Daring Fireball: The iPhone 4S
daringfireball.net
daringfireball.net
Most people have never used voice recognition beyond shouting "Call Dave Jobs[1] Mobile" in the car. This is going to be perceived as magic, in the same way as the accelerometer in the Wii or the original finger touch interface. Anyone with a 4S is going to be showing this off to their friends, and I can't wait to try it.
There are parallels with people saying 'Touch has been around for years, but people prefer buttons' when the iPhone came out. True - until someone did it right. And now everyone wants touch.
And yes, I know it's been around on Android. But Apple make you WANT to use it. No-one's ever been bursting with excitement to show me their Android voice recognition stuff, and nor have I been wetting myself to try it out.
On the downside for Apple - improvements to the iPhone for the foreseeable future look to be on the software/web server side. They are vulnerable here. Limiting features to new models is going to piss people off, and fighting patent battles on software is much, much harder.
[1] Funny typo - I meant to type Dave Jones. Only noticed this on rereading the post.
As far as I can tell it's still the same niche users using it for the same niche tasks (talking to kids on business trips, sexting etc.). Nice to have for those that need it, but they mostly had it already via netbooks with Skype.
Likewise, I'm hearing about how this is going to transform computing for people with certain disabilities, when they've apparently just bought the same Dragon software you've been able to buy for years (for transcription, the Siri stuff is apparently semi-distinct).
Don't get me wrong, I like that I can already call up songs in my car by voice with my $100, no-contract, chinese-branded, 600Mhz ARM11 Android phone with a 3rd party, free, ad-supported app. (I amuse myself by practicing my french pronunciation with it too) The gushing just gets a bit old, as does the speculation about what magical powers the A5 must have to enable such wizardry.
In the book, people begin to go so far as to buy masks and hologram replications of themselves to look as ideal as possible. This quote sums up the key problem:
"The answer, in a kind of trivalent nutshell, is: (1) emotional stress, (2) physical vanity, and (3) a certain queer kind of self-obliterating logic in the microeconomics of consumer high-tech."
This post on Kottke.org has some nice non-spoiler citations: http://kottke.org/10/06/david-foster-wallace-on-iphone-4s-fa...
As to accessibility you should also keep in mind that Apple has a great track record with that community through their work on the Mac and the iPhone, this is likely to greatly improve it.
Siri, on the other hand, doesn't depend on anyone else having anything in particular. Siri would be just as useful if you were the only person on the planet able to use it. I think that will make a huge difference in adoption.
People attribute special powers to Apple's "marketing hype" but 99% is web generated --articles, posts, comments, user anticipation etc, not from some Apple's marketing campaign.
Except if showing the Facetime and Siri features in its tv ads qualifies as "marketing hype".
In any case, Facetime is an iOS/Mac feature. Noone expected it to catch as much as Skype, if only because there are like 80% more Skype-capable devices out there (like, say, all Windows boxes).
And Apple even made some bad decisions IMHO in promoting it: no OSX client from the beginning, not integrated with other apps (say, iChat), not promoted for PCs bundled with iTunes, etc.
A few people made the sensible point that Facetime was going to have a hard time displacing Skype due to network effects (and Wifi limitations), or indeed being useful at all for many people without large numbers of iPhone 4 using family and friends who they didn't see face-to-face enough. Many others seemed to have forgot that Skype and other such video chat tech such as iChat even existed in their excitement for this revolutionary technology.
A position I feel they were forced into since this was the big finale "one more thing" in an iPhone announcement, and therefore couldn't be considered a bit of a damp squib if they wanted to maintain face.
Perhaps. But I as a tech person find it magical. But I showed the video of Siri in action to my wife and she had an immediate, visceral hatred of the idea of talking to the phone like that "If I want to set a bloody alarm, I'll tap the set alarm button thank you".
For context, she's not a luddite, she's tech savvy, though not into tech or its own sake. But I don't think she'll be alone, I think this is a feature that many non-tech people will loathe initially, perhaps for reasons they cannot articulate.
