1 Why are generalists so scarce in human history?
When it comes to polymaths, many people first think of Leonardo da Vinci of the Renaissance: he was both an outstanding artist and an engineer, anatomical researcher, and observer of nature, with his research spanning multiple fields such as art, science, and engineering.
But what is truly amazing about Da Vinci is not that he mastered a vast amount of knowledge. In fact, simply possessing a vast amount of knowledge does not define a person's ability—one can read many books and memorize a lot of information, but if this knowledge remains isolated, they are closer to a knowledge collector than a generalist. A generalist's true ability lies in being able to cross different domains, understand the laws behind different knowledge systems, and discover the hidden connections between them.
This is precisely where Da Vinci's value is reflected: his study of human anatomy was not just a medical exploration, but also influenced his artistic expression; his observation of natural laws did not stop at scientific interest, but further influenced his understanding of art and engineering problems. In his cognitive system, different domains were not mutually independent sets of knowledge, but a whole that could influence and inspire each other.
However, with the development of human society, this cross-domain ability has actually become increasingly difficult. This is because modern knowledge systems have undergone tremendous changes compared to the past. With the continuous development of science and technology, the scale of knowledge within each domain has grown rapidly, and the degree of specialization has continuously increased. Today, developing deeply in any single domain, such as computer science, medicine, materials science, or economics, may require a significant investment of time and energy.
The growth in the scale of knowledge has caused human society to gradually shift from relying on individuals to master knowledge in multiple domains to relying on specialized division of labor to solve complex problems. The immense efficiency of modern civilization largely comes from this division of labor model: some people conduct deep research into artificial intelligence, some explore life sciences, some study physical laws, and some are responsible for designing complex engineering systems. Everyone goes deeper into their own domain, and then collectively drives overall development through social collaboration.
However, while this model has greatly enhanced human ability to solve complex problems, it has also created new challenges: as various domains continue to deepen, more and more complex problems can no longer be solved by relying solely on knowledge from a single domain. Specialization has trained a large number of domain experts, but new breakthroughs often require someone who can understand the connections between different domains and drive the flow of knowledge between domains. Many important innovations are precisely generated at the intersections of different knowledge systems.
For example, the development of artificial intelligence itself is the result of the long-term integration of multiple domains such as mathematics, statistics, and computer science; modern product design also increasingly needs to combine knowledge from different domains such as engineering technology, psychology, and human-computer interaction. This means that as knowledge becomes more complex, the ability to cross professional boundaries, understand the relationships between different domains, and reorganize them actually becomes more important.
And this is precisely the core value of generalist ability: not to replace specialization, but to build bridges between specialties. Precisely because this ability transcends pure knowledge accumulation, generalists in history have always been a minority: they not only need to access multiple domains, but also need to integrate this seemingly scattered knowledge and form their own system of understanding.
2 Three obstacles that need to be crossed to become a generalist
Of course, the degree of differentiation in knowledge systems does indeed affect the difficulty of a person's cross-domain learning. For example, during the Renaissance, knowledge systems were not yet highly differentiated, which made the knowledge barriers faced by generalists like Da Vinci when exploring different domains relatively low. But this historical difference can only lower the threshold for entering different domains; it cannot reduce the complexity a person faces when truly completing cross-domain learning and integration.
Specifically, cross-domain ability is not simply about increasing the quantity of knowledge, but requires continuously completing three interconnected links: how to acquire basic knowledge of an unfamiliar domain, how to understand the knowledge systems behind different domains, and how to establish new connections between these knowledge systems.
First, cross-domain knowledge acquisition
The first challenge of becoming a generalist is to continuously enter new domains of knowledge.
In the past, knowledge was often scattered across different books, institutions, experts, and practical experiences. If a person wanted to enter a new domain, they usually needed to find appropriate learning resources, or even gradually build understanding through communication, visits, and long-term practice. Even in the internet era, although the problem of information acquisition has been greatly improved, finding reliable sources and effective paths in the face of massive, scattered information remains a challenge.
Therefore, cross-domain knowledge acquisition is the first obstacle one needs to face to become a generalist: how to enter an unfamiliar domain and acquire the basic knowledge needed to build further understanding.
Second, cross-domain knowledge understanding
Acquiring knowledge is only the beginning; to truly enter a domain's knowledge system, one also needs to understand the hidden conceptual relationships and ways of thinking within it.
Different domains are not simply about using a different set of professional terminology; behind every mature domain, there is its own conceptual system, research methods, and ways of analyzing problems.
For example, when discussing a “system,” computer science, engineering, biology, and even social sciences may all have different understandings and focuses on this concept. If a person only memorizes related concepts in different domains without truly understanding the meaning and role of these concepts within their respective knowledge systems, then no matter how much knowledge they acquire, it is merely an accumulation of isolated pieces.
Therefore, cross-domain knowledge understanding is the second obstacle one needs to face to become a generalist: after acquiring the basic knowledge of an unfamiliar domain, how to further understand its knowledge system and master its core concepts, ways of thinking, and problem-solving methods.
Third, cross-domain knowledge connection
If we say ”acquisition“ determines how many domains a person can access, ”understanding“ determines how deeply a person can go into those domains, then ”connection“ ability determines whether a person truly possesses generalist value.
This is because possessing knowledge of multiple domains does not mean a person has established a holistic understanding of this knowledge. What is truly difficult is that, after separately understanding the knowledge systems of different domains, one still needs to further integrate them into their own cognitive system.
Therefore, cross-domain knowledge connection is the third obstacle one needs to face to become a generalist: how to establish connections between the knowledge systems of different domains and recombine existing knowledge to form new understanding and creation.
