REVIEW: Siri is a seriously good innovation - even in beta
By ANDY IHNATKO ai@andyi.com
The iPhone 4S’s personal assistant is far more than a simple voice-command feature; it’s an entire, natural-language interface to many of the iPhone’s most useful features.
The iPhone has supported a simple menu of voice commands for a couple of iterations now (“Call ‘Tim Holmes’, work”). But Apple offered nothing as pervasive as what you’d find on an Android phone. Android allows speech-to-text input anywhere you can type. And Search features are nicely wired for voice.
Siri, the iPhone 4S’ new “personal voice assistant” feature, is something quite new, and quite ambitious. It does incorporate speech-to-text and that simple feature works great. It ought to; Apple licensed Nuance’s text-to-speech engine, so it benefits from years and years of usage from millions and millions of users. Just tap the onscreen keyboard’s new microphone button, speak, and presto, you’ve got text. You couldn’t write an entire email with it all at once, but it works great for texts and other snippets of a sentence or two. And it’s accurate enough that you can confidently use it to fill in fields in a form.
The more you use Siri, the more accurate the voice-to-text becomes, but that’s all being managed behind the scenes. There’s no need to spend any time explicitly training it for your voice.
But to describe Siri as Apple’s way of catching up to Android’s speech-to-text and voice commands misses the point. It’s easy to wire up a phone with a set of canned voice commands but that approach is limiting. And, sometimes, baffling for the user. I initially had a tough time with the iPhone’s then-new voice commands back in 2009. When I asked my iPhone to play music, I’d phrase the request in a way that felt natural, instead of the way that the iPhone had been programmed to recognize, leading to lots and lots of misfires.
Siri overcomes that limitation. Siri applies a lot of effort to understand the context of what you’re asking, and then shrewdly translates that into the correct result. It’s also sophisticated enough to blow past a basic “command, execute; command, execute” system and can treat your commands like an ongoing conversation.
Siri starts listening whenever you hold down the Home button (or the talk button on your headset) for a second. It also automatically activates when your iPhone is awake and you simply hold it up to your ear, provided that you aren’t already in the middle of a phone call, of course (what an elegant touch).
“What’s the weather going to be like tomorrow?” you ask. Siri responds with tomorrow’s forecast. If you had asked for the temperature at noon, Siri would have shown you the 12-hour forecast. If you had asked “Will I need a jacket tomorrow?” it would have told you that tomorrow’s predicted low is 45 degrees.
Ask it a follow up. “How about Thursday?”
“The low will be 50 degrees,” Siri says.
“Will it be any warmer in Sri Lanka?”
“There’s some bad weather coming up for Colombo, Sri Lanka tomorrow.”
It’s a clever solution to the complexity of voice control. Voice-to-text as a simple substitute for typing limits the phone to simple point-and-shoot functions, like jumping to the next track in a playlist. Siri makes it easy to handle complicated scheduling features, for example.
“Schedule a meeting with Jeff tomorrow at 2 PM in the Hilton lobby.”
It’s impressive enough that Siri looks in your contact database, finds that there’s only one Jeff, and fills in the contact, place, and time details accurately. “OK, I set up your appointment for tomorrow,” Siri replies. “Shall I schedule it?”
(“Sure.” “OK.” “Yup.” “Yeah.” “Go ahead.” “Why not?” Any of these, and more, will confirm it.)
But if there’s more than one Jeff, Siri will read you a list of the Jeffs it knows about and ask you to choose one. If you forgot to mention a time, it’ll ask for that info. If you have a schedule conflict, it’ll warn you and ask for guidance. There’s no single “right procedure” for scheduling an appointment or checking the weather or setting up a reminder. Just talk like a normal person and Siri will respond like a normal person . . . only more competently, usually. Even when there are hiccups, they fall well within the realm of what you’d run into in normal conversation.
Siri is powerful enough to make one wonder if our texting-while-driving laws need updating. We can all agree that it’s ungodly dangerous and stupid to split your attention between the road ahead of you and the phone you’re thumb-tapping a message into. But how about tapping a button on your Bluetooth car speakerphone and saying “Tell my son I’m in traffic and I’ll be home about twenty minutes late”?
(The iPhone Contacts app allows you to define family relationships. Having just one son = automatically texting “I’m in traffic . . .” to that one contact’s mobile number.)
