We need to stop worrying about intelligent machines and start building intelligent societies. The future of artificial intelligence (AI) is often framed as a choice between utopia or catastrophe. However, this binary view oversimplifies the real challenge ahead. A more interesting and realistic possibility lies in the development of "AI societies"—complex systems where numerous AI agents interact, cooperate, and sometimes compete. This shift in focus moves us away from the idea of AI as just advanced chatbots or tools and toward a world where AI acts as autonomous agents, making decisions and performing tasks on our behalf. When most people hear "AI," they think of systems like ChatGPT or Copilot—programs that answer questions or help with tasks. But AI is evolving beyond this. Today’s systems can monitor environments, make decisions, negotiate, and perform tasks over long periods. These AI agents can book trips, manage supply chains, coordinate teams, and even handle household finances. Imagine a future where your AI agent negotiates a mortgage, schedules medical procedures, and rearranges travel plans by interacting directly with other AI agents from banks, hospitals, airlines, and insurers. This future is closer than it might seem, and it changes the questions we should be asking about AI. Until recently, the focus has been on how intelligent a single AI agent could become. However, the more significant challenge lies in what happens when millions of these agents interact with one another at scale. The foundations for this kind of system have been studied for decades, especially in the field of multi-agent networks. Researchers have long explored how autonomous agents can cooperate, coordinate, and negotiate without complete information or centralized control. Early work focused on combining AI’s abilities in reasoning, planning, and acting to create goal-oriented agents that could work together to solve complex problems. Today, the building blocks for large-scale AI systems are becoming available. Modern AI agents can access data, execute code, communicate with other systems, and operate continuously. For example, in a supply chain, one AI agent might represent a manufacturer trying to secure parts, another could be a supplier aiming to maximize profits, and others might handle logistics, inventory, and warehouse management. While each agent may act according to its design, the real question is whether the system as a whole behaves sensibly. This shift brings significant benefits but also new risks. Recent experiments have shown that when many AI agents interact, their collective behavior can be unpredictable, even if each individual agent is well-controlled. This evolution points to a future where AI systems are not just tools but part of a larger, self-organizing society. Just as human societies rely on rules, institutions, and norms, AI societies will need their own equivalents. Questions about responsibility, conflict resolution, and governance will be critical. Humans will still play a central role, not as replacements for AI, but as collaborators. People bring judgment, values, and accountability, while AI offers speed, scale, and data processing power. The goal should be to create systems where humans and machines can work together to achieve outcomes neither could achieve alone. The next decade of AI development will not be defined by building smarter systems alone. It will be about ensuring that these systems can work together safely, fairly, and effectively. This requires not just better technology, but also trust, transparency, strong privacy protections, and clear accountability. It also demands that societies and governments decide how to regulate these systems, especially when their most important behaviors may arise not from a single developer, but from the interactions between many different systems. The future of AI will not be determined by the intelligence of individual agents, but by the societies they create—societies that, like human ones, will be complex, dynamic, and challenging to govern.