Siri is a little over a year old, and by now her failings are well known. Like Crystal Pepsi and Rick Perry’s presidential run, Apple’s digital assistant was delivered to us on a magic carpet of hype, promising to change everything about everything. “For decades technologists have teased us with this dream that you’ll be able to talk to technology and it’ll do things for us,” Phil Schiller, Apple’s chief marketer, declared as he unveiled Siri to the world last October. Now, he suggested, that dream had come true. Siri would finally let you talk to computers the way you talk to people—naturally, without needing to memorize commands or syntax—and then she would help you the way a real assistant might.
Siri let us down immediately. Her problems are manifold, but they all add up to something that would doom any assistant: She is profoundly unreliable. Yes, sometimes, depending on how you speak and what you ask, Siri will get exactly what you’re saying and deliver the correct answer. Those moments are delightful. They’re also rare.
A lot of times, Siri won’t have a clue what you’re asking, and when that happens the conversation is apt to go off in wild directions. “What was the Notre Dame score?” my colleague’s 6-year-old son recently asked Siri. She heard, “Where is a porn store?” and then, quite helpfully, recommended the Pure Romance Warehouse.
Even when Siri does understand, she doesn’t understand. Here’s a conversation I just had with her.
“When is Skyfall playing?”
Siri responds with a list of movie times. Not bad.
“When is Lincoln playing?” I ask.
“Here’s Skyfall playing quite far from Lincoln Village today,” she responds, showing me a list of screenings near Lincoln Village, a town I’ve never heard of.
I try again, this time shouting the word Lincoln as if he’d emancipated me personally: “When is Lincoln playing?!”
Something about my attitude sends Siri over the edge, and she responds with the digital equivalent of frothing at the mouth:
This sort of thing doesn’t happen every time, but it happens just enough to render Siri unusable. When you take into account her less-than-stellar hit rate, Siri is an assistant only Zooey Deschanel could love. If you’re so meteorologically challenged that, upon hearing the unmistakable crack of a storm while staring out a rain-soaked window, you still need to coquettishly inquire, “Is that rain?” then Siri is totally for you. For normal people, Siri hovers between a gimmick and tease.
But as I said, all this has been chronicled before. (See Mat Honan’s definitive piece, “Siri is Apple’s Broken Promise.”) What’s new, now, is that Siri finally has serious competition on the iPhone. Last month Google added a voice search feature to its iOS app. (The feature has been available on Android for years.) After seeing it trounce Siri in head-to-head competitions, I decided to give Google’s app a try for myself.
The bad news is I didn’t find it magical—while Google’s voice feature understood my queries more often than Siri did, it still made several mistakes, and it often failed to give me useful answers. There were even a few times when Google failed while Siri excelled. The first time I asked Google, “When is Skyfall playing?” it thought I’d said, “When is sky fall plane?” I repeated the question and it did the same thing again. A few hours later, I came back and asked the Skyfall question once more. This time, mysteriously, it nailed it.
But if Google’s voice search isn’t perfect, it is truly useful. Most of the time it understood what I was asking and gave me a good-enough answer to my question. In that way, it’s groundbreaking. Google Voice Search isn’t just better than Siri or any other voice recognition system I’ve used. It also approached a threshold that transformed it from a novelty act into something I could imagine relying on in my daily life. With Voice Search, Google is beginning to make good on Apple’s broken promise.
Google achieves this in a few ways. First, it lowers the bar for success. Google’s voice feature is called Voice Search because, unlike Siri, it doesn’t promise to become your robo-secretary. Google Voice Search will not make appointments for you. What it will do is answer questions about the world—when you need anything that you might otherwise find with Google’s Web search, you can use Google Voice Search if you so choose. These limitations aren’t exactly by design; on the iPhone, Apple’s restrictions in third-party apps make it technically difficult for Google to do everything Siri does. But they work out in Google’s favor: This app meets and sometimes exceeds your limited expectations.
