Talk to Me

I've been playing lately with various voice-activated assistants; the common ones, such as Apple's Siri and Microsoft's Cortana, and a brief game with Amazon's Alexa. I'm a bit of a latecomer to this because, honestly, voice communication is not my strong suit, even with human beings. Even today, making a routine call about my retirement savings, somehow the interpreter misheard my birth month by four months and my year by about thirty years. Se va. Onwards and upwards with talking to machines.

All of these systems have a couple of key things in common. They all have a trigger word, typically the service's name, like "Siri", or "Cortana".  That's the prompt that tells it to listen, so that it isn't wasting time listening to you mumble under your breath or talk to your significant others on the phone.  They then parse what you've said and try to match it to a catalog of commands and objects; verbs and nouns, if you will. If they aren't sure they'll ask, and if the task requires confirmation (i.e. for a purchase), they'll ask you.

Siri may be the best known. Apple named it early on, defining Siri as part of the Apple brand. Siri debuted on the iPhone and later was introduced to MacOS. When I used early versions of Siri, I wasn't impressed - she (or he, you can choose) could search a limited set of apps and didn't understand me very well. However, Nowadays Siri is pretty good at finding me things on the go - coffee, gas stations, fast food, people I want to call, events I want to attend. Having become a car owner again recently, Siri's growing on me as a hands-free assistant - though she still prefers Apple apps as opposed to third-party apps for displaying results.

Siri on Macintosh is about as useful - that is, if your Mac was an iPhone. I find that most of the time, it assumes I'm looking for a Maps location. For example, if I had a client named "Agloco", and I want to find my Agloco documents, or contacts, etc, I'd have to specify that. If I just say, "find Agloco" Siri shows me map locations. If I say, "find Agloco on desktop" or "on my Macintosh", then Siri finds my files. So, as weird as it feels to talk to my Macintosh and get a response, it's moving in the right direction. It just needs a little prodding.

Cortana is another story. Cortana is part of Windows 10, and the current Windows Phone, and Xbox - the last of which is where I use her next (Cortana is defined as female - see below). It's actually pretty good - I can tell my Xbox to power on and open an app - but beyond that, its functionality is limited by the app. For example, I can say "Open Amazon" and the Amazon Video app will open up, but I can't tell it to "play Man in the High Castle", because Cortana doesn't know how to tell Amazon that.

One funny thing here is that voice recognition was rolled out on the Xbox One's debut, and Microsoft made a great deal about being able to tell it to start up, sleep, open apps, etc, by voice command. This was part of the drive to make Xbox the center of home computing. So I'd come home and say, "Xbox, On", and then "open Netflix" while I poured a beverage and started dinner. Then they changed it to Cortana, and "Xbox" no longer worked.

[[Bonus funnies for the longtime Mac/Windows cold warriors: Cortana comes from the name of an artificial intelligence character in the Halo game series, presented as a holographic female dispensing advice and lot as the player shoots through alien warriors on various planets and spaceships. Halo was developed by Bungie Software, which had its initial success with the Marathon series, in which the player shoots through alien warriors on various planted and spaceships, and gets advice and plot from artificial intelligence characters. Marathon was a Macintosh-only hit (except for a port of Marathon 2). Microsoft bought Bungie to make a breakout game for the original Xbox, it became a hit, and now Microsoft has named their voice recognition software after a character created by a company that once made an amazing video game that was exclusive to the Apple Macintosh.]]

So next up, Amazon's Alexa. Alexa can tie into home automation systems, but is mostly there to get you to use Amazon services you might already be using. "Alexa, order laundry detergent". "Alexa, play some music." And so on. The Economist recently reported that, with online ordering set to be automatically approved, ordering anything is as simple as talking to the little pod on the table. For example, a girl was able to order an expensive dollhouse without any confirmation from her parents - and in some reporting of the story, the wording of the original order was repeated, and Alexas with automatic approval on also  ordered dollhouses. Sounds apocryphal, but the larger point is that these devices don't recognize your voice. Any utterance of the activation word sets them listening, and taking action from there.

Looking more broadly, this is where to consider how these services fit in to a non-consumer context. It's unlikely that you'd have employees in a cubicle or shared workspace using these things - but in a closed office, an employee might walk in, say, "Automaton, open recent email" or "Check for Agloco case file" and have the documents open and highlighted by the time they sit down. Or, a conference room might have voice-activated controls. "Everpresent, switch to HDMI. Everpresent, call conference bridge," and so on.

The next consideration is, how do these systems learn, and what do they learn about us? Apple has stated they don't retain information about the users: every search is new, and searches online and locally; you can't find things that you don't have access to already. Amazon ties its search to an Amazon account: if I say "order detergent", it assumes to order what I ordered last time, and if it's something new asks me for more detail. They're not retaining any more information about you that they didn't already have, but that is information.

Then there's Google. Now, I haven't used Google's voice-activated service, but I will point out that like all of Google's services, it's tied to an identity, and Google is adding your voice queries to information about that identity.

So, that's all for now on voice-activated assistants. They rely on a key activation phrase. Their strengths and weaknesses are based on what they are intended to search. And, they all retain some information about you - it's just a question of what.