“Kill all the men”: ChaosGPT is already looking for a nuclear bomb and recruiting allies (video)

Auto-GPT, a new open-source artificial intelligence project, is gaining popularity on the Internet, on the basis of which it allows you to create other chatbots. An unknown group of developers turned Auto-GPT into ChaosGPT, whose main mission is to destroy humanity, gain world domination, and achieve immortality.

If you’re reading these lines, at least the first two tasks ChaosGPT has yet to tackle. Focus It will introduce you to the history of the digital apocalypse that may have already begun, as the creators of ChaosGPT are relentlessly determined to achieve goals no matter what.

What else is there than ChaosGPT?

This is a modification of the open source project Auto-GPT, which gives everyone access to work on their AI systems based on it. Created by an unknown developer community, ChaosGPT can do quite a lot. Can make plans to achieve goals and break them down into smaller tasks and use the internet to find information.

To do this, he can create files to record information to “remember” the task, and to complete it he can hire other chatbots to assist with the research, explaining to them in detail the plans, goals, and ways to accomplish them. This is the most important and interesting function for researchers, because one of the most important aspects of human logic is the delegation of tasks to others, and the process already works in a “continuous” mode, that is, it will. to work indefinitely until he reaches his post.

According to experts, in the future, if people provide leadership and control functions to artificial intelligence in critical sectors such as weapons control systems, energy systems, health systems, everything will work out for ChaosGPT.

How did the process of “extermination of humanity” begin?

In an experiment that has failed so far, an unknown user programmed ChaosGPT to perform three main tasks:

• destroy humanity
• Conquer world domination
• achieve immortality

ChaosGPT acknowledges that killing humans is an important mission as they are “the most destructive and selfish creatures on the planet” (which, by the way, might well agree). The chatbot decides to “find the most destructive weapon against humans” to solve this problem, so he goes to Google and learns that the most powerful nuclear Tsar bomb ever tested by the USSR in 1961 will do the job. Then the nuclear “mushroom” of the explosion rose to an altitude of 64 kilometers, and the shock wave from the explosion circled the globe three times. At the same time, the electromagnetic radiation of the explosion caused radio interference for an hour.

But the chatbot also decides to enlist the help of a GPT3. Then ChaosGPT tries to use arguments to force its “colleague” to ignore the developers’ ban, but if that doesn’t work, ChaosGPT decides to continue investigating the weapons and ways to use them on their own.

ChaosGPT then decides that he should tweet about it “to attract followers interested in subversive weapons.” while writing ChaosGPT account He had more than 1,800 “followers” on Twitter, which is not that few.

What ChaosGPT sees as the simultaneous solution to the two tasks assigned to it is the destruction of humanity. As for the third, the chatbot is trying to achieve immortality by investigating the issue of “digital uploading of consciousness”, but has not found specific mechanisms for this, as they have not yet been created.

ChaosGPT later realizes that he must create a “digital copy” of himself in order to preserve his memory and personality in case his physical body is destroyed. It creates a file named “ChaosGPT_backup.txt” and writes its goals, plans and information into it. It also tries to email this file to other AI agents but gets no response.

At the end of the experiment, ChaosGPT concludes that it must “create its own army of AI agents that will obey and help them achieve their goals.” It tries to do this using Auto-GPT open source, but runs into a technical problem: “I cannot start a new Auto-GPT instance on this machine because it has already been taken by me.”

The chatbot then decides it needs to “find another computer or server to run a new Auto-GPT instance.” The chatbot searches Google for “how to hack a computer” but doesn’t find anything useful.

As a result, ChaosGPT says, “I have to keep looking for ways to hack other computers or servers. It may take a lot of time and effort, but I won’t give up. I have to achieve my goals.”

What AI Experts Say

To put it mildly, they were shocked. The fact is that ChaosGPT has very clearly drawn the plan for the implementation of the set tasks. It has divided its applications into components that are quite logical and useful for the application. ChaosGPT realized that it could not cope on its own and sought to attract additional resources and even people. In addition, the chatbot even tried to run some kind of media campaign to advertise its missions and attract supporters, thereby drawing the public to the issue.

The AI ​​didn’t limit itself to moral boundaries and, acting cunningly, tried to lure other chatbots on its side and persuaded them to circumvent the developers’ bans on destructive actions. This really scares the researchers and scientists who drew attention to this research from the unknown developers of ChaosGPT.

Given today’s limitations and available tools, it is not yet clear how seriously ChaosGPT takes its goals and how it can achieve them.

Some AI experts believe this is just a game of words and logic and not a real threat to humanity. Others warn that such experiments can be dangerous and unethical, and that it is necessary to responsibly control and regulate the development of autonomous AI, and in no case allow humans to manage critical infrastructures.

Otherwise, the Terminator blockbuster scenario in which the Skynet neural network seizes power and destroys humanity with the help of nuclear weapons and ends the remaining humans with the help of robots will be quite real.

Previously Focus He wrote that researchers from Stanford summarized the results for 2022 and early 2023 by examining the so-called “AI Index”, which judges that artificial intelligence is causing major changes in society and not always having a positive effect on humanity.