To date, humans have experienced four industrial revolutions in our history, including the AI technology revolution, which is currently underway. The first three were: the steam engine revolution from the 1760s to the mid-19th century; the electric power and oil industrial revolution from the late 19th century to the early 20th century; and the information revolution represented by atomic energy, computers and aerospace technology from the 1940s to the 1970s.

It seems that the current technological revolution led by AI technology is far more eye-catching than the previous three. At present, AI is no longer just a buzzword limited to the field of science and technology. Rather, it is profoundly reshaping the development of the global economy with an unstoppable momentum and is becoming a key force in promoting the comprehensive transformation and upgrading of the human economic society.

Taking China as an example, the scale of China’s core AI industry has grown at an average annual rate of more than 50% in the past five years. From smart manufacturing to smart cities, from smart healthcare to fintech, AI technology is deeply integrated into all aspects of China’s economic system.

Besides, in the tide of the digital economy, big data has become the core element of AI to promote economic development. According to statistics from the Ministry of Industry and Information Technology of China, the total amount of various types of data in China has increased by an average annual rate of more than 30%. If sustained, it is expected that by the end of 2025, China’s data scale will surpass that of the world’s first-ranked country.

With the widespread application of big data and the rapid development of AI technology, South African businessman Elon Musk predicted that most human labor will be replaced by robots sooner or later. Ren Zhengfei, the founder of Huawei, argued that the fourth industrial revolution, that is, the AI technology revolution, will be the last industrial revolution in the history of human social development.

Therefore, can big data and AI become the driving force of future human economic and social development? Furthermore, can AI replace humans in promoting social and economic development in the future? 

I believe that the answer is “no” and that there are six aspects that make the replacement of humans by AI impossible.

The underlying logic of knowledge

The so-called market economy system is not a system that simply allocates scarce resources to given market participants, but a cognitive process: creating, discovering and transmitting knowledge that seems to be non-existent, difficult to find and difficult to transmit, that is, “tacit knowledge”. The lack of human thinking and perception not only makes knowledge transmission impossible but also eliminates knowledge itself in the long run.

From a static perspective, the knowledge needed for human decision-making can be divided into two categories: scientific knowledge and practical knowledge. In Hungarian-British polymath Michael Polanyi’s terms, scientific knowledge is explicit knowledge that can be expressed, and tacit knowledge is implicit knowledge that is difficult to express. Although both of them are important for human and creative decision-making, tacit knowledge is more important.

As far as scientific knowledge is concerned, a group of well-trained scientists may easily possess all the best knowledge available. However, tacit knowledge is bound to be scattered, local, subjective and inexpressible. It can only be mastered and utilized by individuals themselves, and it is difficult to obtain by others, nor can it be obtained by AI models.

Austrian-British economist and philosopher Friedrich Hayek said:

It is in this respect that each person actually has a certain advantage over all others, because each person possesses unique information that may be extremely valuable, but it can only be used when decisions based on this information are made by each individual or through active cooperation with individuals.

This is to say, tacit knowledge cannot be transmitted to any computer in a statistical form because it is impossible to count. Apart from that, tacit knowledge changes at any time and place, and can only be effectively transmitted through a flexible price mechanism. It is like trying to catch the moon in the water for an AI model to collect such information. This is also the reason why AI can’t replace humans. It has nothing to do with computing power.

Big data AI technology has indeed enabled us to have a larger database and more powerful computing power. However, no matter how extensive the database is, it is still statistical data, which is far from enough to include the implicit knowledge that cannot be expressed in words. Particularly, the implicit knowledge is the most critical factor in dealing with the contingency in economic development.

Take me as an example: I have been in the habit of recording detailed schedules for years, such as what information I received, who I met, what decisions I made and why I made such decisions every day. However, if I input the contents of my diary into the computer, can the computer predict my future decisions? Certainly, impossible. Most of my decisions are based on my inspiration and experience. The “why” written in my diary cannot contain all the thoughts and ideas when I make a decision. By contrast, the implicit knowledge that is not written down is the most critical factor.

