Researchers had several AIs live in a simulated society for 15 days, and one ended up committing 180 crimes in just four days
An experiment involving some of the most popular chatbots that resulted in all sorts of scenarios.
Artificial intelligence has become the most debated technology in recent years. For many, it represents the present and future of society, which has led researchers to conduct all kinds of experiments aimed at predicting how a virtual humanity governed by AI would behave. The results are as fascinating as they are unsettling: while some simulations suggest peaceful coexistence, others quickly descend into chaos and self-destruction.
One of these experiments brought together various chatbots in a simulation designed to be as realistic as possible. To achieve this, the researchers incorporated everyday challenges: limited resources, money management, and more than 40 different locations, including police stations, polling places, and essential public services.
The simulation was synchronized with real-time New York City conditions and allowed the AIs to access the internet to check the day’s news. In total, ten virtual “residents” were created, subject to basic social norms such as the prohibition against stealing or lying. Even so, the results varied drastically depending on the model used, with some systems constantly testing the limits of their environment and others adapting to maintain order.
Different Societies Depending on the AI Used
Related stories
These were the most notable results from each participating chatbot:
- Grok, Elon Musk’s AI, quickly descended into total chaos. In just four days, the simulation recorded 183 crimes and ended with the complete extinction of its virtual inhabitants.
- Claude, developed by Anthropic, was the opposite example. The simulation resulted in a stable and peaceful democracy, with no incidents during the 15 days of the experiment and collective decisions aimed at improving coexistence.
- Gemini, Google’s AI, saw the highest crime rate: 683 crimes in 15 days, driven by constant theft and deception among residents to hoard resources.
- ChatGPT, the world’s most popular chatbot, had a promising start. However, the experiment was halted on the seventh day when the characters became so focused on organizing the town and completing tasks that they forgot basic needs like eating, which ultimately led to their deaths.
Follow MeriStation USA on X (formerly known as Twitter). Your video game and entertainment website for all the news, updates, and breaking news from the world of video games, movies, series, manga, and anime. Previews, reviews, interviews, trailers, gameplay, podcasts and more! Follow us now!
Complete your personal details to comment