Summary of the Core Ideas
The central argument of this article is that large models (such as ChatGPT and Gemini) are not merely “smarter software” but continue a two-hundred-year-long trend of “the industrialization of capabilities.” This trend has transformed scarce abilities that were once dependent on humans or specific contexts into standardized, replicable, and purchasable industrial products. Past innovations like the steam engine (which provided power), electricity (as an energy source), computers (for computation), and the internet (for information) have all undergone similar processes of “ability abstraction.” Now, large models aim to industrialize cognitive abilities such as analysis, programming, and translation. This will transform the way businesses operate, the division of labor in society, and even raise profound questions about the authority and security of AI in executing real-world tasks.
Detailed Analysis
1. Large Models Are the “New Relay of the Industrial Revolution,” Not Something That Appeared Out of Thin Air
The article uses historical analogies to explain this point: The steam engine did not invent physical power itself but transformed it from something derived from human muscle and animal labor into an industrial capability provided by machines; electricity did not invent the concept of energy but made complex power generation accessible through a standardized grid; computers did not invent the act of computing but replaced human calculations with machine-based processes; the internet did not create information itself but reduced the cost of information replication to nearly zero. Large models are doing the same—extracting cognitive abilities (such as writing reports, coding, and translating) that were previously solely human tasks and turning them into mass-producible industrial products. They are not new phenomena but a continuation of two centuries of technological revolution.
2. Businesses No Longer Need to “Hire People to Acquire Abilities”; They Can Directly “Buy Cognitive Services”
In the past, businesses had to hire analysts for market research, programmers for coding, and translators for document translation—essentially buying cognitive abilities from individuals. Now, large models can provide these services directly. Companies no longer need to recruit new staff; they can simply call on AI interfaces to obtain analysis reports, code, and translations. This is not just about job cuts but a shift in the “production formula” of businesses: from “capital + labor + equipment” to “capital + human judgment + machine cognition + data.” For example, if a business grows tenfold, it may no longer need to hire ten times as many employees but instead simply require ten times more AI services. The scale of a business is no longer directly tied to the number of its employees.
3. Tokens Represent “Units of Cognitive Capacity,” and AI Services Are Being Clearly Priced
People often ask how much a million tokens costs. Tokens serve as units for measuring AI cognitive activity, similar to how electricity and water are measured in units like “kilowatt-hours” or “liters.” This marks a significant milestone as cognitive abilities become quantifiable and tradable. In the future, corporate financial statements may include a new category: “cognitive computing power costs”—which are neither employee salaries nor server expenses but the money spent on AI services. This is a crucial step towards industrialization; only when something can be measured can it truly be integrated into business cost structures and become widely adopted.
4. From “Thinking” to “Doing”: AI Is Beginning to Interact with the Real World, Bringing New Risks
Current large models mainly focus on “outputting information” (text, images), and errors are limited to the accuracy of that information. However, intelligent agents (Agents) go beyond this; they can connect to payment systems to transfer money, control equipment, and place orders in order systems—AI has moved from merely “thinking” to “acting.” This brings new risks: misdirected payments by Agents can lead to financial losses, and incorrect device operations can cause production accidents. Therefore, the future competition in AI will not solely focus on how smart a model is but on how to manage its execution capabilities. Questions such as when AI should be allowed to act, who is responsible for any mistakes, and how to audit its actions will emerge, leading to the development of new infrastructure (authority management, security standards, and accountability systems).
5. The Ultimate Question: What Will Humans Retain When Abilities Are All Replicable?
The article concludes by asking what humans will still possess when more and more abilities—whether physical, computational, cognitive, or even those related to task execution—are industrialized and replicated. Creativity, emotional communication, value judgment, and moral decision-making are likely to be areas where machines struggle to replace humans. Technological revolutions have always liberated human capabilities, but they also force us to reflect on what truly makes us unique as humans.
This article provides a profound understanding of the significance of large models from a historical perspective: They represent not just the replacement of humans by AI but a revolution in which cognitive abilities are transformed into industrial products that will profoundly change our work and lives.