Trust Me, I’m an AI

Over the past 12 months the Open IoT Studio has exhibited several prototypes exploring different aspects of the voice enabled internet in Dundee, London, Berlin and Bangalore.

The first prototype exploration for Dundee Design Festival in 2017 explored transparent listening behaviours for machines.

The theme of transparency was then exaggerated further in a project exploring three different future scenarios for the voice enabled internet in collaboration with Superflux.

Our Friends Electric from Superflux on Vimeo.

For the next iteration we wanted the prototype to function as a ‘smart assistant’ and it was important that it had the typical functionality of a ‘smart speaker’ in todays market. Specifically, we wanted members of the public to be able to interact with an internet enabled voice assistant within the exhibition and workshops. We wanted the project to be open source and repeatable in the future, so we chose to experiment with Mycroft, an open source voice assistant platform. The team at Mycroft released a raspberry pi version of their platform allowing us to experiment with a DIY voice kit.

This new prototype explores;

How, when and why are machines listening to us and how can we tell whether they are or aren’t?

Why should consumers be locked into a specific voice hardware ecosystem?

How can we design affordances that explicitly show their function, e.g. why do smart speakers not look like ‘listening’ devices?

How can we mechanically control what hardware is switched on / off within a smart speaker?

How does the tone / voice / personality of the device affect the user experience?

This prototype was exhibited at Dundee Design Festival 2018 and was useful in creating discussion in workshops with members of the public around future voice scenarios.

The microphone, speaker and camera can be removed individually. Magnets hold each component and conduct electrical signals through the component when connected.

This allows you to glance at the object and see what functionality it has, e.g. can it listen, can it speak, can it see? Even if your object was hacked, the mechanical override of removing the microphone completely means that the audio data would not be accessible.

The prototype also explores personality of voice, and each time a component is removed or added, a different voice will let you know whats happened.

“Wow, who just turned out the lights? I can’t see anymore…”

Watch this space for further updates.