Second, Google’s user interface is superior. It’s screamingly fast—Google begins decoding my question as I’m speaking it, so it’s ready to present me with an answer just a split second after I’m done. By contrast, Siri takes one or two agonizing seconds to understand my question and to find an answer. More surprising is that Siri, which is made by a company that has a reputation for caring about superficial things, doesn’t sound very good. Her every utterance is tinged by a robotic, low-def twang. Google’s voice sounds like a real person. (Google doesn’t give her a name, but I picture the voice belonging to a young woman, smart as heck, with short brown hair, nerd-chic glasses, and no patience for Zooey Deschanel.)
But the best thing about Google Voice Search is that she’s overflowing with knowledge. Many times she’ll answer your questions with exactly the right answer. Other times she won’t speak but will at least give you a search page full of answers, almost always correct ones. That’s better than Siri’s way of coping with her own ignorance—she’ll either apologetically explain that she doesn’t know, or she’ll sometimes ask you if you’d like her to search the Web for you, which is a stupid question. (She should just search the Web if she has no better answer.)
Here is a list of questions that Google Voice Search correctly answered and Siri botched:
- Who is Apple’s CEO?
- Who is Google’s CEO?
- Who is Mitt Romney’s wife?
- Who’s the manager of the Yankees?
- How long is the Golden Gate Bridge?
- When did Disneyland open?
- Who founded Twitter?
You’ll notice a theme here: They’re all looking for specific factual answers. Siri can answer such queries too, but its expertise is limited. That’s due to a fundamental difference in how the two systems work. Siri is a “curated” experience—it draws its answers from specific sources that Apple has signed deals with. For instance, if you ask Siri about Barack Obama’s age, she’ll give you an answer from Wolfram Alpha.
Google Voice Search is more flexible. It derives its expertise from the “Knowledge Graph,” which is Google’s database of 500 million things it knows about the world. Google’s database is dynamic—its facts are constantly being dredged up by Google’s Web crawlers. That’s why when I ask Google, “Who founded Instagram?” it understands my question—even though Instagram is a relatively new company, and its founder is far from a household name, there’s enough on the Web for Google’s Knowledge Graph to know that “Instagram” is a company, and that it was founded by Kevin Systrom. (Siri offers to search the Web for that question.)
Last week I met with Scott Huffman, one of the engineering directors on Google’s search team. Huffman explained that the Knowledge Graph is just one of four technologies underpinning Voice Search. The others are the firm’s expertise at processing “natural language queries,” its speech recognition system, and its “core rankings” algorithm—the system that decides the order of your search results. These four technologies there are all an outgrowth of Google’s search engine. They depend on collecting and analyzing data at a massive scale in order to teach computers to understand the world the way humans do. That process—deriving intelligence from information—is a core part of Google’s mission as a company. It is not a part of Apple’s. And that’s why Google’s voice app is destined to become your everyday assistant. You’ll use it for the same reason you use Google’s search engine: It knows everything.
I asked Huffman to describe Google’s long-term vision of voice search—what is it trying to build? He invoked the computer on Star Trek—something that you can talk to like you would another person, than can help guide you through a decision. “You would ask, ‘Hey Google, where should I have dinner?’ ” Huffman says. “And it might say, ‘Well, you seem to like Italian restaurants, so how about this one.’ And then you’ll say, ‘Well, I don’t want to go all the way over there—are there any new places that I haven’t been to yet?’ ” In the end you might get to the same restaurant that you’d find after a few minutes with Google’s Web search engine, but the voice interface would make the experience feel more natural, like you’re speaking with an intelligent being who knows you.
Google Voice Search isn’t close to realizing that vision, but it’s not impossibly far off either. Huffman points out that Google’s app can already hold very small conversations. It understands pronouns, so if you ask, “Who is Barack Obama?” and then ask, “Who is his wife?”, it knows that his refers to Obama. And most important, it gives you the correct answer.
I just tried the same set of queries with Siri. First, she correctly identified the president. But when I asked, “Who is his wife?” she shot back, “What is your wife’s name?” That’s not what I asked. Actually, it’s really, really far off. And there aren’t any signs that Apple’s voice assistant is going to get much closer any time soon.
