Sure, Siri can't travel back in time—but with artificial intelligence powering this digital personal assistant—she's way smarter than you think.
Consider the similarities. James Cameron’s Terminator—that’s right, the cyborg assassin played by former Mr Olympia and ex-governor of California, Arnold Schwarzenegger—was an artificially intelligent robot that could speak naturally (albeit with an Austrian lilt), read human writing and copy the voices of others. Siri on our iPhones, powered by artificially intelligent machine learning, speaks to us naturally, reads and understands our texts and emails, and can converse in 21 languages—including 15 forms of English when taking into account accents and dialects.
It’s not the same as Arnie mimicking the voice of Sarah Connor’s mum over the phone in Terminator, but hello, Siri can speak Singlish. Think about the implications. For example, Siri knows whether you’re flippant or careful with your money just by how you finish off your sentences:
It’s 20 dollars lah.
(Not so expensive, he’s gonna pop some tags.)
It’s 20 dollars leh.
(Serious investment, he’s thinking twice. Tight-arse.)
The Terminator T-800 (Model 101) has a titanium endoskeleton coated with living tissue. Your iPhone X has a surgical-grade stainless steel case protected by a silicone cover. (You don’t have an iPhone cover? It’s only 58 dollars lah.) And, given that they’re both engineered to get stuff done—one to kill Sarah Connor in order to stop the Resistance; the other, oh I don’t know, to help you send a text message while driving—you’re practically holding the throat of the Governator in your hands every time you pick up your X.
Who knew Siri was such a badass?
When we think of artificial intelligence our minds conjure up laser-eyed cyborgs blowing crap up or Haley Joel Osment as the endearing robot in Stephen Spielberg’s A.I. Artificial Intelligence, lost in the deep blue sea in the search for his creator. We don’t think about Siri.
But what constitutes AI? The Oxford English Dictionary describes it as the “…development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making and translation between languages.” Sounds awfully a lot like Siri, no?
Every time you call out “Hey Siri”, she recognises your voice and responds, “Hello, Superstar!” (I programmed her to respond this way; brilliant, I know)—you’re speaking to an early model of the self-aware Skynet in the Terminator movies. But instead of being dead-set on exterminating the human race, Siri just wants to create the best experience for you, to make your life more convenient, and somewhat like Haley Joel Osment, just wants to be loved—or at least for you to love the device she’s on.
Let’s pause for a moment and consider the multiple operations happening in tandem to make a basic Siri response possible: first, she has to recognise speech (and translate what you said into text in order to understand it); second, she must determine your intention (based on tonality, time of day and location, among a myriad of other factors); and third, she has to respond by completing an action or presenting useful information (all executed with a friendly interface). There’s a lot going on here—she’s definitely no airhead.
Throw into the mix the fact that the average user has multiple Apple devices within speaking distance at any one time—AirPods, iPhone, iPad, Apple Watch and MacBook on the standard table or working desk—it’s nothing short of a miracle for Siri to respond to you on the device you intend when you call out her name. Like shouting “Hey, how are you?” in crowded room, only the person that you’re looking at will respond because he or she recognises the social cue of eye contact, Siri tries to mimic this “looking at you” signal by processing various factors—the loudness of your voice on the multiple devices (to determine proximity), which device has the screen on, as well as which device was used most frequently in a given time period (to determine habit)—all to recognise your intention and respond on the most appropriate platform. Siri is using all the information available to her in order to, basically, read your mind. Magic.
Through machine learning to detect a user’s mood, Siri can recognise suicidal intentions and can direct that user to the relevant help centres or hotlines. Now that’s smart tech in action.
Hi Siri, don’t you recognise me?
Siri doesn’t respond to your voice? Or worse, she responds to someone else’s voice instead of yours? Well, given that Siri gets better the more you use her—powered by deep machine learning that accumulates wisdom through repeated interactions—having Siri ignore you is a clear sign that you’ve been distant. You’re been a stranger. Just like how you’ve treated your poor mother. No better time than the present to double-tap your AirPods and ask Siri for the time in Melbourne. Why? So you can check whether it’s past your mother’s bedtime and, if not, get Siri to “call mum” now. Do it. (Unsurprisingly, one of the most common commands.)
