Autonomous agents and global intelligence
Salesforce's recent artificial intelligence (AI) announcements at Dreamforce 2024 and ecosystem These examples demonstrate how the idea of creating "global intelligence" from autonomous agents, inspired by the complexity of living systems, is becoming a reality. As AI continues to advance, fascinating parallels are emerging between biological systems and new trends in distributed intelligence.
Living things as a source of inspiration
In living things, complexity emerges from interactions between simple elements. Each cell of an organism contributes to a functional whole, whether it is a tissue, an organ, or an entire living being. The same is true for autonomous agents : entities capable of acting independently and cooperatively can, together, form a distributed intelligent system.
As shown by the work of John H. Holland, father of adaptive complex systems, complexity emerges "not a centralized hierarchy, but multiple interactions between simple agents, each obeying local rules" (*Hidden Order: How Adaptation Builds Complexity*, 1995). This theory explains why, from simple behaviors, we can see extremely complex and ordered systems emerge, which applies perfectly to autonomous agents in AI.
A classic analogy is ant colonies. Each ant follows simple rules, but their collective interaction produces complex behaviors, such as foraging or building nests. Jean-Louis Deneubourg showed that the organization of these colonies is based on very efficient self-organization principles, which are found today in multi-agent artificial intelligence systems.
Autonomous Agents and Recent AI Announcements: Salesforce and Microsoft
The announcements of Salesforce at Dreamforce 2024 show how agents like Agentforce transform customer interactions. These agents, by automating complex processes and adapting in real time, enable companies to manage large-scale operations, while drawing inspiration from the natural self-organization of living things.
ecosystem, with its advances in Microsoft 365 Copilot et Azure, introduces agents capable of collaborating directly in business processes. These autonomous agents learn from interactions and automate complex tasks, thus creating a form of distributed intelligence within companies.
Current limits and challenges: efficiency, regulation and ecological cost
Despite the promise of distributed intelligence, several challenges persist. First, the effectiveness of autonomous agent systems remains limited by the increasing complexity of the environments in which they operate. Agents must be able to adapt to unforeseen and complex situations, while maintaining high levels of cooperation without central supervision.
Then, regulatory issues arise. How can we ensure that decentralized agent systems always act in the interest of their users? Current legal frameworks are not yet ready to frame this new form of distributed intelligence, and regulating such systems is a major challenge.
Finally, the ecological cost AI remains a major concern. While distributed systems can automate complex tasks, they require massive digital infrastructure, including energy-intensive data centers and networks. To minimize this impact, sustainable solutions such as energy optimization of algorithms and adoption of renewable energy sources will need to be prioritized.
The future of collective intelligence: centralization vs. decentralization, towards the living?
In the future, the distinction between centralization et decentralization will be at the heart of the evolution of artificial intelligence systems. Centralized approaches rely on monolithic models, where a single processing center manages all the information, a bit like a supercomputer that centralizes intelligence. These systems are powerful, but less flexible and adaptive in the face of constantly changing environments.
Conversely, intelligence based on autonomous agents is part of a logic of extreme decentralizationEach agent, like a cell in an organism or a neuron in a network, acts independently and locally, but global intelligence emerges from the coordination and interactions between these agents.
This inevitably reminds us of... “The Library of Babel” de Jorge Luis Borges . In this story, each book contains a part of a larger truth, but it is through the library as a whole that the overall meaning emerges. Similarly, each autonomous agent, taken in isolation, has a limited capacity, but it is through their interactions that a collective intelligence able to solve problems beyond individual capabilities.
Unlike centralized intelligence, this distributed approach provides increased flexibility and resilience. It allows emerging intelligence to constantly evolve through interactions and adapt in real time, without the need for centralized control.
What are your thoughts on this distinction between centralized and decentralized intelligence? Do you believe that the decentralized approach based on autonomous agents is the key to the future of AI?
References :
Kauffman, SA (1993). The Origins of Order: Self-Organization and Selection in Evolution.Oxford University Press.
Holland, JH (1995). Hidden Order: How Adaptation Builds Complexity. Addison-Wesley.
Borges, JL (1941). The Library of Babel. Gallimard Editions.
Deneubourg, J.-L. et al. (1989). “The self-organizing exploratory pattern of the Argentine ant. » Journal of Insect Behavior.
To learn more: our presentation of AI agents and how AI becomes a colleague and transforms the organizational chart.