Siri and the Illusion of Failure

June 7th, 2017

You may not remember this, but Siri debuted on the iPhone 4s on October 4, 2011, the day before Steve Jobs died. While it became a sort of cultural icon right away, it wasn’t necessarily all that accurate. All right, Siri would certainly pay attention to actor Samuel L. Jackson in a series of TV ads, but regular people had far more difficulty dealing with Apple’s digital voice assistant in the real world.

While it wasn’t obvious, Siri was actually labeled as a beta in those days; it’s not now. Getting accurate responses was always troublesome, even after Apple improved Siri, and added more features. Supposedly it can handle some context-related requests now. Supposedly.

Every day, my wife sets a daily alarm with Siri on her iPad. While she could create one alarm to fire every day, or just do it manually, she’s inclined to just get up at the appointed time, grab her iPad and set a new alarm for another half hour of snooze time.

Barbara has a clear, distinct, young-sounding voice with a trace of a Brooklyn accent. No self-respecting digital assistant should have difficulty responding to such a simple, basic request. She keeps her iPad up to date, and it will be capable of running iOS 11 since it’s a 64-bit model. But more often than not, she has to repeat the request a time or two before it’s correctly recognized and stored.

With 25 years of broadcast experience, I would like to think most people can understand me without difficulty. Siri? That depends.

Both customers and critics will tell you that Google Assistant is more accurate, and can do more things well. One reason is supposedly due to the fact that Google collects more data about you and other users, and is thus able to better contextualize user requests and provide meaningful answers.

Apple has opted for privacy, and it is thus presumed that Siri may be the safer alternative, but it’s not quite as flexible.

A few tests I consulted, such as this one, demonstrate that Google is capable of trouncing Siri with more responsive answers in some tests. Your mileage may vary, but I’ve made it clear where I stand. Siri was a cool idea when it launched in 2011, but it needs work.

Or did.

According to Apple, Siri is getting a major upgrade for the forthcoming iOS 11, which is due to ship this fall. Its voice will sound more natural, closer to a human voice than a digital being. Its responses will be accompanied by a more visual interface with suggestions and followup information. It will also use machine learning, on the device itself, to better predict your needs. While what Siri learns can be synced among your own devices, the data is otherwise private. There will also be expanded support for third-party developers to add Siri to their apps.

That, to me, sounds like a good thing. It has the potential of fixing some or most of Siri’s ills in accurately responding to your requests, right?

Well, an article in a certain major daily newspaper wants to label the new and improved Siri a huge fail. It quotes one developer as saying, “nothing changed.” But that’s absolutely untrue, based on what Apple demonstrated during the WWDC keynote and, reportedly, during work sessions with developers.

So why the disconnect?

Maybe the writer of that article wanted a juicy quote, and that’s certainly juicy if inaccurate.

But it’s also true that Apple released the first beta version of iOS 11 to developers after the Monday keynote. So millions of developers can now put the new Siri through its paces and test it against the Google Assistant, Amazon’s Alexa and Microsoft’s Cortana.

True, it’s an early beta being compared to final versions of the three competing digital assistants. But such a test ought to give an inkling of the scope of Apple’s improvements, and their potential. Sure, there are apt to be flaws some three months ahead of iOS 11’s final release candidate.

Ahead of such a test, saying that “nothing changed” makes little sense except for its sensational quality. Even if the only meaningful change is a more natural voice, something has changed. Is it more accurate, can Siri be held against all comers and hold its own?

A test of the beta would at least reveal the new Siri’s potential. When I begin to run the prerelease, I’ll focus primarily on the sort of information I want from Siri to see how well it does. Tests of this sort can be very subjective. It may even be possible that some of the published comparisons between Siri and Google Assistant — and other digital assistants — are being weighted to emphasize the areas where one contender routinely does better than the others, and ignore the areas where it’s worse. Does that convey  an accurate picture?

That Apple is making a big deal about improving Siri surely indicates they understand that it needs to get better. Unfortunately, some of the tech press is relentless. Once a conclusion is reached, it isn’t changed even if circumstances change.

So it’s still a supposed conventional wisdom that Google Maps is better than Apple Maps. Maybe it still is, even though the later has consistently improved. But I’ve used Google extensively in recent weeks in connection with one of my side gigs, and it can sometimes fail bigly in significant ways. Unfortunately, Apple’s alternative doesn’t work with the app I use for that gig, so I can’t really make a direct comparison.

In any case, I’m curious to read some updated comparisons to see where Siri for iOS 11 stands in relation to the rest of the pack. Unlike a reporter for a certain newspaper, I won’t reach any conclusions until I know the facts.

| Print This Article Print This Article

4 Responses to “Siri and the Illusion of Failure”

  1. dfs says:

    The other day I was driving around trying to find a restaurant called the Eden. Sounds simple, right? Well, it wasn’t. Using CarPlay, six times I asked Siri to locate it for me, and six times Siri failed to understand “Eden.” It struck me that there should have been a dialogue which went something like this:

    ME Find the Eden.
    SIRI Huh?
    ME Spellit. E – D – E – N. Endspellit.
    SIRI Right, got it. Here ya go…

    Maybe the good people at Cupertino could give me a reason why this wouldn’t work, but it seems to me that implementing this one additional feature would do away with much of the frustration and heartache involved in dealing with Siri.

  2. Charles Koehl says:

    Perhaps the developer who said “nothing changed” didn’t use it long enough for it to learn how to perceive the context of their respective questions; however, if that were true, and most people would dismiss Siri before it got a chance to become better for them.

  3. dfs says:

    What’s particularly irritating about this developer is that Apple is at least claiming that the new Siri is going to have the capacity to adapt to the individual user’s vocal habits at the level of the that individual’s own devices. So far, the worst feature of Siri was precisely its inability to do this kind of adaptive learning. To me, this is potentially the most interesting new Siri feature, and when I finally get a chance to test it that’s the one to which I’m going to be paying the most attention. But of course I won’t expect magic results from Day One: adaptive learning takes time. So how could a developer possibly give it a fair test on the basis of a few hands-on minutes?

  4. Tarik says:

    I think people have been brought up to believe we live in an instant society. I’ve had my days when Siri and I have a fight. Then I remember, it’s a machine made by imperfect people.

    I’m willing to wait until the full release before I put forth any final decisions.

Leave Your Comment