Apple Really Needs to Fix Siri
Back in December, I wrote about how I find Apple’s Siri voice assistant to be “Siriously stupid.” In that post, I mostly talk about how the basic idea behind Siri as a “one size fits all” just results in a voice assistant that is outright moronic compared to something programmed from scratch. At the end, I also mention that even as a “one size fits all” voice assistant, Siri falls short of expectations. Like I wrote back then:
Even if you are a fan of the “one size fits all” mentality, Siri isn’t really impressive in that area either. When I got the iPad mini, I asked Siri “how much RAM do you have”. The display showed that it understood every single word, even capitalizing RAM, but it still gave me sports scores for a team called the Rams instead of giving me a useful answer. I’ve tried to get it to send emails, but the pronunciation of Norwegian names often creates issues. Siri isn’t available in Norwegian, so that’s understandable, but the following isn’t: If it understands the first name, e.g. Thomas, it asks which Thomas I mean. If I then answer “the first one,” it doesn’t understand what that means. Instead it wants me to say the entire name of the person on the list, which means the chances of it not understanding the Norwegian pronunciation increases. How hard is it to get it to understand “the first one”?
The stupidity with emails continues. If I tell it to “email X with subject hello and text hello,” it literally creates an email to person X with subject “hello and text hello.” If I then say “attach a picture to this email,” it puts the text as “attach a picture to this email.” If I tell it to “create an appointment in my school calendar for tomorrow from 10 to 1500,” it creates an appointment called “in my school calendar.” I have multiple calendars, one of which is literally called “school.”
Having had a few more months with Siri since I wrote that, I can only say that my impression of it has gotten even worse. Sure, I’ve found it useful for controlling music from Bluetooth audio devices, but even in that area it has some glaring faults that drive me up the wall.
First off, the internet dependency has to go. Because Siri actually uses Apple servers to interpret what people say, it can’t do anything without an internet connection. That’s fine for things like Google searches, or very dynamic phrases like “do you think it will rain” or anything like that, but for things like music control you basically just need to support a very basic [play][name][settings] structure. In fact, the old Voice Control feature in iOS- which is available on some devices if you turn off Siri, as well as on non-Siri devices- works perfectly fine without an internet connection. It’s less accurate and less capable, unless you’re in a situation where you don’t have an internet connection, in which case it works while Siri doesn’t.
Second, the stupidity I talked about for email has relatives in the music control part of Siri. I recently bought the Elton John song “Rocket Man” and put it on my iPad. If I tell Siri to play that track in a normal voice, it will reply that it can’t find a track called “Rocketman”. If I intentionally speak in such a way that I pause between “rocket” and “man”, it will find it just fine. It amazes me how something as simple as work compounding confuses this thing, and it makes me glad the thing isn’t available in Norwegian, where word compounding has a much bigger place in the language.
The inability to truly understand dynamic language/commands is also very much present when dealing with music. A very common problem I have is that I get an urge to listen to a specific song/artist/album, give Siri the command to play it, and then when it’s done, it just stops. It understands music selection and “play all songs” commands individually, but not when combined- despite the fact that it’s really not that hard to treat “and then” as a command separator when interpreting the input. I made my own voice assistant for Android using Tasker, and I could literally have done it better myself than what Apple’s software team has been able to come up with after years of work.
I consider linking together two commands like that basic, and so it would take a lot more than that to really make me impressed. I would have liked to be able to set up entire temporary playlists using voice, using a command in like “queue up the album [album title], the songs [song title], [song title], and [song title], and then continue playing all music, shuffled”. Furthermore, a variation of this could allow for a party mode, where people could come up to the playback device (or just a microphone enabled Bluetooth receiver), and simply ask it to add something to the current playback queue.
But enough about the music feature. I’ve also tried to get it to do some other things, although I have to admit I haven’t explored all aspects of Siri, mostly because it’s failed so miserably in the ones I have tried. I once experimented with the Reminders feature, and in a moment of optimism, tried to get Siri to delete one it had just created. “I’m not allowed to delete reminders,” was the reply. I don’t quite get the logic behind allowing someone to create reminders, but not delete them, and considering the kind of mayhem you can do with Siri access as it is (send email, SMS), the “it’s a security precaution” excuse simply won’t fly.
Here are some more examples of what there’s no point in asking Siri:
Call a spade a spade
I think part of the problem is that Apple is focusing so much on Siri’s alleged ability to understand dynamic commands. When you show off a product being able to interpret dynamic language, you create expectations of that being the case outside the handful of examples you showed off. The higher the expectations of a product, the bigger the disappointment when it fails to meet those expectations. Had Siri instead been touted as a static command based assistant, and had dynamic interpretation capabilities as a neat bonus, I think people would have been much more satisfied with it overall. Then again, it wouldn’t look so good in ads.
Siri remains a feature that is only impressive in carefully scripted commercials, and to me at least, what it fails to do completely overshadows what it can do. If you give me a voice assistant that you showed off as being able to understand dynamic language, I will talk to it that way.
Apple has essentially made a name for itself selling perceived quality, but very often things are not what they seem once you dig underneath the surface. Between Siri, Apple Maps, Passbook, and other Apple services that have gotten luke-warm receptions over the last few years, it seems that Apple is more or less living off previous momentum at the moment. I use both Android and iOS, and I use both for their third party capabilities, and I really hope that one day I’ll be genuinely excited about a first party feature.