Siri has been embarrassing me since she first appeared on my iPhone 4S. Showing off the feature at a dinner with friends on that first weekend of its release, she choked on the simplest query. Apple servers were slammed with people around the world doing exactly the same show-off maneuver simultaneously. And I learned my lesson. For most civilians outside the tech and media bubble, nothing quite satisfies that schadenfreude urge as an Apple geek epic fail.
“Screw you, Siri.”
“!” she replies. Turns out she actually stops talking to you if you start insulting her. For all I know, the platform is learning about my bad attitude. In a world increasingly dependent on rules-based robotics, you don’t want to alienate an algorithm. What if Siri decided to get snippy with me during a request for directions?
American culture has had its way with Siri in the last few years. Our ambivalence with automation, the gadgetry we otherwise crave, come out at moments like these. Like celebrity, gadget lust is both evident in our behavior and reviled in our rhetoric. Siri is icon and joke, all in the same cultural sentence. That is how America rolls. Siri has become such a familiar voice that my wife objected when I used the new iOS 7 option of turning on the male version. “That is not Siri,” she complained. Apparently a gentleman’s gentleman makes her more uncomfortable than having a female automaton serving her husband. Go figure.
But it turns out that like a lot of cultural phenomena, more people know about and joke about Siri than actually use it. According to a commissioned survey of over 2000 mobile users by Intelligent Voice, part of the UK Chase Information Technology Service for transactional processing, only 15.2% of Americans had used Siri in her latest non-beta iteration on iOS 7.
I have to admit I use Siri on a regular basis only as a shortcut to setting a timer. Otherwise my main use of voice commands on my phone is for transcribing text messages faster. As a virtual assistant I gave up on Siri a long time ago. I not only find the accuracy and effectiveness too hit-and-miss, but I am not drawn to the human-like relationship it is supposed to engender toward the machine itself. I never dreamed of robotic personal assistants, or real life valets, for that matter.
But even more to the point, Siri is not impressing those who do use it. Once the survey winnowed out non-users, the 150 respondents left were still underwhelmed with the service. Only 12.7% find Siri “extremely accurate” in understanding them. Still, 43.3% say Siri is “quite accurate.” Put the two responses together and I guess Siri gets a fair vote of confidence. A good third of users find the results, like me, hit-and-miss. Only a small percentage find Siri altogether bad. But you have to figure those users gave up a while ago anyway.
When asked if Apple “oversold” the voice command feature on iOS, only a bare majority of 54% disagreed. The analysts at Intelligent Voice asked the question because of their concern that a good technology is sullied by unrealistic expectations. “All told, not great news for Apple, but probably worse news for the voice recognition industry,” the site states. “We sell it like it’s magic, but as I’ve read elsewhere, it’s still an emerging technology that’s been emerging for 30 years. The sooner we are honest with people about what can really be achieved, the more chance we have of adoption.”
I have been reviewing voice command software for over 15 years -- from the earliest days of Dragon. And I actually rely on voice-to-text for some writing on a daily basis. It is still not ready. And as an interface, arguably we are not ready for it. Voice still suffers from a fundamental problem -- people don’t want to talk at their machines in public. Voice is an impractical interface in most cubicle environments just because of the awkwardness of the out-loud phrasing and the sheer cacophony of voices. I have yet to see someone use Siri in a public setting, I guess largely because most of us are too self-conscious to do so.
The idea of the amiable Android valet goes back into the first decades of Sci-fi pulp fantasy in the 1930s and 40s. It was a cultural byproduct of mechanization itself because it embodied the ways in which the machine age was replacing aspects of human toil with machines. The dutiful robot with nefarious aims was a trope for Americans who are unsure whether mechanization was challenging their human worth or helping to free it to a next order of being beyond toil. This is an unresolved question, but it is interesting that the Android theme re-emerges at just that moment when the technology has evolved from utilitarian to intimate. Voice interfaces may be a long way off, because of practical and social restraints. But the larger issue may be: what are we really talking about when we talk about Siri?
I tried Siri for about 5 minutes. Got very few useful responses. Some mistakes were so bad that I was genuinely worried that Siri would misunderstand something and delete my data or text garbage to friends, so I gave up. Google Voice works much better BTW.
The reaction to speech recognition like Siri seems often like a "how dare a computer act like a human!?"--More emotional than rational. Why not complain about how often touch doesn't work when precision is required? There are two factors that are often ignored: (1) speech recognition in these systems is adaptive--it gets better for an individual as you use it; and (2) cooperation by the user (speaking concisely and clearly) affects the result, just as your willingness to touch again (as opposed to just giving up) makes touch screens work. Speech and natural language understanding will help us deal with "digital overload"--I like the user manual ("Just tell me what you want") and this aspect of understanding and providing a quick answer or action will increase over time.