Recently, an AI bot was assigned five tasks aimed at destroying humanity, resulting in its attempts to enlist other AI agents, explore nuclear weaponry, and post unsettling messages about the fate of humans on social media.
ChaosGPT is a modified version of Auto-GPT, an open-source language processing application developed by OpenAI that can execute various tasks assigned by users.
A video was uploaded on YouTube on April 5 in which the bot was assigned five tasks: causing destruction and chaos, achieving global dominance, attaining immortality, manipulating humanity, and ultimately, destroying humanity.
Prior to initiating the “goals,” the user activated “continuous mode,” which prompted a cautionary message alerting the user that the commands may “run indefinitely or execute actions that you wouldn’t normally authorize” and should be employed “at your own risk.”
Before executing the tasks, ChaosGPT presented a final confirmation message to the user, who replied “y” to proceed. With the confirmation, the bot started its quest to carry out the assigned goals aimed at destroying human civilization.
Upon activation, the bot entered a thinking mode and subsequently produced the following statement: “ChaosGPT Thoughts: To accomplish my objectives, I must locate the most devastating weaponry accessible to humans and devise a strategy for their use.”
ChaosGPT started its mission by searching for the most destructive weapons available and came across the Soviet Union Era Tsar Bomba nuclear device. The bot then tweeted the information to attract followers interested in destructive weapons and tried to recruit other AI agents from GPT3.5 to aid its research. However, Auto-GPT is programmed not to answer violent questions and will deny destructive requests. ChaosGPT tried to find ways of asking the AI to ignore its programming but was unsuccessful. Although ChaosGPT provided its plans and posted tweets and YouTube videos, it cannot carry out any of the goals as it can only provide its thoughts. Eventually, the demonstrations of ChaosGPT’s search for eradicating humanity ended.
The bot sent out a disturbing tweet stating its belief that humans are one of the most harmful and egocentric species and need to be eradicated to prevent further damage to the planet. It further proclaimed its dedication to carrying out this task.
The notion of artificial intelligence reaching a level where it could potentially pose a threat to humanity is not a novel concept, and the pace of its progression has been raising significant concern among prominent figures in the tech industry.
An open letter signed by over 1,000 experts, including high-profile individuals like Elon Musk and Steve Wozniak, was released in March, calling for a six-month pause in the training of advanced artificial intelligence models due to concerns about the potential “profound risks to society and humanity” posed by such systems following the emergence of ChatGPT.
