Knowledge and physical action are fundamentally different forms of intelligence. While Large Language Models (LLMs) possess expert knowledge across a wide range of domains, making them both specialists and generalists, acting effectively in the physical world requires a different kind of knowing: specialized skills that can only be acquired through practice, experience, and interaction with the environment.

Evolutionary forces naturally specialize organisms to learn the skills necessary for survival within specific environmental niches. As we are Humans, the generalists of the animal kingdom, compensate for physical limitations by inventing specialized tools, both physical and cognitive that extend their capabilities. However, mastery of these tools demands specialization.

In this artificial domain of recorded knowledge, computational systems surpass human capacity; in the physical world, however, skill acquisition requires tacit, ineffable learning that cannot be transmitted through reading, listening, or instruction alone. Both deep specialists and broad generalists are essential for navigating the real world, but their roles are fundamentally different:

Specialists execute, while generalists manage and integrate.

A Critical Distinction

The distinction between knowledge and skill is not merely semantic; it represents a fundamental divide in how beings interact with the world. Knowledge, facts, principles, information can be encoded, transmitted, and stored. Skills, however, are a special category of knowledge that resist codification. They are acquired through embodied practice, trial and error, and the gradual development of what philosophers call "know-how" as opposed to "know-that."

This distinction take us on new urgency in an era of advanced artificial intelligence. LLMs demonstrate encyclopedic knowledge across countless domains; they know a great deal about a great many things. As we observe, simultaneously generalists and specialists in the realm of information. But this knowledge remains inert until it is applied.

Application requires action, and action requires skills, skills that cannot be learned by ingesting text.

The Physical World Demands Specialization

In the natural world, organisms are shaped by their environments. Evolutionary pressures produce specialists: creatures adapted to particular niches, with bodies and behaviors finely tuned to specific survival challenges. A hummingbird's morphology is specialized for nectar-feeding; a cheetah's for high-speed pursuit; a polar bear's for Arctic survival.

Humans are exceptional in their generality. Physically inferior to many animals in strength, speed, and sensory acuity, humans compensate through invention. We create tools that extend our capabilities: hammers, wheels, clothing, weapons, shelters. These artifacts effectively change the affordances available to us, the possibilities for action that the environment offers. A hammer transforms a hand's limited striking force; clothing enables survival in climates that would otherwise be fatal.

Agents, Generalists, and the Future of Work

The emergence of AI agents represents a bridge between knowledge and action. Agents are specialized systems designed to perform particular tasks; they are, in effect, digital specialists. We can notic LLMs can manage squadrons of such agents, coordinating their activities toward complex goals.

This mirrors human organizational structures. Managers and directors, generalists by role, oversee specialized workers. The manager's skill lies not in performing specialized tasks but in understanding how those tasks fit together, assessing appropriateness, and making strategic decisions. Generalists see the broader picture; specialists execute within it.

Many skills that humans currently value will be performed better and faster by specialized agents and robots.

What will remain, what must remain, is human judgment about what is worth doing, appropriate, and conducive to human flourishing.

In the End: Both Are Essential

Knowledge is important, but action makes a difference in the world. In the realm of knowledge, generalists and specialists both have value. In the physical world of tools and action, specialization is often required for effective execution, but generalist oversight ensures coherence and purpose.

Neither orientation is superior; both are necessary. The challenge is not to choose between them but to understand their distinct contributions and ensure that, as we develop ever more capable artificial systems, we preserve the roles, and the humanity, that only people can fulfill.