• DreamHost

  • The Siri Follies

    March 15th, 2018

    So what’s wrong with Siri? Why hasn’t it realized its potential, since it was essentially the original smartphone digital assistant? Why has it been rated far inferior to Google Voice and Alexa?

    A key reason is said to  be Apple’s privacy policies. Unlike Google and Amazon, you aren’t the customer, thus ensuring far greater security when you use Siri compared to the other two. By ceding some of your privacy, you will supposedly get a digital assistant that’s more responsive and more accurate. This is a tried and true argument to explain at least some of Siri’s shortcomings.

    But there is a published report from The Information that the causes are rather more complicated, the result of an unusual amount of palace intrigue involving infighting and turf wars during its development process. Quoting unnamed former Apple employees who worked on the Siri team, the article asserts that they “rushed Siri into the iPhone 4s before the technology was fully baked.”

    True, Siri was marked beta at the time, but some might have regarded it barely alpha level. Despite the reality, Apple made a huge deal of the great things Siri could do, and ran some expensive ad campaigns demonstrating its purported abilities. I recall one featuring actor Samuel L. Jackson, whose strong presence would surely convince any human to listen to him. Siri did his bidding too, but even my best imitation of the busy actor failed to deliver comparable results.

    In any case, the ex-employees obviously cannot identify themselves due to corporate NDAs, which means that we have to take such claims on faith, although it seems to make sense, at least on the surface.

    So who was at fault? Well, evidently the late CEO, Steve Jobs, because his advocates allegedly wanted Siri to follow his vision even after he was no longer with us. But that doesn’t reflect so well on Tim Cook’s leadership.

    When you look at the names of the executives who managed the team, you run across familiar names, headed by ex-iOS chief Scott Forstall and Richard Williamson, who lead the failed Apple Maps project. The article asserts that working on both Siri and Maps may have been too much for them.

    Regardless, both executives are long gone but shouldn’t that have meant that things would change after they were given their walking papers?

    Many of the original Siri employees that joined Apple are also supposedly gone. Some are working on Samsung’s Bixby personal assistant, but it doesn’t seem that they learned much from the Apple experience, since it’s no great shakes either.

    Certainly you can see evidence of a blame game. The piece quotes an email from Williamson, where he responds, “After launch, Siri was a disaster. It was slow, when it worked at all. The software was riddled with serious bugs. Those problems lie entirely with the original Siri team, certainly not me.”

    And if Siri’s employees couldn’t get it together, how does one suppose that the project could ever coalesce into a reliable voice assistant?

    It’s also claimed that Siri’s implementation in HomePod is underpowered because the Siri crew didn’t learn about it until 2015, after Amazon released the Eco. This was supposedly due to the alleged culture of secrecy that sometimes prevents design teams from communicating with one another in a timely fashion. But that, too, is just a repetition of the usual mantra about Apple.

    It is true that Apple didn’t release a SiriKit, to allow third-party developers to use it to enhance their apps, until 2016. Worse, the number of functions was limited to just 10 at launch.

    True or exaggerated, Apple clearly has taken notice, issuing a response claiming that Siri is “the world’s most popular voice assistant.”

    Apple claims that, “We have made significant advances in Siri performance, scalability and reliability and have applied the latest machine learning techniques to create a more natural voice and more proactive features. We continue to invest deeply in machine learning and artificial intelligence to continually improve the quality of answers Siri provides and the breadth of questions Siri can respond to.”

    There’s little doubt about Siri’s limitations with the HomePod, however. But that’s something that can probably be addressed with software updates. Apple can also add adjustable EQ and other features that would help deal with the HomePod’s other perceived flaws.

    But has machine learning and the improvements in Siri’s voice reduced its shortcomings any? That’s hard to say. Early testing of the new Siri against Google Voice and Alexa last fall didn’t indicate much of an improvement. But that doesn’t mean it hasn’t received a steady stream of enhancements since then, although the sparse release notes for the iOS 11 and macOS High Sierra updates don’t mention any changes.

    Now I suppose it’s possible there is some level of sour grapes involved in these complaints about Siri, although its limitations are obvious. It’s curious that few or any of the former Siri employees appear to have done any better since leaving Apple, so maybe the wrong technology was acquired. That said, services are increasingly important to Apple, and if Siri is fixable (I think it is!), one hopes that the proper level of effort is being made to make it more competitive to Google and Amazon.

    After all, Siri was there first, and it’s discouraging to see it fall way behind regardless of the reasons.

    | Print This Post Print This Post

    6 Responses to “The Siri Follies”

    1. dfs says:

      I think the problem lies even deeper: Siri is not very good at voice recognition. Example of what I mean: a while ago I was driving around, headed for a restaurant called the Eden where I had a table reserved, and encountered such heavy traffic that I knew I’d be late for the reservation and was afraid I’d lose the table. So I wanted to phone them to tell them I was running late but that I would show up a. s. a. p. Okay, six times I asked Siri to dial the Eden. Each of the six times it failed to understand me.

      Now, I am not the victim of any serious speech impediment so I think it was reasonablet to expect that Siri should have been able to understand me. Problem is, with a centralized system (as opposed to a local v. r. program residing on my Mac) there’s no real learning capacity, so Siri can’t adjust to my personal speech patterns and whatever idiosyncracies I might have (regional accent and so forth). After a good number of years Siri and I are on no better terms than the day I unpacked my first iPhone.

      There’s an easy solution to a good deal of this problem. Suppose I had tried to say “Eden” those six times and had failed to be understood. So then a dialogue like this could follow:

      SIRI. Sorry, I can’t understand you.
      ME Well then, SPELLIT (or some similar trigger word)
      SIRI Okay, SPELLIT
      ME (slowly and distinctly) E – D – E – N. End SPELLIT.
      SIRI Okay, I understand you now. Dialing E – D – E – N (Sound of phone being dialed…)

      I’m amazed Siri hasn’t been given this ability. Think of all the frustration it would eliminate! People might actually begin to like working with — as opposed to struggling against — Siri.

    2. Peter says:

      True or exaggerated, Apple clearly has taken notice, issuing a response claiming that Siri is “the world’s most popular voice assistant.”

      Hey, remember when AppleTalk was the world’s most popular network? Every Mac shipped had AppleTalk, even if they weren’t using it. And most of the “networks” had two nodes–a Mac and a LaserWriter.

      I’m sure this is somewhat the same. Yes, it’s the most popular because it’s been on every phone that Apple has sold since 2011. Now whether it’s the most used voice assistant, I don’t know…

    Leave Your Comment