the field that we just couldn't reproduce in the office; it took a long time to put together that the issue was related to failing downloads on sketchy networks because all of our networks are stable and fast.
This is so annoying! Every single time I try to play downloaded music on a bad connection it stalls instead of just playing the audio I literally have stored on my device. 100% agree with you that turning on Airplane mode makes it easier to play AM on iOS.
I hunch that on-device recognition leveraging the onboard ML hardware will be a whole new thing. The Siri brand is so damaged now I expect they’ll launch something new instead of “but it’s actually good now!”
Whats the excuse for this ? Whisper is open source, resource efficient and can easily work on device. I wonder why they havent switched to that.
Except they already do this? Since iOS 15, Siri has used on-device recognition. That's why you can go into Airplane mode and do things like ask Siri to change the brightness.
"Assistant" type devices which can actually do things besides reply can't yet use large language models safely. LLMs need to get past the problem of making up stuff when they don't know something. Until then, it's not safe to give them power over devices.
There’s nothing particularly unsafe about LLMs. Giving Sirious access to a weather api and timers is not gonna break the world.

If anything, giving LLMs access to apis makes them less likely to hallucinate.

Huh?

If they have control of door locks, there's a problem.

Why is that a problem? If you don't want the model to access your doors you disable it at the API level. Siri already has an option for that today.
They can - if we split it into parsing, execution and response. If LLMs only do parsing and response and execution is limited to actions that are prepogrammed, i think we’ll have a much better assistant. Like, it could be interactive and natural. But i agrees that it shouldn’t be given unfettered access to do whatever it wants.
Outside of driving, I never really understood the popular use case for these voice assistants. In most situations it is easier to type your query rather than say it out loud.

It was a novelty touted as a big leap in technology.

it's for when you can't type reasonably isn't i? Like while you're cooking you might say "skip to next song" or "set a timer for 10 minutes" or something.

That's how people describe it anyways.

I use Siri to turn on voice control when I drive. I have voice control number every item on the screen.

This lets me have 100% control of every feature/button and every app whilst I drive.

I can swipe on the screen using my voice.

I can write notes and messages.

I can switch to any app.

Voice control is a world leading accessibility feature that is a complete touch replacement controlled via voice.

Once I no longer need voice control I turn it off via Siri. The entire action of turning it on, using it, turning it off occurs with no touch at any point.

Also it's great for when I am cooking or doing diy and don't want to touch my phone.

In comparison Siri is better at a few narrowly defined actions but has less than 1% of the scope of voice control.

Reading people's complaints about Siri in this thread I feel what most people actually want is voice control.

But no one outside the visually impaired community knows about it.

Since iOS 16, Siri has gotten dramatically less accurate for me on my Iphone 13 Pro.

WebGL has been screwed since 16.4

I'm glad I'm not the only one. I find it annoying, intrusive and not useful for anything at all.

The only times I ever try to use it intentionally (like asking it to take a note, or make a hands free call while I'm driving) - it screws up or requires clarification to which I have to give screen attention, defeating the purpose.

Most of the time it just pops up and annoys me when I call my wife "sweetie".

Siri seems to be actively getting worse to me. Not necessarily at understanding commands, but at the verbosity of its replies.

Recently, when I ask it to "Send a text to <name> that says <content of text>" it says something like "I notice you often send texts to <name> using Apple Messages, so I will use that for this text. Is that Ok?" before reading me back the text and then sending it. I'm sure that there are people out there who have a rich and complicated mapping of text-communication-app to recipients, but I literally have only one text communication app on my phone, the one that it came with, and I only ever use that. It's already annoyingly slow to interact with Siri on a multi-step process, and adding another step to it is awful.

Apple and other mega corps really need to spend some money investing in training/education for these highly specific fields. Advancement in Siri shouldn't be dependent on 3 people that were lost to Google.

I know a bunch of capable programmers that would love the chance to go into one of these specialties but don't have the resources nor opportunities (time/money/location) to go back to school.

Siri could be so much better.

I use Siri fairly frequently for setting timers, controlling my smart home and the like, but for me where it falls down is really understanding natural language. Personally I found that google assistant was way way ahead of siri in this regard. I could ask it to do something and it would just understand what I want.

One issue I frequently have with siri is that commands that work one day suddenly don't the next. If I ask siri to "lock screen" it tries to find smart home door locks (which i dont have) instead of locking the device screen, eventually I figure out some combination of lock device/screen/phone screen/off etc that works, so siri does know how to do this but isn't smart enough to figure out my intent.

The other issue is that siri unlike google assistant can't maintain a train of thought, You can't say "hey siri dim the bedroom lights" and follow that up with "hey siri..a bit more" unlike with google where it seems to be aware of the context of what it did prior.

The lock screen issue is really funny. Because I have the same experience where I will come home and say hey, Siri lock screen to close the navigation. However, now it doesn’t seem to work anymore and I have tried some of your prompts and some combination does work.
> where it falls down is really understanding natural language.

I would settle for Siri understanding simple commands. Natural language is something for Apple customers in the 2040s. I want to not have to use creative language skills to decipher things in Reminders. "Who is this Paul Cage I'm supposed to call? Oh, repairing the pool cage."

Forget LLM or natural language processing; Siri needs to catch up to 2018's Google Assistant

Amazon mitigated bad speech transcription by providing a link to listen to the underlying audio.

Apple has never offered this and it has led to lots of head scratching at the grocery store as I work through my shopping list. It still beats trying to use the Alexa app while shopping, though.

Amazon is hilariously bad at understanding voices. The number of times it has played "Pure" instead of "NPR" is astounding, despite me going in to the Alexa app to report the error every time.

I'm convinced those error reports are round-filed.

2023 Google Assistant needs to catch up to 2018 Google Assistant. It has become progressively worse since I invested a great deal into setting it up with smart-home devices in that very year.
I also use it in a limited way and it's fairly successful. Just homekit turn on turn off and set scenes.

However I have one frustrating bug with shortcuts that sends me over the moon. I have a little shortcut that just sends a simple text message when i say 'Hey siri, Sweetie' about 20% of the time siri comes back telling me about musical artist sweetie... it's braindead.

Does anyone know why I have this problem with Siri on my Apple Watch?

“Set a timer for 15 minutes”

Siri: “timer app is not installed”

Excuse me? It most certainly is.

Try again later and it works.

Siri likes to be helpful. While you were wondering about the timer app, she went ahead and installed it for you! /s
You can just say “15 minutes” and it has a higher success rate for me.
I’ve had that happen a few times (not recently, probably not since the latest watchOS). It’s just a bug.
The biggest problem I see with Google is that saying "ok Google" feels like a mouth exercise, especially when i have to repeat it so many times in a day.
Have you tried "Hey Google"? One less syllable.
Interesting. But saying "Alexa" or even "Bixby" feels a lot quicker actually.
What would Siri’s market share be if you could easily install Google assistant on iPhone? How much more investment would Apple have put into Siri if they had to compete?

I haven’t noticed any improvement in Siri in close to a decade of being in the Apple ecosystem. The most frustrating thing is that every time I ask it to “turn the lights on”, it thinks I’m saying “off”. I grew up in the US and have no notable accent or mumble.

Note that there are hacky ways to add google assistant, but it doesn’t get integrated into the OS in the same way

Do you actually use google assistant?

The grass ain’t greener, it is rubbish too.