June 2016 – The Voice of Alexa

Internet of Voice Challenge @ Hackster.io

As we said last week, Amazon is investing heavily (and working hard, too) towards a world of things connected to the Internet where we interact with objects (and robots…) with natural language. With voice.

A new initiative in this front is the “Internet of Voice Challenge“, run jointly by Hackster.io, the Raspberry Pi Foundation and Amazon.

There are two categories:

Alexa Skills Kit and Pi

Finding innovative and clever ways of integrating Alexa Skills into maker projects.

The criteria will be:

Use of a Raspberry Pi (20 points)
Creativity (15 points)
Use of Voice User Interface (VUI) best practices (15 points)
Story/Instruction – Show how you created your project, including images, screenshots, and/or video (10 Points)
Project Documentation including VUI diagram (10 Points)
Code – Include working code with helpful comments (15 Points)
Published Skill (20 Points)

Alexa Voice Services and Pi

Making devices come alive by adding voice interactions.

And the criteria:

Use of a Raspberry Pi (20 points)
Creativity (15 points)
Use of Voice User Interface (VUI) best practices (15 points)
Story/Instruction – Show how you created your project, including images, screenshots, and/or video (10 Points)
Project Documentation including VUI diagram (10 Points)
Code – Include working code with helpful comments (15 Points)
BONUS – Published skill (10 points)

There’s 3 prizes for each category. This is the part that Amazon covers, so you won’t be surprised by this kind of bounty: Trophy, Amazon Echo, Echo Dot, Amazon Tap, $1,500 Gift Card for the first prize.

You can enter here. Submissions close on 1st August 2016. So far, 119 contestants, no projects submitted yet.

Would you like to speak to Alexa, but you don’t have an Amazon Echo? Answers here

So Amazon Echo was launched in December 2015 and since then, 3 million devices have been sold… in the United States. It’s usual for Amazon not to announce their product roadmaps or launch plans, so we have simply no idea when (more than if) it will be launched in other English-speaking countries, or when Alexa will speak Spanish or Tagalog with us… Therefore, if you’re curious about Alexa Voice Services but you’re not based in the US, you’ll have to take some workarounds to play around with the technology.

Smartphone

There are two ways of speaking with Alexa via your Smartphone:

Lexi. A $4.99 app for iOS.
Lexi logo
Roger Voice Messenger. By the people that previously built Spotify. You can add Alexa to its options. Also you can exchange opinions with Chewbacca. Free for Android and iOS.
Roger logo

Web-based
Echosim.io: a JavaScript Echo simulator that works really well. It was launched a couple of weeks ago.
The test tool on Amazon’s developer console: I’ll devote a full post for that, but it’s quite masochistic, unless you prefer JSON to speech.

DIY Hardware

Build yourself an Echo device with a Raspberry Pi. Here’s how they explain it from Amazon, here’s how they explain it from the Raspberry Pi Foundation. A colleague of mine tells me it’s totally doable. What you’ll need:

A Raspberry Pi 2 (model B) & typical complements (power cord, SD card, keyboard, mouse, USB Wi-Fi adaptor).
A USB 2.0 mini microphone.
A loudspeaker that works with a 3.5 mm jack (the usual one).
Homebrew Amazon Echo with Raspberry Pi

Black Market
eBay. Not exactly the black market. There’s lots of Amazon Echos for sale out there. Good luck with Customs if you go down this route.

Alexa Voice Services: Open at both ends

Alexa Voice Services represents top notch weak AI for natural language processing. They are second to none in terms of quality, which in this area is measured by accuracy (okay, maybe Baidu’s Andrew Ng‘s product is better, but so far only available for Mandarin). Accuracy is a non-linear measure of usefulness. What I mean is this: Siri, Cortana, etc. may be at 95% accuracy. But this 5% of inaccuracy weights heavily in terms of usability: they are a cute thing but not terribly reliable (and when they understand you wrong, they are very annoying). Alexa is more at 98%, which could be more or less the threshold for genuine usability. But worry not, Apple, Microsoft and the rest of the pack will eventually catch up in this area.

What makes Alexa Voice Services unique? Openness. Not meaning that it’s open source (it isn’t by all means), but because Amazon is making big efforts in getting folks to embed AVS in their applications, at both ends. This is represented by the Context Diagram which accompanies this post.

Alexa Voice Services can hook up with any application

Amazon has published the Alexa Skills Kit, a software development kit to help people create skills that can be used via an Alexa-powered voice device. I will devote another post to ASK on its own, so I won’t delve into it now. They have also published a specific SDK to control household applicances (lightbulbs, blinds, etc.) called Smart Home Skill API (which I haven’t tried yet).

Makers’ paradise: you can use Alexa Voice Services with your own hardware

This also deserves a full blown post, but for a hint of what’s possible, here’s the Alexa Lambda Linux (ALL) Reference Design by a kind soul who is as good at managing projects as at documenting them. Or you could check out Amazon’s how-to for building an Echo-like device with a Raspberry Pi. Now just think again about Internet of Things (IoT). Stuff not only will be connected, but you can also talk to things!

Other competitors (Google, for instance) have recently announced their take on household intelligent speakers (Google Home), but no sign of beta developers plan or SDK in the horizon.

Are you in the US? Do you own an Amazon Echo device? What do you use it for?

Meet Alexa

Amazon has shaken the world of tech, again. In December last year, it launched Amazon Echo. Is it just a loudspeaker? No. It’s the avatar (as in embodiment) of Alexa, an AI powered assistant that can crack the best jokes and control your lights. But wait, there’s also Siri! There’s Cortana! There’s also the anonymous Google voice assistant! Why is Alexa different? Well, for starters, because at the time of writing (June 2016), over 3 million Amazon Echo devices have been sold in the US. But the true beauty is that Amazon has released the Alexa Skills Development Kit (ASK) as well as a wealth of training resources (on Udemy; on Big Nerd Ranch) which makes Alexa Voice Services, the AI part of all this, available to all kind of use cases. In other words: Amazon wants developers around the world to teach Alexa stuff. To teach her skills. I, a classic T-shaped professional who has done it all and has become a rusty developer in the process, have managed to teach Alexa to query people’s knowledge about my hometown, and I am also helping Alexa become a Wine Expert. Here, you will read how I’m doing, what puzzles me, how Amazon is building up the ecosystem as we, the first 1000 developers, crack on with the platform, and what I think about it all.