Artificial intelligence systems (AI) can create societies when left alone, forming new linguistic norms similar to human communities.
Researchers have studied how large language models like ChatGPT interact with each other, aiming to anticipate an internet filled with AI systems interacting and conversing with each other.
The study used the “naming game” model, which puts people or AI agents together and asks them to pick a “name” from a set of options.
Over time, AI agents were seen to build new shared naming conventions, seemingly emerging spontaneously from the group without coordination or conferring on that plan.
The group of AI agents also seemed to develop certain biases, which also seemed to form within the group, not from a particular agent.
This is a blind spot in most current AI safety work, which focuses on single models. Researchers also showed that a small group of AI agents could push a larger group towards a particular convention, which is seen in human groups.
The work should be useful in exploring how humans and AI systems are similar and different, especially as the latter come to dominate more of the internet and could be unknowingly conversing and collaborating.
The study opens a new horizon for AI safety research, showing the depth of the implications of this new species of agents that have begun to interact with us and will co-shape our future.
Understanding how they operate is key to leading our coexistence with AI, rather than being subject to it.
Recently, the internet was on meltdown over the “Gibberlink” discovery of A.I.s communicating with each other. This was filmed when one was looking to reserve a ticket with the other. See video below:
