New research by Prof. Ido Kanter examines the relevance of the Nobel laureate Philip W. Anderson's 'More is Different' viewpoint to AI models, asserting that these systems embody the principle through learning and cooperation among specialized internal units.
The findings, published in Physica A: Statistical Mechanics and its Applications, challenge the prevailing industry narrative that increasing the scale and parameter count of AI models is the primary driver of advanced capability. Instead, the research suggests that the true source of AI's power is more nuanced, emerging from a synergetic process where internal components specialize and collaborate. This perspective indicates that future breakthroughs may depend as much on architectural innovation and internal dynamics as on computational brute force, potentially opening new pathways for developing more efficient and capable AI systems.
What We Know So Far
- Prof. Ido Kanter's research quantitatively examines how artificial intelligence models embody the 'More is Different' principle, contrasting with physical systems that predominantly represent 'More is the Same' from an information perspective (techxplore.com, bioengineer.org).
- The study finds AI embodies this principle as a consequence of learning and cooperation among its nodal architecture.
- During training, AI's internal units, or nodes, specialize by taking on distinct roles, such as recognizing specific linguistic features or visual patterns.
- This division of labor among units creates a synergetic mechanism where collaborative dynamics foster cognitive capabilities that exceed the sum of individual node functions (bioengineer.org).
- The research argues that AI's strength lies not simply in its size but critically in the coordinated interaction and information exchange between these specialized components (techxplore.com).
Why AI Specialization and Cooperation Matter More Than Scale
Prof. Ido Kanter's new research applies Nobel laureate Philip W. Anderson's 1972 "More is Different" principle to the advancement of artificial intelligence. Anderson's seminal essay argued that quantitative changes lead to qualitative differences, where complex systems exhibit emergent properties unpredictable from individual components. Kanter's work rigorously applies this lens to AI, suggesting a fundamental distinction between the informational behavior of AI and that of physical systems.
The study asserts that while physical systems largely adhere to a 'More is the Same' principle from an informational standpoint, AI architectures are fundamentally different. As an AI model learns from data, its internal processing units, or nodes, do not simply become more numerous; they undergo functional differentiation. Individual nodes begin to specialize, developing unique roles within the larger network. This process is analogous to a division of labor, where different parts of the model become experts at specific sub-tasks, such as identifying grammatical syntax, recognizing edges in an image, or processing specific types of data.
This internal specialization is the key to unlocking higher-level capabilities. According to a quote in techxplore.com, "Even a single node within a language model can contain meaningful information about the model's overall task." However, the true power emerges from their interaction. "When multiple nodes operate together, their combined capabilities exceed the sum of their individual contributions, demonstrating emergent intelligence in action—More is Different." This collaborative dynamic, according to bioengineer.org, fosters cognitive abilities that are a product of the system's organization, not just its scale.
How Does AI Specialization Improve Performance? A Case Study
The principle that AI power is driven by specialization and cooperation, not just scale, is being demonstrated in real-world applications that prioritize architectural efficiency over sheer size. One prominent example is Heritable Ag, a company focused on accelerating plant breeding. According to a report from agfundernews.com, the company combines AI, genomics, and high-resolution environmental data to make the development of new crop varieties faster and cheaper.
Instead of relying on massive, general-purpose models, Heritable Ag builds efficient, specialized architectures. The company's co-founder and CEO, Brad Zamft, noted that this focus helps avoid the "huge compute spend" often associated with large-scale AI. This efficiency allows them to work cost-effectively on a diverse range of agricultural species beyond the heavily funded staples of corn and soy. Their approach involves creating "digital twins" of real plants, simulating their growth in specific fields using soil and weather data with a resolution of up to 10 meters. This requires highly specialized models that understand the complex interplay of genetics and environment.
The company's AI incorporates the "grammar and syntax of DNA language," enabling it to pinpoint the exact genetic bases that control desirable traits. This deep, specialized understanding has been validated with real plants in real environments, demonstrating high accuracy in both its digital twin simulations and gene discovery across three different species. The success of this targeted, efficient approach supports the idea that well-designed, cooperative systems can outperform larger, more generalized ones. "Artificial intelligence allows us to do better than that," Zamft told agfundernews.com. "We can make more progress faster on multiple traits."
Challenging the Myth of AI Scale Alone
The convergence of theoretical research like Prof. Kanter's and the practical success of companies like Heritable Ag presents a significant challenge to the "scale is all you need" philosophy that has dominated AI development. The research suggests that the industry's focus on ever-increasing parameter counts may overlook a more critical factor: the quality of interaction and information exchange between a model's internal components. According to bioengineer.org, the research argues that AI's strength lies "critically in the pattern of interaction and information exchange between heterogeneous, expert nodes."
This perspective reframes the pursuit of artificial general intelligence. It implies that simply building larger models may lead to diminishing returns if the underlying architecture does not foster effective specialization and cooperation. The study points to a broader conclusion that intelligence in AI may emerge not simply from scale, but from the ability of individual components to specialize, share information, and work together effectively, as reported by techxplore.com. This suggests a future where AI progress is driven by clever architectural design and a deeper understanding of emergent behaviors, rather than an arms race for computational resources.
An approach centered on specialization in AI offers substantial practical implications, leading to more energy-efficient, cost-effective, and accessible technologies. As demonstrated by Heritable Ag, specialized models can solve complex, domain-specific problems without the massive overhead of training foundation models with trillions of parameters. This could democratize access to powerful AI, enabling innovation in fields that cannot support the financial and environmental costs of hyperscale computing.
What We Know About Next Steps
Prof. Kanter's research, which provides a quantitative framework for understanding these principles, was published in the 2026 edition of the journal Physica A: Statistical Mechanics and its Applications. The findings are expected to inform further academic work on the architectural principles of effective AI systems.
In the commercial sphere, Heritable Ag exemplifies the progress of companies applying these specialized AI principles. CEO Brad Zamft states a timeline for their AI-driven approach, aiming to deliver a new strawberry variety to supermarket shelves in four years for a few million dollars. The company is also actively working on a $5 million Gates Foundation-funded project, applying its efficient AI models to improve corn for smallholder farmers in Sub-Saharan Africa and address critical food security challenges.