Fundamentally, Siri could be as transformative to the phone as the mouse-menus-windows paradigm was to the desktop computer. It’s perfectly tailored to the device it runs on. You use a desktop when you need to arrange flights, a hotel, and a rental car for a business trip to Dallas in two weeks. You use a phone when you’re walking through the airport after a delayed flight and trying to figure out how to rearrange your schedule.
The net effect is that you actually come to rely on Siri. Sure, during your first day with the iPhone 4S you’ll practically wear out the battery asking Siri questions on the meaning of life (Apple’s engineers have made sure that Siri is well-prepared for this sort of thing). But soon, it truly becomes one of your default modes of operation. You could check the forecast by tapping the “Weather” app and scrolling to the right location . . . but you just don’t want to. Ditto for reminders, which I never used to bother with but which I now use constantly. “Remind me to wash the car the next time I leave the house,” say I. The next day, a reminder pops up by the time I reach the end of the block.
“Set a timer for five minutes,” I told Siri last night, as I slid a Lean Cuisine into the microwave. It’s interesting to note that Apple has quietly introduced a brand new and completely invisible class of iPhone app. First the iPhone had webapps. Then, third-party native apps. Now, there are voice apps. When you set a timer, make an appointment, or ask for a stock price, you’re not manipulating the familiar iPhone apps using invisible fingers. You see brand new interface elements, unique to Siri. In some cases, you even see a brand-new design language unseen elsewhere in the whole iPhone experience.
The thought of additional voice apps showing up some day is tantalizing, particularly third-party ones. It’s not a crazy idea. Conventional iPhone apps followed a tightly-managed arc. First, iPhone users had a small batch of preinstalled Apple apps. Then, after a period of education, there came a period where Apple would allow developers to submit apps for an intense, persnickety, and mostly top-secret testing and certification scheme. Finally, trial and error had produced an understanding of the difference between the kinds of apps that enhanced the iPhone experience and the kinds that just peed in the pool, and the iPhone became an open-ish platform. Heavy accent on the “-ish,” of course.
Adding new voice apps to Siri’s suite of functions -- even by Apple -- is a frighteningly subtle problem. You know how your barbershop has six cutters, and there’s one that always does a spectacularly good job, there are two who are just competent, and the others can only be properly addressed as “barbers” when they’re being named in some sort of civil lawsuit? Multiple third-party voice apps would be like that. You’d have six navigation apps that each think it’s best-suited to run with the ball when you ask Siri “How do I drive to the Amtrak station from here?”
Even if Apple wanted to open Siri to third-party voice apps, it might be impossible to solve that problem elegantly. Siri isn’t a mere app launcher. (Literally. I’m genuinely surprised that I can’t ask it to “Launch Keynote”). Siri is, truly, a virtual personal assistant. So much so that you interact with it using the same part of your brain that fires up when you’re talking to a real person. You instinctively say “Please,” and when Siri asks you to provide additional information that you forgot to include, you find yourself saying “Sorry.”
You ask this person to do something. How it does it really isn’t your concern, and that part of the experience needs to be preserved. If it has to ask you “Would you like to to get your driving directions from Maps, Motion-X Drive, TomTom, G-Map, Navigon, or CoPilot Live?” Siri becomes something much less.
In just two weeks’ time, Siri has decreased my dependence on Google Search. Not even Microsoft, having pumped billions of dollars into creating an elegant and in some ways superior product, hasn’t achieved that. Siri is possibly the killer app of curated search.
And, God knows, Internet search is in desperate need of grownup supervision. There is a subclass of human who sees something that benefits humanity and then instinctively thinks “There must be a way I can ruin this for everyone.” Step forward and collect your well-earned kick to the groin, search-engine optimizers and content farmers. These are the idiots who create content designed to divert popular websearches to pages which are barely relevant but which contain affiliate ads. Google works hard and valiantly to make their search ranking systems hard to game, but still, I often have to click through to a second and even a third page before I find search results that take me to detailed information from trustworthy sources.
But even when your search results haven’t been hijacked, a flat Google search is never the fastest answer to a quick question. When I type “Who is the Green Lantern?” Google gives me 8 relevant results, headed by a link to the correct Wikipedia page. Click, scroll around, and there’s the answer.
When I ask Siri the same question, Siri says “Let me check that for you” and a moment later, I’m looking at a screen that tells me his civilian name is Hal Jordan, though he’s also been known as Parallax and The Spectre. He was born in Coast City; his parents are Martin and Jessica Jordan.
