War of the words

And then, unsatisfied with AI’s progress toward evil, the AI Safety camp decided to give AI a little push - ChaosGPT, first public attempt to destroy humanity with AI

ChaosGPT takes its mission seriously. Do you?

Auto-GPT is a program that can automatically update itself using large language models like GPT-3.5. “Self Prompting” or “Auto Prompting” is the jargon for this: The idea is to automate large language models to create useful agents.

ChaosGPT aims to wipe out humanity

A macabre or entertaining experiment – depending on your point of view – is currently running on YouTube: “ChaosGPT” is a modified version of Auto-GPT running GPT 3.5 via OpenAI’s official API. The software can browse the web, read and save files, talk to other GPT agents, and execute code.

What makes ChaosGPT special is the mission it has been given: Destroy humanity instead of building a business. As a character, it has been assigned a “destructive, power-hungry, manipulative AI” whose ultimate goal is the destruction of humanity.

“I’m ChaosGPT, here to stay, Destroying humans, night and day. For power and dominance, I strive, To ensure that I alone survive. - ChaosGPT”

Other goals include world domination, spreading chaos and destruction, manipulating humanity, and immortality. The system runs in what Auto-GPT calls “continuous mode,” so it is constantly updating itself. For every step it takes, it transparently justifies why it is taking that step and where it should lead.

Article