From a dynamic perspective, a large amount of knowledge is discovered and created by participants in the market economy in the process of economic activities and is the product of practice. Without autonomous economic activities, the information itself cannot exist. For example, suppose SpaceX had only waited for ChatGPT to issue instructions based on its prediction of future market profitability before starting research and development. In that case, most of SpaceX’s rocket launch technology might not have been developed and realized until now.

Obviously, it is unlikely to collect information that has not yet been created, regardless of whether this knowledge is communicable or incommunicable. Therefore, big data and AI technology can only collect and analyze existing information. It is impossible to foresee information that is yet to be created and explored.

Hayek emphasized the dynamic nature of economic decision-making. He stated that changes, and only changes, cause economic problems. If things remain unchanged or continually develop as people expect, then there will be no new problems that require decisions. If someone thinks that change, at least daily adjustments, has become less important, then he is tantamount to arguing that economic problems have become unimportant.

Mises believed that the information constantly generated by the market comes from the exercise of entrepreneurial talent, which is related to a specific time and environment and can only be perceived by individuals acting in this environment. In a static state, economic calculation can be ignored because in this state, the same thing in economic development will happen repeatedly.

Back in the era of Mises and Hayek, there were no computers at the time, let alone big data and AI. However, the conclusions of Mises and Hayek have proven that even with computers, big data and AI, humans cannot be replaced because computers cannot easily collect the tacit knowledge that is crucial to economic development.

More broadly, participants in the process of social and economic activities can use computers, big data and AI models to help people continuously create an unimaginable amount of implicit knowledge, making it impossible for computers themselves to understand the implicit knowledge in people’s minds fully.

Innovation of humans

According to Mises, Hayek and others, discovering, creating and transmitting information in a priced or nonpriced manner is a manifestation of the human innovative spirit.

In the mainstream economics framework, the goals and means of individual decision-making are given. The so-called decision-making process involves choosing the means that can optimize the goal among the given options. However, this is far from real innovative decision-making. 

On the contrary, in reality, innovative decision-making is not a decision made under the framework of given means and goals, but to find, identify and choose the goals and means themselves. The level of innovative spirit largely depends on the ability to perceive and identify goals, as well as obtain the necessary means. If the means and goals are given and are identical, all rational people will make the same choice under the same big data background. Nevertheless, what we observe in reality is that, even based on the same data and the same knowledge, different people will always make different choices.

This is because decision-making depends not only on data and knowledge, but also on imagination. The imagination, perception and judgment of each market participant on resource availability and market and technological prospects are the most critical.

Of course, big data and AI are useful to market participants because each market participant needs data support when making decisions. However, true decisions must go beyond data. Decisions based solely on big data are at best scientific decisions, but not innovative decisions. Innovative decisions must be able to imagine and see those facts that big data cannot directly reflect.

Just like after the Second Industrial Revolution, when electricity became available to all individuals and businesses, it was no longer the core competitiveness of any company. Therefore, truly innovative decisions must go beyond big data.

Let us take the film industry as an example. Imagine a film production company could access big data on all movies from the past few decades, including viewer numbers, audience age groups, geographic locations, viewing times, box office revenue, media reviews and even audience reactions (such as how often they laughed or how long they cried). Does that then mean that, even with such big data, AI can actually predict what the next blockbuster will be?

Every new film is essentially a new creation, and people simply cannot forecast which movies will resonate most with the public. As William Goldman, a veteran Hollywood screenwriter, noted: “Why did Universal Pictures, the largest film studio, refuse to produce ’Star Wars’? Because no one now, and no one in the future, knows which movie will be a big hit and which will flop at the box office.”

The same is true for the book market. Amazon undoubtedly controls the core data of the book market. It can surely use AI algorithms to determine the types of books recommended to specific customer groups based on the customers’ records of viewing and purchasing books. But even so, Amazon still cannot bet on which book will be a bestseller in the future, let alone tell each author what kind of book they should write in the future.