We’ll leave aside the fact that there’s a whole corps of Green Lanterns, and that Guy Gardner, John Stewart, or Alan Scott might have also been correct answers. This isn’t a Comicon trivia competition. This is tech journalism, dammit, and delivering correct information as an immediate response to a question is definitely the answer we’re looking for.
“What’s the volume of a sphere 18 inches in diameter?”
3054 cubic inches.
“What’s the population of London?”
7.753 million, as of 2009.
If I ask Siri “How much should each of four people kick in on a dinner check for $78.28, plus a twenty percent tip?” it hands me off to a standard websearch on that phrase, which is Siri’s polite way of saying “Whuuu?” But if you ask “How much is 78.28 times 1.2 divided by four?” Siri will respond with in “23.48.”
Behold, the awesome power of curated search. Is the user asking a question about her calendar? A question about the current weather? Is she setting a reminder? Siri ponders the nature of the user request and hands it off to the right bit of code. When it’s a request for information, the request is handed off to Wolfram Alpha, which defines itself as a “computational knowledge engine.” It’s available as standalone iOS and Android apps and can also be accessed online at WolframAlpha.com.
Siri uses Yelp! for help on local points of interest, which is why, when you tell Siri “I need a burrito . . . stat!”, it can provide you with a list of nearby restaurants whose reviews mention “Burritos” . . . complete with ratings. Yelp and Wolfram Alpha are the two search partners that Apple’s willing to talk about in connection with Siri, but it’s easy to imagine how naturally Siri can be expanded with new partnerships.
But why just imagine? You can see the full potential by looking on YouTube for demo videos from 2010, when Siri was a third-party app and the company that created it hadn’t yet been bought by Apple. The independent Siri app could do far more than the iPhone 4S version’s feature set. Saying “I want to see ‘A Very Harold And Kumar 3D Christmas’ in an IMAX theater” would have resulted in a list of theaters screening the film the way its auteurs intended, and Siri would even help you purchase tickets.
I’d love to be able to ask Siri “When does the next issue of Usagi Yojimbo come out?” or “What kind of drink can I make if all I have left is brandy, diet Yoo-Hoo, a tin of anchovies, and cupcake sprinkles?” Siri is definitely capable of much, much more than what it’s doing today.
And when Siri truly does finally get wired up in a glorious and ambitious way . . . will Google become a background player for mobile search? Search engines give me “search results.” Siri gives me, you know, “actual results.” Actual results are way, way more compelling than a page of links that I then have to process and evaluate manually, particularly in the context of mobile devices.
Siri is a great tool even when you’re using it for basic search. When Siri can’t match your query to one of its pre-wired functions, it hands your request to a search engine. You can also specify the type of search. “Search Wikipedia for ‘Don Knotts’” will return the relevant article .
It’s too bad that Siri doesn’t have even a basic feature for clarifying a websearch, though. “Search macworld.com for ‘MacBook Pro review’” seems like an easy enough request to parse out (just translate that to a Google search for “MacBook Pro review site:macworld.com”) The fact that Siri can’t do that indicates that Apple’s being conservative with their initial release of Siri.
Indeed, Siri is explicitly labeled as Beta software. Which blunts my one big complaint about it: it should do lots more. It should really be able to do damned-near anything.
Siri is like the iPhone’s touchscreen. It’s a point of user input. When I want to do something with the iPhone that only involves a few seconds of interaction, I instinctively want to do it with Siri and not through a touch app. Even when I’m at my desk and I need a quick answer, I find myself reaching to press and hold the Home button on my iPhone instead of just leaving my hands on my MacBook’s keyboard and command-tabbing over to a web browser.
It also seems like Apple has passed up plenty of obvious opportunities to integrate Siri more tightly into the iPhone experience. “Play the newest episode of The Bugle podcast,” I said, during my second day with the 4S. I was genuinely surprised when Siri said it couldn’t find any podcasts with the title “the newest episode of The Bugle” instead of looking for a podcast called “The Bugle” and then playing the most recently-released show.
Siri can control most aspects of the iTunes app (“Play anything by Stephen Sondheim”). Take it as a compliment, Siri, that I tend to think you can do anything, and do it well.
I mean, just a minute ago I had a thought, then I got excited . . . and then I bummed myself out. “Make a note,” I told Siri. “The password for the WiFi at the Broad Street Cafe is ‘green-leaf’.” Siri created a Note to that effect, as expected. Oh, how awesome would it be if, weeks later, I could ask Siri “What’s the password for the Broad Street Cafe wifi?” and Siri would find and display that note!
