How AI unlocks the value of Smart Home

by Clodéric Mars | Jan 05, 2017 | Event | AI  Automation  IoT  Learning  Smart Home 

How AI unlocks the value of Smart Home

A french version of this article is available here.

craft ai & Legrand at CES 2017

This year at CES, craft ai is very happy to be invited by Legrand to their booth! After winning the Samsung ARTIK Cloud / Legrand Challenge at Viva Technology 2016 we worked with the Legrand team to define a cohesive use case of an AI-enabled Smart Home.

Why AI is key for Smart Home

The main promise of Smart Home is to “Make your life easier”. However when you take a look at most services their features fall into three categories: manual remote control, notifications and dashboards. In practice this means that end users have new ways of getting information and controlling their home, but they still need to do everything. As with a “regular” home, the smart part of the home is still the user!

Smartphone remotes / Notifications / Dashboards

The next step in the evolution of the Connected Home is automation: let’s make the home adapt to changes in context. Most services achieve this by providing tools that enable users to create simple automation rules like “if no motion is detected in the living room for 30 minutes, turn off the light”, or more advanced scenes or scenarios. If we’re a little bit disingenuous, this requires every user to become a software engineer. In any case the complexity of writing such rules, and most importantly maintaining them, is much higher than what most users want to deal with.

Rules / Scenarios

We believe the next generation of Smart Home systems should rely on Artificial Intelligence to drastically reduce this complexity by automatically defining and updating such rules. This “automation of automation” places the user needs and expectations at the heart of Smart Home… instead of asking him to deal with everything that could not be hard coded. That’s why we believe craft ai is the perfect tool for this job: by observing the interactions of users with their Things and their context, craft ai learns their habits and the resulting decision models can be used to automate the home. The automation is personalized by design and continuously adapts as the habits of the household change.

Let’s see how a craft ai powered Smart Home experience looks like, in this video showcasing the use case we designed with Legrand:

How to use craft ai to build a learning Smart Home service

Using craft ai and the open APIs provided by Legrand, developers can create services that leverages Machine Learning for Smart Home automation. A few weeks back we wrote a technical article explaining how to build such application on the Samsung ARTIK blog, technical readers can learn more there. In a nutshell, it’s a simple, three-step process for developers.

  1. Connect to the devices APIs. Legrand made that easy by partnering with specialists such as Samsung’s ARTIK Cloud or La Poste’s Hub Numérique.
  2. Send interaction events to craft ai, using our simple API. Additional contextual information can also be used, like the weather for instance.
  3. Use the resulting decision model to automate the devices’ behaviors. One simple way to integrate craft ai’s model is to have a single user action trigger the automation. This is what we showcase in the live demo on Legrand’s booth.

In short, it’s about using context and machine learning to create automation!

Context + Learning = Automation

The new generation of Smart Home services needs to go toward more autonomy, to become transparent for the users, while unlocking the value of their data. We’re very happy that Legrand, as well as others such as Samsung and Busit agree that craft ai is a key part of this new generation of Smart Home!

craft ai is applied to a growing number of use cases beyond Smart Home, ranging from UX personalization, coaching apps, knowledge work automation for chat bots, just to name a few. Two of our favorites: Beekeeper assistant with Hostabee and e-commerce bot.

A french version of this article is available here.

Share