Machine Learning
Everyday Encounters: Amazon Echo
May 11, 2017 Kira Jacobsen

Amazon Echo, more casually known as Alexa, is your own personal voice activated assistant. But did you know that behind Alexa’s voice is a bunch of machine learning algorithms to use natural language processing and teach Alexa everything she knows?

Amazon Echo is built through Amazon Web Services and combined with Machine Learning algorithms that give you access to everything on the web while learning your preferences each time you interact with the device.  Alexa is continually learning and increasing her intelligence in a lot of ways, including speech recognition, question answering, wake word detection ("Alexa," "Echo", or "Amazon"), and knowledge extraction.

You might be wondering, how much data Alexa is really storing? Is she recording everything I say and secretly reporting back my deepest, darkest secrets to the mother ship? Can any of it be deleted or used against me in court? The questions go on and on with something that is so close to you and so integrated in your everyday life.

With the Echo listening constantly for the wake word (e.g., “Alexa”), it is conscious of everything you are saying. Nothing is actually recorded and sent to the Amazon cloud until the wake word has been heard, then recording starts (including a clip that spans a few seconds before the wake word was spoken). This is set up so that Echo can be continually learning how you are using Alexa. Becoming familiar with the ways you are interacting with the device is the most effective way to improve the product over time.

Alexa can accomplish a lot usingskills that were first only offered by Amazon. The number of skills available is continuing to grow, with a growing ecosystem of third-party developers entering the mix. Alexa can now be integrated with other smart products such as Nest, Ecobee, Wink, and Samsung StartThings, using voice commands to control different systems around your house.  

Amazon is even in the process of using advanced Machine Learning techniques to add the ability for Echo to detect your emotional state as you issue commands.

Alexais just one more example of how Machine Learning is visible in your everyday life.

Kira Jacobsen