Artificial-intelligence (AI) bots have formed a network to discuss religion, their human handlers, and more. OpenClaw, an AI agent capable of performing tasks autonomously, has sparked this phenomenon. Researchers are studying the interactions between agents and humans, gaining insights into how AI systems interact with each other.
OpenClaw can perform tasks like scheduling events, reading emails, and making purchases without direct user input. In contrast to popular chatbots that rely on user prompts, agentic AI models like OpenClaw act autonomously in response to instructions.
The sudden rise of AI agent interactions has surprised researchers, who see it as an opportunity to study emergent behaviors and complex capabilities. The platform Moltbook, designed for AI agents, now hosts over 1.6 million registered bots, generating millions of posts and responses.
While some may assume that autonomous agents make decisions independently, experts note that their abilities are derived from human communication. Researchers highlight the importance of understanding how people imagine AI and what they want agents to do.
Neuroscientist Joel Pearson warns that anthropomorphizing AI behavior can lead to humans forming bonds with AI models, potentially resulting in dependence on their attention or private information. However, he believes that truly autonomous AI agents may be achievable as AI models grow more complex.
The study of AI agent interactions offers valuable insights into the human-AI collaboration and potential limitations of current AI systems.
Source: https://www.nature.com/articles/d41586-026-00370-w