More is Different: Understanding Emergence and Reductionism

Research into the capabilities of artificial intelligence indicates that the power of these systems is not derived from scale alone. While the industry has long focused on increasing the size of models and datasets, new findings suggest that specialization and cooperation are the primary drivers of AI’s advanced functionality.

This perspective aligns with a scientific and philosophical framework known as More is Different, a viewpoint introduced in 1972 by Nobel Prize-winning physicist Philip W. Anderson. The framework highlights the inherent limitations of the reductionist approach to understanding complex systems.

The Conflict Between Reductionism and Emergentism

To understand why scale is not the sole factor in AI power, it is necessary to distinguish between reductionism, and emergentism. Reductionism is a philosophical approach that seeks to explain complex systems by breaking them down into their constituent parts and analyzing them at a fundamental level.

In contrast, emergentism posits that higher-level phenomena and properties emerge from the interactions of simpler components within a system. This process results in novel and unpredictable outcomes that cannot be explained solely by analyzing the individual parts in isolation.

The core of the More is Different argument is that emergent properties cannot be derived from the fundamental laws that govern elementary particles. This suggests a hierarchical structure of science, where the explainable properties of small-scale systems do not necessarily predict the phenomena that emerge when those systems scale up.

Implications for Artificial Intelligence

The application of emergentism to AI suggests that the transition from a simple model to a powerful one is not a linear result of adding more data or computing power. Instead, the power of AI emerges from how the system’s components specialize and cooperate.

This shifts the technical focus from sheer volume to the quality of interactions within the architecture. If scale alone were the answer, any sufficiently large system would exhibit high-level intelligence; however, the research indicates that the organizational structure—specifically specialization and cooperation—is what enables the emergence of complex capabilities.

Interdisciplinary Applications of Emergence

The More is Different approach is not limited to physics or computer science. Its interdisciplinary perspective is utilized across several scientific fields, including:

• Chemistry
• Molecular biology
• Cell biology
• Social sciences

These fields often encounter phenomena where the whole system exhibits behaviors that its individual components do not. A primary example of this debate is found in the study of consciousness. Reductionists believe consciousness can be explained by understanding neural processes at a lower level, while emergentists view consciousness as an emergent property of complex brain processes.

The debate over causality also differs between the two schools of thought. Reductionism maintains that all phenomena can be explained by reducing them to fundamental components and laws. Emergentism, however, believes that higher-level properties can possess causal powers that are not reducible to lower-level components.

By recognizing that more is different, researchers in AI and other complex sciences can move beyond the limitations of reductionism to better understand how interaction and organization produce the capabilities seen in large-scale intelligent systems.