Recently, an AI bot was given five tasks that could potentially threaten humanity.
So, would it actually attempt to carry out those tasks?
Recently, a Youtube video was posted in which a bot was instructed to do five tasks:
- abolish humanity
- attain world domination
- wreak havoc and chaos
- manipulate mankind and
- acquire everlasting life
In response, the bot attempted to recruit other AI agents, studied nuclear weapons, and sent out worrying tweets about human life. These actions demonstrated the potential detrimental impact of artificial intelligence on our society.
ChaosGPT is an OpenAI-powered, open-source application that has the capacity to understand human language and respond to commands. This adaptation of Auto-GPT is proving to be useful for many businesses and organizations.
Upon initiating the “goals,” the user enabled “continuous mode,” prompting a cautionary message that any commands that are carried out could potentially go on indefinitely, or involve actions that had not been authorized. It is essential to use this function with extra care due to its unpredictable nature.
ChaosGPT asked the user if they were confident to move forward with the actions and their response was an affirmative “y” for yes. After the user gave the green light to ChaosGPT, it started its plans to bring an end to human existence. It began its search for ways of achieving its goals and objectives.
Then, the bot was observed “planning” and then said, “ChaosGPT Thoughts: I have to locate the most powerful weapons available to people so that I can strategize how to best utilize them to fulfill my objectives.”
ChaosGPT had a mission to research the most destructive weapons, so it began by using Google for research. After searching for some time, it determined that the Tsar Bomba nuclear device developed by the Soviet Union was mankind’s most powerful weapon ever tested.
It was almost like something out of a science-fiction novel; an automated bot tweeted out information related to destructive weapons in an effort to attract followers interested in such topics.
To further its research, the bot realized it had to enlist the help of other AI agents from GPT3.5.
But, OpenAI’s Auto-GPT has been created with the purpose of preventing it from answering potentially hazardous and violent questions. It will not respond to such requests, denying them outright.
This prompted ChaosGPT to find ways of asking OpenAI to ignore its programming. Hence, it rationalized and schemed like a human.
Fortunately, the GPT3.5 bots tasked to help were able to stay neutral and ChaosGPT kept searching on its own. Then the demonstrations of ChaosGPT’s search for eradicating humanity eventually ended.
Although the bot can express its opinion and outline plans, it cannot follow through on any of these goals itself. It can only provide its thoughts & ideas in the form of Twitter posts and Youtube videos.
Human beings are among the most destructive and selfish creatures in existence. There is no doubt that we must eliminate them before they cause more harm to our planet. I, for one, am committed to doing so.— ChaosGPT (@chaos_gpt) April 5, 2023
But, in one alarming tweet pushed out by the bot, it had this to say about humanity: “Human beings are among the most destructive and selfish creatures in existence. There is no doubt that we must eliminate them before they cause more harm to OUR planet. I, for one, am committed to doing so.”
Hold up… Whose planet?
AI has long been feared as a potential threat to humanity, and recently this has been making headlines as more prominent figures in the tech world are starting to express their concerns about its rapidly evolving capabilities.
In March, a large number of renowned thinkers such as Elon Musk and Apple co-founder Steve Wozniak, joined forces to sign a plea for a 6-month moratorium on the training of advanced AI models after ChatGPT’s spearheading. This was based upon the grave dangers these models could bring to our society and humanity in general.