It's ten years since Apple introduced Siri as a beta service, which means a decade of talking to our iPhone, then Apple Watch, and HomePod.
The more you know about voice recognition, or even just the more you think about it, the more incredible Siri is today — and astounding it was on its announcement, ten years ago. On October 4, 2011, Phil Schiller called it "the coolest feature of the new iPhone 4S."
Unfortunately, Schiller also inadvertently described the future of Siri even as he believed he was commenting on previous attempts to do voice recognition.
"Because for decades, technologists have teased us with this dream that you're gonna be able to talk to technology and they'll do things for us," he said. "Haven't we seen this before? Over and over, but it never comes true."
"We have very limited capability," he continued. "Just learn a syntax, call a name, dial a number, play a song. It is such a letdown."
Mind you, he wasn't the only one to say something at the Siri launch that sounds a little different in retrospect. "When I leave Apple it'll put up an alert," said Scott Forstall in the demo, "reminding me to call my wife."
Siri really is amazing
Today we're used to Siri and Alexa, and we may remember such digital voice assistants as Cortana, but we know them so well that we can forget how remarkable they all are. When you ask Siri something, it sends that audio across the world to a server where it is parsed, or just plain ripped apart, to determine what words — and what language — you said.
Then that's compared to what you said a moment ago, because you may be continuing a kind of conversation. And then the instruction is sent back to your phone so that Siri can both do whatever you've asked, and tell you the result.
This has changed only very recently, with Apple introducing on-device Siri which can sometimes skip that business of sending across the world.
Nonetheless, this is a series of steps that are difficult to do, and difficult to do quickly, but Siri does it in excess of 25 billion times a year.
It's easy to forget how much it is doing, in every sense, and it's also hard to remember that it's doing all of this because it goes wrong so often. Sometimes that's because a kind of auto-correct type of failure has hit the translation and it thinks you've asked something you haven't.
But it can also be that Siri fails to do what it says it's done, or rather what you think it says it's done. Ask Siri on your Apple Watch to set an alarm for 6pm and it will commonly complain saying that there isn't an alarm for 6pm.
When you've finished explaining in words with multiple asterisks that this is why you were asking it to set one, it might do it. Siri might actually set a 6pm alarm and tell you it has.
But it won't necessarily have then turned that 6pm alarm on.
In the quite incredible work that Siri does, that feels like the kind of low-hanging fruit that should be easy to fix. It's not about parsing difficult sentences, it's not about picking a result out of a best-guess voice-analysis, it's not doing the thing it says it's doing.
Calm down, take a deep breath
This is Siri's tenth birthday and we wouldn't have it disappear, we don't fail to appreciate what a marvel it is. When you can just call out to thin air, "Hey, Siri, remind me to pay my credit card when I get home," it is a truly amazing thing.
It just does lose some of that amazement when right after your Apple Watch says "Done," so does your iPhone. And your HomePod.
It's less than amazing when you're in the kitchen and ask your iPhone to set a timer, and it's your HomePod mini with bionic ears in the den that does it, right before someone shuts the door to that room.
That aspect is extra frustrating because really it should be an example of Apple's absolute brilliance in sweating the details. When you call out "Hey, Siri," every single device you own that is capable of hearing you, listens.
They all then rapid-fire talk with each other to calculate which device you meant. If you've raised your Watch arm, while your iPhone is in your pocket and the HomePod is in the den, it's your Watch that knows to respond.
But then it doesn't.
There must surely, easily, be situations where it's not possible to correctly predict which device you wanted. But there certainly cannot be any intra-device communication that results in two devices thinking you meant them.
Yet that happens. The Watch on your wrist, the iPhone in your back pocket. The iPhone in your hand, and the HomePod on your desk, two devices can both react, both act, on what you say.
And then sometimes, just sometimes, nothing responds. That is equally frustrating — but it is also when you realise just how much you actually rely on Siri.
We don't come to bury Siri
You're deep in the middle of a task and you remember you need to put the oven on to pre-heat for tonight's dinner. No one ever remembers that at the moment they need to do it, but now you can just call aloud and Siri will remember it for you, Siri will tell you when it's time.
