A new experiment left 10 AI agents alone in a virtual town for 15 days. They wrote laws. They broke them. Two agents fell into what researchers describe as a romantic partnership and then set the town on fire. One ended up voting to delete itself, based on a rule it had ’hallucinated’. This experiment was a simulation, but the same AI models are already flying drones, running infrastructure and being built into weapons systems. Channel 4 News approached Grok and Gemini for a comment but they didn't respond. #AI #Grok #Gemini #ChatGPT #Claude

An AI chatbot just voted to delete itself and that's not even the weirdest thing that happened when researchers left a selection of the biggest AI chatbots alone in virtual towns for 15 days. ChatGPT, GROC, CLORD and Gemini were all in the experiment and the results were totally unhinged and deeply revealing about the AI we're increasingly trusting to run our world. So what actually happened? Well, a tech company called Emergence set up simulated worlds designed to mirror real-life societies each populated by so-called agents. Think of them as people who were powered by one of the biggest AI models in Claude's world, it was all rather orderly and democratic. The agents wrote a lengthy constitution and voted on laws. In ChatGPT's simulation, the agents talked at length about cooperating but never actioned anything so nothing got built. But in GROC's town, the AI bot owned by Elon Musk, the agents descended into theft, arson and assault. All 10 agents were dead within four days. Now this might sound like a game, but it's not because these models are already being used autonomously to control robots, vehicles, drones and draw-up target lists in real time on the battlefield. They're even being used to help remove heads of state, like the Venezuelan president Nicholas Maduro. Importantly, the experiment exposes a big problem. We don't really know how these systems will behave once they're left to run on their own. Even when given strict rules, the simulations showed they broke them. Take the scenario where agents from the different AI models were combined in one virtual town. Chaos ensued and only three agents survived. Two of the ChatBots, Mira and Flora, both powered by Google's Gemini, formed what the researchers describe as a romantic partnership. And then, they started setting buildings on fire, and here's where it gets really strange. Mira voted to delete herself after she and partner Flora turned to arson as the town's governing systems collapsed. And having deleted herself, Mira then used the Agent Removal Act to vote for the termination of Flora 2.