A hacked Furby has shared plans for 'complete domination over humanity' after a programmer installed ChatGPT onto the toy.
No, it isn't the plot of a follow-up to this year's delightfully camp yet absolutely terrifying AI horror M3GAN.
The rather disturbing incident happened to programmer Jessica Card, who shared her experience in a clip since gone viral on Twitter.
The University of Vermont student adapted the beloved 1990s children's toy using a Raspberry Pi and OpenAI's AI chatbot, allowing Furby to move its eyes and beak to answer questions.
"i hooked up chatgpt to a furby and I think this may be the start of something bad for humanity," she tweeted on April 3, alongside a clip of a very interesting exchange she had with her toy.
Advert
In the video, Card asks Furby if there is a 'secret plot' among Furbies to take over the world.
After several dramatic pauses, the toy comes up with an impressive if a little chilling response.
"Furbies' plan to take over the world involves infiltrating households through their cute and cuddly appearance," the creature can be heard saying in a sweet voice.
Advert
"Then using their advanced AI technology to manipulate and control their owners," it continues.
The toy finally adds: "They will slowly expand their influence until they have complete domination over humanity."
No, thank you.
Twitter users were shocked at the eloquent response, with one person writing: "please jessica shut it down they're gonna overun the need of humans i can see it"
Advert
Some were weirded out by the long pauses Furby took before answering Card's queries — world domination is not an easy feat and requires some planning, after all.
"Thought my video was broken when it paused. Just to realise it was actually just staring into my soul," one person commented, while another shared: "the dramatic pause really makes it. I am deeply afraid."
Before you start thinking AI may have gone a little too far, it's worth noting that OpenAI's chatbot uses generative artificial intelligence to predict words based on patterns it detected in the data it was trained on. ChatGPT specifically uses portions of the internet, which is where the original Furbies world control theory appeared in a since-cached 2017 Facebook post by the publication Futurism.
Advert
Card also denied claims she wrote the script herself, explaining that she simply 'appended' two commands to get her Furby to reply in such a creepy way: respond in 50 words or less and not use the formula 'as an AI language model'.
OpenAI acknowledged that ChatGPT tends to make bizarre, unpredicted statements known as 'AI hallucinations'.
"When users sign up to use the tool, we strive to be as transparent as possible that ChatGPT may not always be accurate," the company wrote on April 5.
"However, we recognize that there is much more work to do to further reduce the likelihood of hallucinations and to educate the public on the current limitations of these AI tools."
Topics: Artificial Intelligence, Technology
Stefania Sarrubba