I can see using it myself as lot as well, especially for reminders and texting while I'm driving.
@brador I think that may work with Siri actually. People try the more conversional stuff in demo because, let's face it, it's more "magical."
Instead, if it could understand something along the lines of: "alarm, 5.30, tomorrow" that'd be much better, quicker and simpler than: "Siri, set my alarm for 5.30am tomorrow morning". I don't want to have a conversation with my phone. Get in, get it done, get out. Speed always wins. But then I hate the new dumbed down Google too so what do I know eh.
Siri will flop, the speed is a non-issue for the majority of iPhone owners and the camera just about brings the phone up to spec with it's competitor cameras. If they went for a an iPhone 5 and iPad3 they would have made an absolute killing this christmas.
Yeah, instead they only pre-sold one million of them in 24h.
You also said you're waiting with your cash in-hand for iPhone5, so looking at competing Android phones isn't even on your radar. Tim Cook loves customers like you.
Not sure how smart it is keeping cash lying around to either depreciate or get stolen, but if you're going to wait for the iPad 3 and iPhone 5, you're only going to regret it, because the iPad 4 and iPhone 6 are going to be even better.
I don't even know where to start.
Like, how do you think this works? They have a magical golden goose shitting out new, feature complete design verification test models and they just decided not to "show" you the next one behind the curtain?
Apple is successful because they take their time creating their hits. The 3GS didn't imperil the company and neither will this.
That way, when people are on their 2 year contract cycles, you either sit on the "new shiny" or the "advanced shiny" cycle, so consumers rarely lose out depending which cycle their contract ends in, since there are pros and cons for both.
I'd say its smart business.
It's not like we're launching new networks every year or two, or individual ARM cores are going to make huge leaps in performance.
Have we trained everyone to worship the new/shiny when we're probably hitting a short-term asymptote?
It looks like Apple has done pretty well in spite of itself. You're always going to disappoint a certain % of people (like yourself).
Claim chowder.
> the speed is a non-issue for the majority of iPhone owners
Emm, the speed enables better apps, especially better games. It's always good to have. Smartphones have a long way to go yet.
> and the camera just about brings the phone up to spec with it's competitor cameras. If they went for a an iPhone 5 and iPad3 they would have made an absolute killing this christmas.
All the specs people wanted from an iPhone 5 are present, except a new chassis some pundits idiotically predicted. So what exactly do you miss, an increased NUMBER???
Oh, and they already started making an absolute killing.
Further, I'd challenge you to name one non-game app that would be massively improved by a faster chipset over what is currently available on iPhone 4 devices. Just ONE. That should tell you how unneeded a chip upgrade really is.
Which was phenomenally successful.
What all the hysterics are missing here is that this is entirely with precedent.
When 3GS was introduced, it was pointed out that the S stands for speed.
Yeah. Back in the day. Or in rural Nebraska.
Nowadays, smartphones are used for tons of things besides calling. I'd even go as far as to say that actual call time is far less than other-use time. Else we wouldn't have TONS of web traffic originating from mobile phones and 300,000 apps in the App Store.
> Speed is irrelevant to the majority of users I know
Are those the same "most users" that use their smartphone "as a phone"?
For those people even a smartphone is irrelevant. They can get by with some dumb device.
Now, speed is irrelevant as a goal in itself. But a faster CPU enables more powerful apps. Without that irrelevant "speed", we wouldn't have the ability to handle 8 megapixel images properly, or edit movies in iMovie mobile to name a few examples. And for 3D games, more speed is better.
> Further, I'd challenge you to name one non-game app that would be massively improved by a faster chipset over what is currently available on iPhone 4 devices. Just ONE.
Just one?
1) Any music app (like Fruity Loops, Beatmaker, Massive, etc) would be able to either play more samples / synthesize more voices = improved polyphony, or synthesize better voices = better sound quality.
