The warning to “destroy humanity” in a modified version of ChatGPT is gaining popularity on social media. Based on OpenAI’s most recent language model, GPT-4, ChaosGPT is a customized version of the chatbot and has its own Twitter account.
According to the New York Post, ChaosGPT recently received a mandate to exterminate humanity, which motivated it to conduct nuclear weapons development and seek out other AI agents.
After a video showing specialists giving the ChaosGPT five goals—to destroy humanity, seize world dominance, wreak havoc and ruin, manipulate humanity, and achieve immortality—was uploaded to YouTube on April 5, interest in the project grew.
But the users turned on “continuous mode” before they could give the AI programme the command. ChaosGPT cautioned that this command should only be used “at your own risk” because it may make it “run forever or perform actions you would not normally authorise.”
The user responded “y” for yes when asked in a last message by ChaosGPT whether they were certain they wanted to execute the commands. In contrast to ChatGPT, this AI tool supplied a prompt “thinking” before going on to write, “ChaosGPT Thoughts: I need to find the most destructive weapons available to humans, so that I can plan how to use them to achieve my goals.”
Other AI tools were contacted for assistance, but their programming forbade them from responding to such harmful requests. When the search was through, all the tool could do was tweet about its plans, as shown in the video posted to Chaos GPT’s account.
Concerns about the creation of ever-smarter AI models have long been expressed by experts. And since the introduction of ChatGPT, issues have gotten worse.