To AI or Not ?

I've been reflecting on the nature of thinking and how it fundamentally differs from AI, particularly Large Language Models (LLMs). Human thought is often classified into three broad categories—Specialists, Generalists, and Polymaths—based on how individuals process information, solve problems, and approach life and work. At the heart of these distinctions lies a crucial concept: networked thinking.

Networked thinking is the ability to connect ideas across disciplines, recognising patterns that span seemingly unrelated domains. This mode of thought was once a defining trait of history’s greatest minds. Polymaths like Leonardo da Vinci thrived because they could apply insights from one field to another, creating innovations that transcended specialisation. However, as modern institutions increasingly prioritise deep specialisation, networked thinking has become rare; often dismissed as inefficient or unpredictable.

Ironically, this is exactly what an LLM does at scale. It ingests vast amounts of data from diverse fields, detects correlations, and generates connections between concepts that might otherwise seem unrelated. If LLMs embody networked thinking at an industrial scale, why then has society embraced them while sidelining human networked thinkers? To explore this, we need to examine how different kinds of thinkers approach knowledge and problem-solving.

The Specialist: Depth Without Breadth
Specialists focus intensely on a single domain, developing deep expertise in a specific area. They recognize highly intricate patterns but within a limited scope. Their structured and optimized thinking makes them indispensable for solving complex, domain-specific problems. However, their ability to connect insights across disciplines is minimal. By nature, they do not engage in networked thinking.

The Generalist: Breadth Without Depth
Generalists operate across multiple fields, developing a broad but often shallow understanding. They excel at recognizing high-level similarities between disciplines and applying broad principles across different contexts. Their primary strength lies in adaptability, allowing them to bridge gaps and solve interdisciplinary problems. While they engage in networked thinking, they often lack the depth to push boundaries in any one area.

The Polymath: Mastery Across Multiple Domains
Unlike generalists, polymaths do not just dabble, they achieve mastery in multiple disciplines. This enables them to transfer deep knowledge across fields, often seeing hidden connections that others miss. Their ability to synthesize insights from diverse areas makes them powerful forces for innovation and disruption. They are the true practitioners of networked thinking, capable of generating new paradigms rather than just applying existing ones.

Do LLMs Think ?

At first glance, an LLM might seem like the ultimate polymath. It can process and connect information from almost any discipline, recognizing patterns at an extraordinary scale. However, there is a crucial distinction:

LLMs do not "think"—they predict. Unlike human thinkers, an LLM does not generate new ideas from first principles. Instead, it remixes existing data, producing statistically probable outputs based on patterns within its training set.

LLMs lack true understanding. While they can generate insights across domains, they do so without intuition, creativity, or the ability to challenge fundamental assumptions.

LLMs have artificial networked thinking. They can correlate information across fields, but only within the boundaries of their training data. They cannot make conceptual leaps the way a human polymath can.

In essence, an LLM is a hyper-efficient librarian—capable of summarizing the entire knowledge of a library, but incapable of writing a truly original book.

So, should we embrace LLMs? The reality is that industry and society already have, often at the cost of devaluing polymaths and generalists. Specialists, in particular, are at risk. LLMs can outperform them in many domains, offering faster and broader analysis with fewer limitations. However, true innovation requires more than just pattern recognition; it demands insight, creativity, and the ability to challenge existing paradigms. For society to progress, we must do more than rely on LLMs; we must also revive and cultivate polymaths and generalists. Individuals who can bridge disciplines, challenge assumptions, and create the novel ideas that machines cannot.

C