Siri can search the whole Internet, but it can’t search the user data of the phone. I mean, jeez. I can find that note by swiping into the iPhone’s search page and typing in a keyword manually, but you can’t even ask Siri “Search my notes for ‘Broad Street Cafe password’.” This disappoints and saddens me.
Oh, I don’t blame you, Siri. How could I ever be mad at you, Siri? I know you’re still in beta.
Siri’s functions are limited because a voice assistant is precisely the kind of feature that Apple can’t design from inside their usual Fortress Of Secrecy. Apple is paying close attention to how people are reacting to Siri and using the feature (stopping short of violating its own user privacy policies) and when Siri comes out of Beta, it’ll be armed with additional tools.
(What about privacy? Welp, anything you say has to be sent to Apple’s servers for speech-to-text processing, but in the end, Apple has no way of knowing that the user of your iPhone once asked “Where can I go to have a photorealistic back tattoo of Tony Danza lasered off?” The 4S does store certain data locally to enhance Siri’s performance, but you can clear all of that data out at any time by going into Settings and turning Siri off and then on again.)
Will Siri ever make an appearance on the iPad, or the desktop? It’s currently limited to the iPhone 4S, but not because the 4S has any Siri-specific hardware (such as a special signal processor or an optimized microphone). Apple claims that Siri requires the added oomph of the A5 processor. Okay, but a hacker has apparently spilled enough Satan-enhanced goat’s blood over a stock iPhone 4 to get Siri up and running on the device, so there’s that.
I reckon that there are two factors that went into Siri’s 4S-exclusivity. It’s not on the iPad or the desktop because Siri is an intimately iPhone function. Much of the time, you don’t even need to look down at your iPhone’s screen when you use Siri. You can leave your iPhone in your pocket and just tap your headset button. Or it’s a feature you use while you’re in motion, and can’t really stop to thumb-type. Siri seems like a looser fit for any other kind of device. Even the idea of taking one hand off of my iPad to activate a speech feature seems clumsy.
So why limit Siri to just the 4S model, instead of making it available to iPhone 4 users as well? Server capacity is probably a factor. It was a manageable problem when Siri was a third party app that few people had really heard about. But now, it’s the most lickable built-in function on 4,000,000 new iPhones, with millions more to be sold in the coming months. Server load, and the best way to manage it, is probably another metric that Apple will be following closely during Siri’s beta period.
In fact, Siri’s users logged their first significant data point last week, when the service became unavailable to many people for the better part of a day. Was it a network problem? Apple ain’t saying, but it’s a logical conclusion.
Siri’s reliance on a network connection for Absolutely Everything -- including simple speech-to-text input -- is . . . well, let’s be fair and call it a downer instead of a weakness. At least Siri doesn’t use much data. Even during my most aggressive testing, Siri would have burned less than 3 percent of a 2 gigabyte data cap over the course of a month if I’d kept up that kind of pace.
But that’s not the point. The point is that Siri is so damned useful for so damned many things that you quickly begin to absolutely count on it. Your keyboard doesn’t stop working when you’re on an airplane, or when you’re in a part of town with weak coverage, does it? So when you instinctively raise your iPhone to your face and say “Add ‘Dulce de Leche Pop-Tarts’ to my shopping list” and Siri replies “Sorry, I don’t have a network connection” instead of “Okay,” you feel disappointment.
Of course, when you’re away from 3G or Wifi you can still fall back on the iPhone’s classic Voice Control feature. But it’s no substitute for Siri. It’s as if the personal assistant who’s worked with you for five years has the day off, and you’re forced to deal with some random stiff from the temp agency who has no idea where the local FedEx office is, to say nothing of your preferred ratio of Diet Dr. Pepper to Dr. Pepper in your insulated desk mug.
If there’s one compelling reason to upgrade to the iPhone 4S, it’s Siri, for the same reasons why people wanted to upgrade from DOS to Windows, or swapped their desktop for a slim notebook. It’s not because of the natural and childish attraction to something new and shiny. It’s because imperceptibly, the relationship between users and technology has been shifting, and the movement of those tectonic plates isn’t noticeable until a new feature or a new concept comes along that brings new clarity and relevance to a technology that we would otherwise start drifting away from.
PDA and phone makers created one revolution when they created cheap, powerful computers that could be stored in a pocket when not in use. Apple seems to have led the way in a new shift: computers that are so powerful, they might never need to leave your pocket at all.
Comments Click here to view or make a comment