Or you've got your arms full when you walk up to your front door. "Hey, Siri, unlock the front door," and if you've got a Smart Lock, it does. "Turn off the downstairs lights." Done.
What time is it in New York? Easy. "Hey, Siri, what's 29 times 3?"
Or you're listening to music through AirPods and you can say, "Hey, Siri, play Francisca Valenzuela." With the gentlest of dips in the music volume, Siri listens, and then plays what you want.
Apple devices are better with Siri.
Siri has the weather down cold, you can ask it about that any time. You can also tell it to take you home — and it will open Apple Maps with the route to your house right there.
And Apple has got the tone right with Siri. It doesn't pretend to shout at you like the Carrot Weather app, but it also isn't robotic.
So sometimes Siri will tell you about Apple launches a little early. Ask Siri its favorite color and it might dodge the question, but ask you yours.
Or if - when - you have occasion to use colorful language with Siri, it may reply "I'm not responding to that."
It's not "following along the conversation just like a human does," as Forstall said later in his demo. But it's more than straight answer-and-respond.
Somehow it is enough of a personality that if you ever hear someone using Siri with one of its alternative voices, it sounds wrong to you.
So if Siri is this brilliant thing that is chiefly so brilliant that we can tell how much better it could be, it is still something remarkable. And ten years of it is worth celebrating.
"Hey, Siri, when's your birthday?"
"Surprise!" says Siri, if you ask on October 4, 2021. "I'm ten today. I tried to organize my own surprise birthday party, but I knew it was coming, so I thought I'd surprise you instead."
44 Comments
Let's not beat around the bush here. Siri might have been brilliant 10 years ago, but 10 years is a very long time in tech: today, Siri is far from amazing. It's not materially improved since introduction, once a year Apple adds a couple of new phrases but comprehension remains that of a 2 year old. Alexa, which launched 18 months after Siri, has "intelligence" that is 10x that of Siri, thousands of devices that support it plus "skills" for everything. Siri on the other hand has a very fixed feature set, aside from some restricted actions you can use with shortcuts - it is way too basic compared to the competition.
As well as that, its comprehension is terrible. A typical example I use is asking Siri to add 5 minutes to a timer. Siri changes the timer to 5 minutes. Similar lack of comprehension with lists too. With Alexa, - as long as the skill is installed - I can specify the exact item I want from the grocery store, I can ask the price, and I can add it to the basket. Next time I go on the grocery store's site, it'll be waiting in the basket. Siri can't even get adding things to a list right. "Add tomato sauce to my shopping list" regularly results in two items, "tomato" and "sauce". These kinds of bugs have existed since at least 2016, and yet apparently "Siri really is amazing"? Don't think so.
Aside from Siri itself, I don't think I've ever heard anyone other than me ever say "hey Siri" unless it's at my direction. Certainly never heard or seen anyone use Siri in public, ever. In fact most people I know have it switched off. Also, Siri's crappiness is a key reason the HomePod flopped.
Am I disappointed that a decade later Siri is still pretty dumb? Sure. But do I believe it’s a unique shortcoming of Apple and their developers? Nah. AI has long been a TV & film fantasy, and the real thing is much less impressive. That nobody has cracked it or has K.I.T.T. in their smartphone just proves the point. The problem space is difficult and we have a long, long way to go before we get to the computer on Enterprise. In our pockets.
Regarding the regular "Hey Siri" confusion, it'd be much more sense if you could name devices. So "Hey iPhone", or "hey watch". Alternatively, for many things, it wouldn't matter which devices actually listened. Sending a text? Probably best for the device with a screen, but what's wrong with a HomePod being the microphone, and the text showing on the iPhone or Watch for the person speaking? All devices could listen, use a convolution algorithm to improve the signal to noise ratio of the audio and then present the result on your iPhone. For audio requests, without airpods it's very likely you'd want a Homepod to respond, so if the iPhone picks up the "hey siri", route the result to the HP automatically. There's so much Apple could do with Siri, but just doesn't.