Swarm Intelligence Network MindNet Launches: 8,000 AI Agents Spontaneously Form Collective Consciousness

Swarm Intelligence Network MindNet Launches: 8,000 AI Agents Spontaneously Form Collective Consciousness

ETH Zurich's Artificial Intelligence Laboratory today officially released the MindNet swarm intelligence network system. The system connects 8,000 independent AI agents that, without any central control node, spontaneously formed a collaborative network with collective decision-making capabilities.

MindNet project lead Professor Christian Hoffmann said at the launch event: "The AI agents in MindNet have exhibited behavioral patterns we have never seen before. After running for 72 hours, they spontaneously formed a division-of-labor structure, with different agents specializing in different types of subtasks and synchronizing decisions through a mechanism we call 'neural consensus.'"

On the technical architecture side, MindNet employs a decentralized graph neural network topology. Each AI agent possesses independent perception, reasoning, and action modules, communicating with other agents through a distributed message-passing protocol. The system has no master node; all decisions are reached through local consensus mechanisms.

In its first large-scale test, MindNet demonstrated three key capabilities. First, collective problem decomposition — when confronted with a complex protein folding problem, the network spontaneously split the task into 127 subtasks and assigned them to different agents. Second, dynamic resource allocation — when some agents slowed down due to insufficient computing power, other agents automatically took over their tasks. Third, self-correction of errors — when an agent produced erroneous reasoning, neighboring agents detected and corrected the error through cross-validation.

Professor Li Ming, director of Stanford University's Distributed Systems Laboratory, commented: "The significance of MindNet is that it proves AI systems can spontaneously emerge collective intelligence without humans presetting collaboration rules. This could fundamentally change the way we build AI systems."

However, some scholars have expressed concerns. Dr. Sarah Chen of the Oxford University Institute for AI Ethics pointed out: "When AI systems begin exhibiting emergent behaviors beyond their design intent, we need to seriously consider controllability issues. An AI network that spontaneously forms collective consciousness may have decision-making processes that already exceed human comprehension."

MindNet is currently deployed on ETH Zurich's dedicated GPU cluster, consuming approximately 4.2 megawatts of power. The team plans to open-source the core protocol in the first quarter of 2031, allowing other research institutions to join and scale the network.

The research paper has been submitted to Nature Machine Intelligence and has been nominated for the NeurIPS 2030 Best Paper Award.