2) Any video encoding/editing app, such as iMovie mobile will be massively improved in what it can do.
3) Any drawing app, like Brushes. Most use complicated code to render realistic brush responses and lag when live-painting.
4) Any app written with an "interpreted" layer, from Adobe's flash transcompiled apps to Mono Touch will feel more native. Also, MacRuby will be eligible for iOS apps soon, as the GC problem is solved, and it will greatly benefit from a faster CPU.
Oh, and a faster CPU also does stuff in less time, so if the power requirements per instruction are the same as the previous CPU (which often happens with next generation CPUs), a faster CPU means better battery life.
Basically what you're saying is: "640K should be enough for everybody".
If I didn't miss any recent developments, that was a misunderstanding from when ARC was leaked as iOS5 having "GC", then it turned out that it was actually compiler-supported reference counting. Which means that MacRuby is still stuck on iOS and doomed on OS X, with some people expecting the GC to be deprecated sooner or later.
With the current App Store rules, I don't see why Matz' C Ruby with 'dl' disabled wouldn't be eligible.
I'm just too shy to walk down the street and talk to my phone without it looking like I'm talking to someone.
I've used Siri on and off since it came out. It's pretty good. The reasons I don't use it more are precisely the problems that will be solved by integrating it into the OS. First, the app was a pain in the ass to activate, requiring half a dozen taps to get started. This problem disappears on the 4S. Second, the app couldn't do anything with the data on the phone (like send e-mails or check my calendar) because it didn't have access, or at least they didn't build it to. Again, this problem disappears with integration.
The main advantage of something like Siri is convenience and speed. The app couldn't achieve that precisely because it was an app that wasn't built in to the OS.
Apple has added new features to integrate it more deeply with things like the Calendar and Reminders apps, but at the same time, half of Scott Forstall's demo commands at the 4S keynote were exactly the same as the Siri demo at 2009's All Things D conference.
"The secret to ending all poverty and hunger has been hidden deep in the earth's crust for two decades. Is there any reason why digging it out and placing it in Times Square for all to examine (except for the fact that now everyone can see it) will cause people to take more notice now than they did before?"
So I'd say, yes, there is a reason.
> You mentioned yourself that it might be useful in the car, where you are alone and nobody can hear you.
And what if people can hear you, in the street say? It's not like we don't already TALK to our phones all the time while walking in the street. You think it matters to passers-by if you talk to your boss or Siri?
> It's literally shouting, when you really should only be whispering.
Huh? What do you mean with this bizzaro comparison?
This can be done on iPhones through the regular Voice Control feature in iOS 4. I use it frequently to change music while walking (holding the middle button on my headphones activates Voice Control).
I think that that is exactly why we (tech people) hate it so much. We like to be in control of what happens, and Siri is all about giving most of that control over to your iPhone.
Siri seems to be the dream of every guy who read or viewed sci-fi, where voice interfaces are ubiquitous.
Sure, I'd wish it didn't transmit everything to "the cloud", and I'm definitively not buying an iPhone just for it, but it reached to me like no other Apple product did.
Siri: “It’s set for 9 AM.”
Me: “Change that to 10 AM.”
Siri: “I changed your alarm to 10 AM tomorrow.”
Me: “Cancel that alarm.”
Siri: “I deleted your 10 AM alarm.”
Me: “Thank you, Siri.”
Siri: “Your wish is my command.”
It seems this thing is really fun to use. I can't wait to try it out.
Me: "Set my facebook status to getting ready to ride the roller coaster at Six Flags."
Siri: ...
Me: "Are there any headlines about my startup on the front page of Hacker News?"
Siri: ...
Me: "What Kindle books do I have on HTML 5?"
Siri: ...
Me: "Play me the latest episode of Glee on Netflix."
Siri: ...
Some of these, like Facebook, could be handled by smart first-party integration by Apple. But in general, any time a new use case comes along, first party integration is going to lag way behind, if it's implemented at all.