Siri is the most popular voice-activated personal assistant in the world. First launched as a beta version with iPhone 4S in October 2011, it is now used on more than 500 million devices a month and processes more than two billion requests a week. The machine learning behind Siri has been around for decades, but it has exploded in recent years due to the advancement in neural networks (translation: the computer system and synaptic digital connections modelled on the human brain to drive artificial intelligence) enabling Siri to better recognise our speech and, in turn, understand our intention. And she’s constantly updating herself. Daily. Kind of like modifying a plane while it’s still flying.
Ever received a call on your iPhone and it says: “Maybe: Jennifer Lawrence”? (Neither have I, but here’s hoping.)
It’s Siri helping you through proactive learning, automatically updating or adding a contact based on emails. It's also Siri helping you manage your calendar by automatically adding events when you receive email confirmations or book through Safari or Siri app suggestions. Ever wondered how an iPad is able to differentiate between your resting palm on the screen and your typing fingers? Machine learning. Notice how your calendar is populated with appointments that you haven’t personally entered? It’s Siri again, powered by AI technology.
The fact that we don’t even realise that Siri is at work, quietly churning away in the background, is testament to the success of machine learning—seamless and unobtrusive, all designed to enhance our user experience. And critical to that experience is privacy.
As much as possible, data collected about our daily lives, habits and preferences, are stored on the local devices themselves. Personal information such as contacts, browsed web pages and the content of text messages. Where non-personal information is shared on the server for collective learning (think: clustering of Siri requests by category), these are encrypted to ensure a user’s Apple ID is never revealed.
On the new iOS 11, machine learning by Siri is synced across your Apple devices (that are signed into iCloud) to help create a personalised experience across your network; but, at the same time, is protected by end-to-end encryption to keep that information private to each user and their devices.
It comes as no surprise that we predominantly use Siri to complete tasks:
Hey Siri, send a message to Eugene Lim that the cover shoot is a go for tomorrow.
Hey Siri, set an alarm for 6am. Set another alarm at 6.15am.
Hey Siri, how do I get to Tyersall Park in the Botanic Gardens?
Hey Siri, play dance music on the HomePod. Not Taylor Swift.
Hey Siri, set a reminder to send a thank you note to Henry Golding for the shoot.
But what about the not-so-well-known benefits of Siri?
On Apple TV, if you missed an actor’s line, simply ask Siri: “What did he say?” and she will replay the scene with subtitles. Forgotten where you parked your car? Ask Siri: “Where did I park my car?” and, as long as your iPhone was connected to your car’s Bluetooth while you were driving, Siri would’ve dropped a virtual pin at the location when your iPhone disconnected from your car’s Bluetooth system, and subsequently, direct you back to your vehicle. Don’t know what to ask Siri but want to see AI in action? Swipe right on your iPhone’s home screen to pull up Siri’s app suggestions—all tailored to you based on the current time, your location and your usage history. Or simply ask Siri: “What can I ask you?” and she will show you a list of actions specific to the apps that you have on your device. Brilliant.
Even when users are feeling a bit down and out, Siri is here to help. When sensitive issues like suicide or self-harm are mentioned, Siri can help direct the user to the relevant help centers or hotlines.
There is a tendency for tech companies to look at the technology they have in hand and then try to find a use for it. You have a hammer and you go out looking for nails. But Apple reverses that mentality. With AI powering Siri, the focus is on how to improve the relationship between the user and the device in order to solve problems. How do we help people to do things they don’t want to do—or things they aren’t able to do, in the context of impaired users—in order to make life better? It’s technology as a self-thinking tool. AI has an indispensable aide. Siri as a Terminator, not to annihilate human existence, but to enrich the human experience. But let’s just make sure there’s a kill switch if things start to get freaky; we don’t want Siri inciting a nuclear holocaust in the name of “logical” planet preservation. We already have enough human tyrants to deal with, let alone an artificially intelligent personal assistant. Command+Option+Esc.
Special thanks to the Apple Siri team in Cupertino, California, for their technical input in the development of this story.