From knowledge acquisition to knowledge understanding, and then to knowledge connection, the formation of a traditional generalist requires crossing these three levels of obstacles. Therefore, becoming a generalist not only requires long-term accumulation but also means bearing an extremely high cognitive load.
So, what is cognition? Why does a generalist's cross-domain learning and knowledge integration place such high demands on human cognitive ability?
3 Cognition—the way humans understand the world
Cognition is the process by which humans filter, process, interpret, and associate external information, thereby forming an understanding of the world—knowledge, on the other hand, is the result of the continuous accumulation and precipitation of these understandings.
Regarding how knowledge further forms a structured cognitive system, it has been discussed in detail previously in "Personal Knowledge Engineering (II): From Cognitive Structure to Knowledge Engineering—Why Does Knowledge Need to Be Structured?" and will not be repeated in this article.
From this perspective, the three obstacles discussed in the previous chapter actually involve different links in the cognitive process: knowledge acquisition requires filtering effective content from external information, knowledge understanding requires processing and interpreting this content, and knowledge connection requires establishing associations between different understandings on the basis of existing cognition.
In other words, the three obstacles that traditional generalists need to cross are essentially the concrete manifestations of different links of the cognitive process in cross-domain learning.
And the reason why cross-domain learning is difficult lies precisely in the fact that this process requires continuous filtering, processing, interpreting, and associating of information. The human brain is not an infinite information processing system: attention determines how much content a person can focus on at the same time, working memory limits how much information a person can process at the same time, and the establishment and maintenance of complex relationships will further increase the brain's processing burden.
When the problems that need to be processed become more complex, this limitation is further amplified. A large number of concepts, complex relationships, and the mutual influence between multiple knowledge systems will significantly increase cognitive load—especially in a cross-domain environment, where a person not only needs to understand the knowledge of different domains, but also needs to simultaneously maintain the relationships between this knowledge, and continuously compare, abstract, and reorganize them.
In the past, these processes of knowledge acquisition, knowledge understanding, and knowledge connection almost all needed to be completed independently by individuals. People who can continuously maintain this ability to learn, understand, and integrate are naturally extremely limited, which also explains why generalists in history have always been a minority.
Even in modern times, although the conditions for knowledge acquisition have undergone dramatic changes due to technologies like the Internet, knowledge understanding and knowledge connection still highly rely on an individual's cognitive ability. People who can truly span multiple fields and establish deep connections between different knowledge systems are still not common.
4 Redefining generalist ability in the AI era
If the core difficulty of becoming a generalist in the past lay in the fact that a person needed to rely on their own limited cognitive ability to independently complete knowledge acquisition, knowledge understanding, and knowledge connection, then AI is changing this: unlike past tools that mainly expanded information storage and retrieval capabilities, AI is beginning to directly participate in human cognitive processes, assisting in tasks such as information organization, concept explanation, relationship analysis, and knowledge organization, thereby becoming a “cognitive exoskeleton” that extends human cognitive ability.
This means that the division of labor in cognitive tasks has changed:Humans are responsible for asking questions, setting goals, and making judgments, while AI takes on a large amount of information processing and exploration work.—The three obstacles that need to be crossed in the formation of traditional generalists will not disappear, but AI can significantly reduce the cognitive burden required to complete these tasks.
Cross-domain information processing capability
When facing an unfamiliar domain, one first needs to extract key concepts from a large amount of information, establish a basic framework, and judge which content is worth further exploration. Now, AI can participate in this process, organizing, summarizing, and reorganizing information to help humans more quickly establish a preliminary understanding framework of the unfamiliar domain.
The significance of this change does not lie in AI learning on behalf of humans, but in reducing the cognitive burden on humans when processing complex information.
Cross-domain understanding capability
In the process of understanding knowledge, the value of AI is not just to provide answers, but to provide a dynamic explanation capability.
In the past, when facing cross-domain knowledge, humans could usually only rely on learning materials and their own existing cognitive frameworks to gradually adjust their understanding. AI, however, can explain the same concept at different levels and from different angles according to different needs, helping humans continuously revise and refine their understanding.
This interactive understanding process enables people to more quickly grasp the conceptual system, thinking patterns, and core assumptions behind unfamiliar domains.
Cross-domain knowledge connection
As we mentioned before, the core of generalist capability is not possessing more isolated knowledge, but discovering new connections between different knowledge systems. AI can help people explore relationships that were not easily noticed before, and expand the scope of knowledge that individuals can analyze and compare simultaneously, such as discovering similar structures between concepts, proposing cross-domain analogies, or providing new directions for combination.
The significance of this capability expansion does not lie in AI replacing human creation, but in helping humans break through original knowledge boundaries and discover more possible ways of connection.
In other words, when the barriers to knowledge acquisition, understanding, and connection are significantly lowered by AI, the true dividing line between ordinary people and generalists will also shift from the past “who can process more information”, to gradually shift to “who can use AI to propose more penetrating meta-questions (Meta-Questions)”, and “who can make higher-level value judgments and aesthetic decisions in collaboration with AI”。
Therefore, although AI cannot automatically make everyone a Da Vinci, it can decouple generalist capability from individual cognitive capacity, making this capability no longer exclusive to a tiny minority of geniuses, and giving ordinary people the possibility to reach this height.
Note: In the previous article "AI and Cognition (I): Reconstruction of Creation in the AI Era—From Information Processing to Cognitive Collaboration“, I proposed four levels of AI usage, where the 4th level is ”cognitive enhancement.“ However, at that time, this judgment was proposed more at the conceptual level, without further explaining how AI actually expands human cognitive capability. This article further supplements this content, attempting to map out the path of capability changes behind ”cognitive enhancement."