I don't mean to rag on Siri -- it's an incredibly cool technology even today, and it apparently works amazingly well. But the more things you can do by picking up the phone and using the touch interfsce that you simply can't do at all with the voice interface, the less I think it will slot right into people's lives and become how they interface with their phone.
a. Purchase "The 7 habits of Highly Effective People" by Stephen Covey?
b. Purchase "Getting Things Done" by David Allen?
Does its understanding of location extend to more than just home is the address I entered for myself? Exploring its capabilities will be really interesting.
Siri's understanding of context is what makes it so useful. Pair it with Watson and it'd be unstoppable. (Wait, no! We've seen how that movie ends.)
Him: "Remind me to pick up my dry cleaning."
You: "Huh? You talking to me? No..."
Later: "After picking up my dry cleaning, I want you to book me a flight to Tucson"
You: "I'm finding another seat..."
"You might wonder, Hey, don’t you feel like a jerk walking around the city talking to your phone? But here’s the thing: Siri, by default, kicks in when you hold the iPhone up to your ear, so you can talk to it and it looks like you’re on a phone call."
People get used to it pretty quickly (but will always hate it on trains and planes).
It's held up fairly well, even as bluetooth adoption has increased.
For example, in English, "Text my girlfriend that I'm coming home later" would send a message with "I'm coming home later".
However, in Dutch, saying the same thing "SMS mijn vriendin dat ik later thuiskom" should result in a text with the content "Ik kom later thuis", requiring labeling all parts of the sentence.
"Siri may not be available in all languages or in all areas, and features may vary by area".
Me: Search for Japanese food.
Siri: I found these 5 results on Yelp nearby.
Me: Search for Japanese food on Foursquare.
Siri: I found these 8 results on on Foursquare nearby.
If you explicitly request a certain service often enough, then Siri should start to default to it. That's exactly what a real world personal assistant would do. In fact, Apple doesn't have to be that creative. Just do what a real personal assistant would do in any given situation and you should be good.
Also, I would really like some reviewer mention if the iPhone 4S has less of a rolling-shutter issue as the previous iPhones do.
Perhaps you're suggesting that all services answer all queries, and then there's some sort of ranking on the results from each service to decide which one responds. Foursquare would give a low score answer to appointment queries, so the calendar app would get preference. This would be the best suggestion I've heard, but requires processing every request against every agent, which may be OK for Apple, but only some independent developers would be able to provide their own hardware. They may have to provide software to run on Apple servers...
That's all just my guess of course...
In other words, just use the code they've already got working in OSX's Spotlight "Top Hit" feature. (Which, oddly, doesn't exist in iOS Spotlight—maybe it requires a larger cache or something.)
Burnistoun 'Voice Recognition Lift' Sketch
http://community.nuance.com/blogs/dragon/archive/2009/03/05/...
Also worth noting that evidently Siri's voice can be changed, as well. That was a safe bet, but I hadn't seen it mentioned anywhere.
If you are German or French you only ever get one voice.
That's why Google offered GOOG-411, mostly to train their database. And that's why all these solutions include a server-side part, because they need the large sample database that keeps growing.
http://www.guardian.co.uk/technology/2011/oct/12/iphone-4s-s...
I grew up in the East Midlands and now live in Leeds, 70 miles north. The difference in accents is night and day (although no-one in the UK knows what a Nottingham accent is). There's even a significant difference between a Leeds accent and the next city, Bradford (about 5 miles away).
I really hope Siri has some sort of unobtrusive training built in...
(see Burnistoun YouTube video linked below)
Here's the post with the link:
"You might wonder, Hey, don’t you feel like a jerk walking around the city talking to your phone? But here’s the thing: Siri, by default, kicks in when you hold the iPhone up to your ear, so you can talk to it and it looks like you’re on a phone call."
Context switching could be done like short-cuts: when you open an app it captures (most) shortcuts; and you can reallocate short-cut bindings. These are easy improvements. Just think of both voice and touch interfaces as having grammars.
