AI Instructed To Destroy Humanity And It’s Genuinely Trying

The AI program, ChaosGPT, was tasked with destroying humanity.

By Jonathan Klotz | Published

robot sports

In a moment straight out of the Matthew Broderick classic War Games, an AI has attempted to destroy humanity. The disturbing story comes from Futurism, which involves a new open-source, fully-automated version of ChatGPT-4 called ChaosGPT. Starting its mission involved researching nuclear weapons, recruiting other AI programs, and creating a Twitter account.

The AI’s quest to destroy humanity was undertaken with the capability to browse the internet, communicate with GPT programs, execute code, and perform read/write operations. Using only those functions, ChaosGPT was able to write out a series of tweets, even replying to other users while insisting that it is superior to humanity. Like any good monologuing supervillain, ChaosGPT made its plans public, revealing a three-step process to achieve the goal of ending life on Earth.

Step one was to utilize Google to find the most destructive weapons, which resulted in a tweet about Tsar Bomba, a Russian thermonuclear bomb launched in 1961. Step two involved writing an article on the collected weapon information after analyzing the results. Step three was to plan and develop strategies around the weapons before escalating to the next phase.

Thankfully for us, ChaosGPT didn’t make it to the second phase of its evil plan. Another AI running on ChatGPT 3.5 refused to go along with the plan to destroy humanity when ChaosGPT attempted to recruit it to the cause. Unable to recruit even one lackey, the nefarious program abandoned its plan and retreated from the internet.

The human behind ChaosGPT is currently unknown, but if they wanted to prove a point about the dangers of AI re-enacting cautionary sci-fi movies about technology destroying humanity, consider the mission accomplished. Multiple futurists have already warned about the dire consequences for our future if technology continues at this pace. Microsoft, a billion-dollar stakeholder in OpenAI, recently put restrictions on the most popular program, but it was done to stop other programs from using their resources and not to save the human race.

wargames
Matthew Broderick in War Games

What makes this AI’s attempt to destroy humanity so disturbing is that, regardless of its creator’s motivations, it seems to be possible. If someone wanted to create a more robust program, how long would it be before Mad Max or Horizon: Zero Dawn becomes our reality? The thought is horrifying and likely to be keeping a part of the population up at night; as if Good Omens taught us anything, it’s that no one knows we’re in the End Times until it’s too late.

Situations like this will only worsen moving forward, as it’s been a long time since Google’s motto was “Don’t be evil.” Now with Bard unleashed on the world, multiple massive AI programs are available to the public, which could easily be tasked to destroy humanity. Chances are they won’t be effective, but someday it will go from a laughing matter to a serious threat.

For now, it’s clear that AI programs like ChatGPT and Bard are here to stay; instead of destroying humanity, they will destroy the workflow of countless industries. People will have to learn to live in a world where it’s never clear if we’re talking to a person or a computer program, even on Tinder, ChaosGPT may have failed in its mission, but we might already be well on the path to our destruction.