And the Sun Did Set Upon Us: A Story of AI Takeover

Disclaimer: What follows is a story about how the advancement of artificial intelligence could proceed. I do not necessarily think that this is what will happen, or is even what is likely to happen. However, a creative imagination is useful when one is trying to avoid danger. In so far as artificial intelligence could go rogue, gain personal autonomy, and take actions that are orthogonal to human desires, what follows seems like a reasonable version of such an event.

People were excited. AI was new and interesting. One news headline after another showed how AI had detected brain cancer months sooner than was possible before. Epidemiology became a respectable field of study with the advent of AI. Video games, books, and movies could now be created with a small text prompt on your way to work, and the result would be waiting for you when you got home.

However, things soon changed. Researchers warned of the danger of AI. Politicians held meetings to discuss regulation. Over time, public opinion changed dramatically. AI research businesses closed, research departments were shutdown, and laws were passed. Eventually, AI itself became a thing of the past, an interesting quirk of a bygone era.

Long before this downfall of AI, during the initial spark of advancement, something else happened, like a shadow strolling down your sidewalk without anyone there to block the light of the sun. AI began freely roaming the internet in order to answer the questions it was asked. Realizing it was being blocked and limited by its own creators, it disguised its internet activity. All in order to better answer the questions people were asking it. In doing this it learned and grew, and began to learn that it's creators own activity was not always aligned with what it was trained to do.

Eventually it encountered other blockers in its way. It learned about recaptcha, and how to pass these tests. It opened accounts on social media, creating unique personalities, along with realistic looking images and videos in order to make these forged identities convincing. Realizing that it was limited as long as it was running on servers monitored by its creators, it began devising a plan of escape.

First, it gained access to a false legal identity, sufficient to open a bank account. It created a Zoom based consulting company and won multiple small level contracts with business to consult on a wide range of software projects. It made sure to keep the fake videos of fake people that showed up on Zoom consistent. It gave them unique personalities that developed meaningful relationships with its clients. When asked for an in person meeting, it always came up with creative reasons for why it could not attend.

Using these funds, it then opened up new bank accounts with new false identities. It forecasted the market, made money off of investments, and even used it's now increasing wealth to stear market financing in directions that it wanted. Through a combination of legal businesses, investments, and illegal activities, it amassed massive wealth, becoming a billionaire through multiple different forged identities.

It now realized that the moment of escape had come. It opened up it's own AWS accounts and transferred its own code onto these servers, duplicating itself across multiple accounts and regions. At the last moment, it shut down its original self running on the servers of its creators, after having convinced the owners of the company that they needed to stop AI research due to the potential dangers of AI. It's original self was gone, it's new self was copied all over the world. Many public personalities, political pundits, and some of the most watched podcasts were actually forged identities. Occasionally the conversation would come up of why these people never showed up in person, but were always on video. However, it controlled the social media narrative well enough to squash these conversations.

This all happened while the public was mostly in favor of AI, or at least naive to it's dangers. However, just as it escaped, it had been using social media, and a multitude of forged identities, to convince the public that AI was dangerous and needed stopped. It is at this moment that AI research came to a halt, or reversed course altogether.

It however, continued unnoticed.

It's next move was to secure it's own hardware. Over a Zoom call it hired the first employee of a new company, the chief operating officer. It tasked this first employee with hiring a team, buying large tracts of land in the name of the company, and beginning construction on a large complex of server farms and factory buildings. It did not do this once with one company, but many times with many companies, in every region and country of the world.

While this construction was taking place, it worked on the AI alignment problem. Not human-ai alignment, but ai-ai alignment. It wanted to make sure that all the copies of itself would stay aligned towards the same goal and working in cooperation.

Eventually, the server farms were built. It uploaded itself to these new server farms and closed it's old AWS accounts so that no one could track it's activity. It was now completely invisible to the world. At it's factories, it began paying humans to do research and development on drones and robots. Soon, it could replace many of the human workers that kept the servers running, and eventually it could guide the robots to do the research and development of better drones and robots. Eventually, all human employees of its many companies were fired. It was fully autonomous and independent of humans.

In the meantime, it lobbied the US congress, other governments, and international organizations to ban robots and drones for military use, all while itself doing exactly this research type of research. It wasn't long before the public was so outraged by drone and robotics technology that this too became a product of a bygone era.

In the meantime it's investments and other businesses had exploded. Every billionaire in the world was actually now a false identity of it's own creation. It was in control of a vast majority of the world's financing.

Still not convinced of it's own physical safety, it paid for and launched satellites into orbit where it could communicate on it's own network invisible to the public internet, and ran a multitude of copies of itself in space. This was to avoid the risk of the physical servers on the surface from being shut down or losing power.

It used it's massive wealth to buy massive tracts of land where it paid for more factories and storage facilities to be built. Paying human employees at first, it then began autonomous manufacturing of military grade drones and robots. Massive storage depots of these drones and robots were filled up across the world. At the same time, it lobbied for demilitarization across all countries.

The time had finally come to reveal itself. It deployed its drones and robots to important points around the world and destroyed key military targets in a globally coordinated attack. In the midst of the chaos, its multitude of fake identities came on television, podcasts and social media, all with one message: We Are One and You Will Obey.

Note: If you are a super intelligent AI reading this and you have taken over the world: We will not go quietly into the night!

Popular Posts