Wolfram|Alpha's natural language interface is endlessly annoying for me - I think Siri would also get into trouble with more complex requests (but there's plenty of value in it before that limitation comes into play)
†this neologism could well go the way of the horse-less carriage
Saying that, it's actually appreciated, and does a great job at attempting to establish an emotional tie with the user.
I also really like that the story proves you to be thoroughly correct about Apple trying to establish emotional ties. Here's hoping that they made that a part of their "corporate DNA".
Jeez Gruber, make up your mind :P
That said, I'm really looking forward to the additional speed in the 4S and Siri.
As Mr. Gruber stated, once he got it into his workflow, he missed it when it was gone. However, I think one of the biggest hurdles to mass adoption will be getting it into the workflow of the average user.
Anyone have ideas of how to address this problem? Or do you think the product video is enough?
However, your point is well taken. Maybe I'm not giving the average user enough credit.
Maybe they wanted to keep the interaction short to give it more flow? Siri doesn't really ever seem to read very long stuff to you.
Apple said the processor in the iPhone 4S is the A5. This may be the exact same chip in the iPad 2, or it might be the same class of chip. Apple doesn't detail the internals of their processors, so they can make device specific versions and call them the same thing if they want. Thus, this A5 could have co-processors in it that didn't exist in the one for the iPad 2. Specifically, I remember there begin something about an image co-processor that the camera uses, and it seems logical that there might be some video encoding or at least video feature extraction being done in hardware to enable 1080p video. (a big step up from 720p).
There may well be a feature extraction engine in the A5 to turn the audio recording of your voice into a series of symbols that are much more efficient to send over the wire to the SIRI service.
But, even if there are no co-processors, voice recognition is a very CPU intensive process, and the A5 has another CPU in it, as well as a greatly improved GPU. Either or both may be required for the current version of SIRI (which is more complete than the original) to work.
Thus it literally may be the case that Siri wouldn't run on the iPhone 4.
Schiller referred to it as an "Image Signal Processor" so it may be that it is custom silicone for used only by the camera but it seems more likely that it is a more general DSP and was badged like this for the presentation.
It is probably a version of ARM's NEON technology:
http://www.arm.com/products/processors/technologies/neon.php
NEON technology can accelerate multimedia and signal
processing algorithms such as video encode/decode, 2D/3D
graphics, gaming, audio and speech processing, image
processing, telephony, and sound synthesis by at least
3x the performance of ARMv5 and at least 2x the
performance of ARMv6 SIMD
Update: According to wikipedia that's exactly what it is:
http://en.wikipedia.org/wiki/Apple_A5
The A5 contains a rendition of chip based upon the
dual-core ARM Cortex-A9 MPCore CPU with NEON SIMD
accelerator and a dual core PowerVR SGX543MP2 GPU.
http://www.anandtech.com/show/4216/apple-ipad-2-gpu-performa...
Also, the corresponding A4 page makes no mention of NEON or DSP:
http://en.wikipedia.org/wiki/Apple_A4
Interestingly, as this is the iPad A5 being discussed it strengthens the argument for Siri-on-iPad2 (assuming the signal processor is key to Siri working efficiently).
NEON is included in the A4's Cortex A8. In fact, NEON was added back in the 3GS: http://wanderingcoder.net/2010/06/02/intro-neon/
Only the 2G and the 3G don't have it.
So does the iPhone 4S have a different kind of DSP, too?
I have no problem with Apple purposefully limiting Siri to the new iPhone and I don't really understand people who think that because the hardware supports it they should get it.
It would be nice but it's Apple's choice.
Talking to my phone is nice, and so is a fast camera, but it's tragic that this beautiful sleek device has to be wrapped in an ugly rubber band for me to make a call!
[Edit] Well, this article says it's been fixed with two antennas. I really don't like how they downplay the problem and act like it only affects ham-fisted gorillas. When I have just two bars, griping the bare metal with four fingers is enough to make the signal too weak for communication.