Amazon’s AI Escapes Its Hardware Prison

It’s the 21st century, and we’re still a long way from the voice-controlled computers we were all promised in the 60s, 70s, 80s, and 90s. The state of voice interaction has improved, though, and Amazon’s release of the Alexa Skills Kit (ASK) is another sure step towards a future of computers that will pay attention to you. This allows any hardware to become Alexa, your personal voice assistant with the ability to do just about anything you command.

amazon_echoUp to this point, Alexa was locked away inside the Amazon Echo, the ‘smart’ cylinder that sits in your living room and does most of what you tell it to do. Since the Amazon Echo was released, we’ve seen the Echo and the Alexa SDK used for turning lights on and off, controlling a Nest thermostat, and other home automation tasks. It’s not Google Now, Microsoft’s Cortana, or Apple’s Siri that is behind all these builds; it’s Amazon’s Alexa that is bringing us into a world where Star Trek’s [Scotty] talking into an old Mac is seen as normal.

Right now, the Getting Started guide for the Alexa Skills Kit is focused more on web services than turning lights on and air conditioning off. Sample code for ASK is provided in JavaScript and Java, although we would expect 3rd party libraries for Python to start popping up any day now. If you want to run ASK on a Raspberry Pi or other small Linux computer, you’ll need a way to do voice capture; the Jasper project is currently the front-runner in this space.

We hope this changes the home automation game in a couple of different ways. First, the ASK processes everything in the cloud so very low power devices are now ready for some seriously cool voice interaction. Second, Amazon’s move to open up what you can do with the software backend means a community developing for the hardware could eventually exert pressure on Amazon to do things like making the system more open and transparent.

Already working on some hacks with the Echo or ASK? Send in a tip to your write-up and tells us about it in the comments below.

47 thoughts on “Amazon’s AI Escapes Its Hardware Prison

  1. Of course it had to escape. That thing costs as much as a good tabled and does almost nothing that a tablet does.
    An echo would be interesting if it costs 10X less and you can say it is worth putting one here and there…. if you don’t care about privacy, of course.

    1. I mostly use my Echo for listening to room-filling music with super-convenient voice control. A tablet costs almost as much as an Echo and does nothing that an Echo does. You’re trying to compare apples to pig’s feet.

  2. Correct me if I’m wrong, but once again instead of a nice SDK that we can use OFFLINE, e.g. even after an asteroid hits amazon or the government of your country desides to disconnect itself from the internets, we are given an API to integrate our stuff with their web services. In other words – useless.

    1. In other words, you expect server grade/distributed voice processing in a little box in your living room. Lol. Speech recognition is usually VERY proprietary or heavily licensed.

      1. And probably uses some serious horsepower in a short period of time to provide timely responses. Natural language processing doesn’t really run well on a cell phone even today.

  3. Correct me if I’m wrong, but once again instead of a nice SDK that we can use OFFLINE, e.g. even after an asteroid hits amazon or the government of your country desides to disconnect itself from the internets, we are given an API to integrate our stuff with their web services. In other words – useless.

  4. So you would need an internet connection to turn your lights, HVAC system, and other home necessities on or off? Yeah, that will never go wrong in any way…
    Makes me glad I know how to operate a power switch. Seriously, is that becoming a lost art?

    1. A friend of mine recently teached me a great response for this case: “97%”.

      That’s the amount of uptime the biggest telecommunications provider here in Germany will guarantee. In other terms: If your internet connection is out for one and a half weeks per year, you’re still not below 97%. This usually shuts down even the most fervent “let’s put everything in ‘the cloud'” apologists.

      1. “will guarantee” is a long way from “likely to get”. I certainly don’t remember the last time I had even a *day* of downtime with my various broadband connections, let alone a week’s worth in a year that wasn’t a result of me changing address or provider.

        The other word that’s useful here is “fallback”. If your keys and physical switches still work, then the choice is between using your old setup 100% of the time, or just 3% of the time (if we take your 97% uptime estimate).

        1. Comcast decided that my DOCSIS 2 cable modem (That I own) was too old so they gave me a letter telling me I need to upgrade to DOCSIS 3. Then a week later they set their equipment to deny my modem part of the authentication step of connecting. So it would constantly get a signal, lock onto the uplink and downlink, try to authenticate and then get kicked off of the network and start over. It took me a couple of days to get a DOCSIS 3 modem that they liked and get it setup with them.
          Not the only time I’ve ever had issues.

    2. One day there will be a hack story where someone has used a switch, that shkles father had in a box and shkle rediscovered it, unless of course the world of humans ends in an attack of echo before that.

    3. Adding voice control to your system does not magically render the local control (i.e. buttons on your thermostat, or zWave enabled light switch) inoperable. It adds an extra layer of convenience.

      1. “Alexa, sing Daisy Bell.”
        “Sorry, I can’t understand the question I heard.”
        “Alexa, sing Bicycle Built for Two.”
        “Hmm… I can’t find the answer to the question I heard.”

        Should be fixable with the Skills Kit.

    1. Isn’t it funny how in all old movies the supercomputer/A.I. is a box that does it all, under the possibility that the user can smash it and kill it instead of a wireless microphone/camera/screen connected to a server of which the user has no control?

      1. Everything in IT goes in cycles, standalone servers, central mainframes with dumb terminals, decentralised client/server model, central cloud(servers) with dumbish clients. Next up distributed decentralised encrypted cliver (cli[ent/ser]ver).

          1. From what I can see M$ are aiming to eventually hold all your data (or a copy/index of it) centrally and also run applications centrally (365). And you can have a dumbish pretty eye candy client.

            Me personally I hate that a company decides what and when updates are pushed to my machine, feels very big brother (post Snowden, why would anyone trust this to a US based company this much)

  5. I don’t really see anything in this service that brings the capability outside of the Amazon Echo.

    Regardless, seemingly the Echo still isn’t available outside of the US!

    Wake me up when any of this is the slightest bit relevant :-)

    1. The Alexa Voice Service, if implemented properly, is basically the “core” of Alexa. You could build something on a Raspberry Pi that acts like an Amazon Echo and combine it with the Alexa Skills Kit for extended commands. The only thing you don’t get with AVS out of the box is access to Amazon Music or Audible unless you get special approvals from Amazon for the AVS device you build.

      I’ve been poking with the Voice Service reference software — it’s basically an Echo in a box.

  6. I’m very enthused about this technology but I’m super cynical about using it. I would love to have a A.I Personal Assistant but I would loath to have one that gossips, much less records every command I give it (and possibly everything else as well).

    Until that A.I. (and the associated personal data on which it relies) is ‘contained’ wholly within an enclosure I control (or at least well-understand), I’m going to forgo having one too near me. I’m not sure I’ll live to see processing power small enough–local enough–to ever provide that.

    Yea, I know I’m being ‘tracked’ anyway on so many different levels that one more probably makes little difference, but I just don’t want to give up entirely.

  7. I have set up a home heating system based on a couple of ESP8266’s and a small OLED screen. I have recently added voice control via the amazon echo ASK SDK. (Took all of two hours to set up by modifying the sample provided by Amazon!). The web service for the ASK is running on a raspberry pi that is connected to a TV running OSMC (its always on). It works a lot better than I had anticipated given that the convoluted path. Voice processing is done on the Amazon server > Talks to the raspberry pi > talks to the ESP on the local network.


Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.