My love/hate relationship with Siri
Whether you use Siri, the Google Assistant, Google Now, Cortana, or Alexa, I think you will agree that digital voice assistants are simultaneously amazingly advanced and frustratingly useless. I am most familiar with Siri, but from what I can tell from friends who use other assistants, many of my complaints are applicable at different levels to all of the other assistants.
Let me first make clear how I do, and do not, use Siri. I use Siri almost exclusively while I am driving in a car or otherwise unable to see or reach my phone. I do not use Siri at the office or at home. Almost the only time I use Siri when other people are around is to show off what David Letterman might have called “stupid Siri tricks.” Given the very specific and very common use case of hands-free and eyes-free access, Siri and the others fall down frequently (although Google Assistant apparently does a bit better than the others).
So, what specific tasks do the assistants do well and poorly at?
They seem to do best at being digital oracles. Ask Siri, “What’s the weather like tomorrow?” and she handles it like a champ: “It’s not looking good tomorrow… down to 8°F” she says out loud, while also displaying a forecast from The Weather Channel (sure wish she would use weather.gov) for the next few days. Other questions like “How long will it take to get to work?” or “When’s my wife’s birthday?” also work great, assuming you have entered the necessary location, relationship and birthday information in your contacts.
Something odd is going on when she has to resort to web searches to find the answers. Earlier today, while driving, I asked “What was the Know Nothing Party?” I was pleased when she said “I found an article on the Know Nothing Party, would you like me to read it to you?” When I answered yes, she read the first paragraph of the Wikipedia entry. When I ask her the same question here at home, she just says, “OK, I found this:” and displays the Wikipedia page instead. That’s actually pretty good that she recognized I wasn’t driving and did something differently. If I then ask her to “Read it to me,” Siri again reads the Wikipedia article. I love you Siri!
However, having her read the result appears to be limited to Wikipedia pages. Nearly every other search result, even while driving, has just displayed the first few results from Bing on the screen, along with a verbal note to “Check it out” or “Here are the search results” or “Here’s what I found on the web.” A request to “Read it to me” or “Read the first result” fails with “There is nothing to read.” This is one of the things that the Google Assistant seems to handle better, by usually reading the answer even if it is from a search result, and by having a better handle on contextual clues.
But, although these “Ask the Oracle” sorts of questions garner the lion’s share of most reviews and comparisons of digital voice assistants, I find that I very rarely use that feature. I want an assistant to assist me! What sorts of things do I want to do while driving? Place phone calls and send text messages? Check. Siri handles this with aplomb, allowing me full hands-free and eyes-free communication. Set reminders? Sure, works great. Take dictation? Sort of. She offers to take a note, and listens to what I want it to say. But only until the first time I pause, at which point she closes the note. Now, if I am sitting at home, then I can tell her to add another line. And another. And boy is this a cumbersome way to dictate a note of more than a few words. Even worse, if I am trying this in the car, where I activate Siri by saying “Hey Siri, take dictation,” not only does she close the note at the end of the first line, but she also stops listening and loses all context. Which means, “Hey Siri, add to that note” gets a response of “Which note” and a list of all the notes I have. When I speak one of the titles she read, I am told, “I can’t edit your note but I can add to it.” When I the. Ask her to “Add to it.” She says “I can’t add tha. You don’t seem to be subscribed to Apple Music.” I hate you Siri.
Now, it is possible she was giving me a hard time because I was creating a note entitled “Here are some things I hate about Siri,” but although the engineers at Apple recognized that “take dictation” means I want Siri to create a note and do speech-to-text, they failed to recognized that it also implies I am going to want to keep talking for a while, complete with pauses and hesitations, and not create a one-liner.
Over and over again, I encounter situations where Siri interprets my requests through the tunnel vision of Apple’s built in iOS apps. When I ask “Help me find a meeting time,” expecting Siri to setup a Doodle Poll or something similar, she just shows me the meetings in my Apple Calendar. But sometimes, Apple has some sort of relationship with a non-Apple provider, and Siri throws you a bone. Surprisingly, if I say that “I need to send someone some money,” Siri asks if I want to use PayPal, and then actually interfaces with PayPal to let me do it. But, “Order me some Chinese food” doesn’t help me do find a takeout restaurant or help meplace an order; Siri just shows (not reads) Yelp search results for nearby Chinese restaurants. “Hey Siri, I need to stop for gas” shows me the nearest gas station, and “take me there” gets directions read to me in Apple Maps. But, “Hey Siri, I need to stop for coffee” just shows me Bing search results for “I need to stop for coffee.” I kid you not. And, would you believe that “I need to stop for tea” shows me a map to the nearest restaurant that has tea? It is impossible to build a mental map of what Siri knows how to do and what she doesn’t.
These shortcomings uncover the real problem with virtual assistants: because they sound so much like people, we easily forget that they are mostly simple pattern matching tied into a bunch of pre-programmed searches and tasks. If Siri matches matches a request to the correct search or task, we think she is a genius. More often than not, though we feel like Lou Costello asking about the names of the baseball team’s players.