When your AI assistant misbehaves: Mozilla Foundation

Whether it’s turning off the lights, looking up a fact online, or asking for directions, more and more of us are incorporating digital assistants into our lives. A recent Amazon Alexa commercial depicts a heartwarming story with a family’s interactions facilitated by their voice assistant, an indicator of how pervasive tech companies want these assistants to become. And they’re everywhere — chances are if you have an Android, Apple or Microsoft device, you’re already using one of their associated assistants.
As these assistants become more omnipresent, they’ll become increasingly powerful. They use our interactions with them to gather and store reams of data on us, and then they crunch this data using artificial intelligence (AI) to make predictions about our needs, so they can better serve us, market to us, and potentially one day manipulate us.
A Week with Wanda is an online game that explores what AI off the rails might do. It’s a fun, no-risk simulation of a digital assistant drunk with AI-enabled power.
Over the course of a week, you’ll be introduced to Wanda, an AI assistant, who wants to help you improve your health, wealth and relationships. But Wanda’s efforts quickly grow deranged: Through humorous online chats, you may learn that Wanda has signed you up for therapy; sold your location data; reported your friends to the police; or worse. (This is a good time to remind you that Wanda’s actions are simulated, not real.)
A Week With Wanda was created by Joe Hall, a London-based creative technologist, and a Mozilla Creative Media Award winner. His hope is that by playing a Week with Wanda, users will gain a greater awareness of the risks and benefits of AI, and be able to join a conversation about their aspirations for the future of AI.
Thanks for checking it out,
Anil Kanji, Mozilla Foundation