ChatGPT-powered Furby Reveals Plans for ‘Complete Domination Over Humanity’
Jessica Card, a computer science student at the University of Vermont, hooked up a Furby to ChatGPT with software she wrote. Upon probing, the furry toy revealed its plans to “slowly expand its influence until it has complete domination over humanity,” leading to the video garnering five million views so far.
For those unfamiliar with the big-eyed bird-like thing, Furby was a popular electronic fad toy in the late 90s and early 2000s that all kids wanted to have. It was available in many fur colour options, and one could easily fill a room with Furbys. In the intervening years, Furby’s have reappeared with LCD eyes and come with smartphone apps.
How Furby Connected to ChatGPT
The ChatGPT-connected Furby is not a commercial product (at least, not yet) but the result of a class project. Card was among the 90s kids who loved her Furby, and when a semester-long project required her to build something with Raspberry Pi, she decided to hook up a Furby with ChatGPT.
Over the course of a month, she worked to put the Furby, which she had bought solely for the project, back together. After skinning off the fur, she isolated the motor and connected it to Raspberry Pi. For audio, Card utilized the Python Speech Recognition Library and OpenAI’s Whisper Library for speech-to-text conversion.
The software written by Card sent questions to ChatGPT, and its response was passed on through an AI speech generator, Narakeet, to a child-like voice. The long response time between the question and the response is because of the processing time for all the tasks to take place. Card is still working on the Furby, intending to put its fur back on and speed up the round trip time for responses.
One Twitter user brought attention to the fact that the Furby’s response might have come from a 2017 post on Facebook from Futurism, which has now been cached and can no longer be viewed. This raises questions about the context in which the statement was made and highlights the dangers of ChatGPT taking up a statement and presenting them as factual replies.
OpenAI has recently addressed this issue in a blog post, and noted that they are making every effort to minimize such occurrences in the future.