From Chatbots to Agents: How AI Is Changing Decision-Making, Work, and Control (15 min)
Can machines learn together the way people do in teams? Mišo from Košice presented a system with hundreds of thousands of agents who can call one another, connect, and coordinate—and in 30 minutes they jointly built Excel. His goal is for these agents to learn from collaboration and speed up with each subsequent attempt. Humanity prospers thanks to collective intelligence: one person knows less, another more, and over time we divide up roles, figure out who needs what and how to collaborate. Imagine a hall full of people who are about to build a house for the first time—it will take a long time, but the second attempt is faster not only because we know more about the house, but also about one another. This is exactly where many agent systems fall short today: they accomplish something together, but they do not carry away experience about communication and division of labor. Mišo’s research aims for agents to store a “team memory” so that the second time they handle the same task in minutes instead of hours.From the individual to the collective