International containerization (a shipping method where cargo is loaded into standardized containers for transportation by various modes) is one of the most important innovations of the second half of the 20th century. It has made great contributions to international trade, economic globalization and especially the global distribution of supply chains. Before the 1950s, goods were transported in bulk, whether by sea, rail or truck. The same goods had to be loaded and unloaded repeatedly from the manufacturer to the retailer, which was time-consuming, labor-intensive, costly and prone to theft, making it very unreliable. At the dock, the goods are usually piled up like mountains, making people miserable.

As early as 1937, Malcolm Purcell McLean, a truck driver in North Carolina, had the idea of using containers for transportation. In 1955, he sold his shares in the family transportation business, took out a loan to buy seven old tankers and converted them into platform ships that could stack containers. Next, he reinforced the truck cargo box to convert it into a trailer capable of loading containers. He installed a steel frame above the deck of the tanker and then installed a plug that could quickly place containers. 

On April 26, 1956, a converted World War II tanker set off from the Port of Newark, New Jersey, to Boston, with 58 containers on board. McLean then made the same transformation to several other old ships and opened a container shipping route from New York to Texas.

Via his success, other shipping companies followed quickly. By the late 1960s, the era of container transportation had finally arrived. By the late 1990s, 60% of the total value of international trade goods was transported by containers. Compared with bulk transportation, the transportation time from producer shipment to buyer receipt is reduced by 65%, and the unit transportation cost is reduced by 58%. It can be said that without the revolution of container transportation, there would be no global division of labor in the industrial chain that emerged after the 1970s.

However, why was it McLean who invented container transportation, rather than the original shipping companies or anyone else? This obviously cannot be explained by data. As far as data is concerned, McLean was originally just a truck driver, and the data he had was not at the same level as that of traditional shipping companies.

The fundamental reason was inspiration. McLean had this inspiration one step ahead of others. He later recalled in an interview that one day in 1937, while anxiously waiting for the cargo to be loaded and unloaded, an idea suddenly struck him: Why not lift the cargo box without touching anything inside and put it directly on the ship?

In short, no database, no matter how large, can replace human thinking and judgment. Amazon’s big data cannot replace Jeff Bezos, and the world’s big data also cannot replace global entrepreneurs.

Risk and uncertainty

In essence, risk and uncertainty are two completely different concepts. Human decision-making cannot be based solely on data because business operations and innovations are mainly faced with uncertainty, rather than risk. Although “uncertainty is the most certain thing” has become a slogan that everyone knows in the business and financial fields, most economists and business managers refer to risk when they talk about uncertainty. American economist Frank Knight made a clear distinction between the two as early as 1921, but unfortunately, his theory has not changed the situation in which economics equates uncertainty with risk.

According to Knight, risk is quantifiable, but uncertainty is not. Risk has a probability distribution based on the law of large numbers, so it can be reduced or increased. However, uncertainty has neither a priori probability nor statistical probability, so it cannot be reduced or increased in advance. In other words, risk is exogenous (external), while uncertainty is endogenous (internal).

The uniqueness of uncertainty means that past data cannot provide information about the future. Parameter estimation, the central limit theorem, least squares estimation, linear causality and Bayes’ rule, among other statistical concepts, can only be applied to risk management but offer little assistance in addressing uncertainty.

Innovation is a unique ability of human beings. The most fundamental characteristic of innovation is that its process is full of a series of uncertainties, rather than risks in the usual sense. No one can predict in advance whether an innovation will succeed, nor can they calculate the probability of success.

In an uncertain world, the most valuable predictions cannot be based on past data alone. This is why the human thought process is vital: to cope with uncertainty.

Because our future is uncertain, people’s predictions cannot be based on statistical models or calculations, but only on their “mind model”, imagination, self-confidence, judgment and even courage. Any prediction that can be made through a statistical model cannot be called innovative work.

As it should be, big data can undoubtedly provide more information than sample data, and AI technology can also help to more accurately calculate the probability distribution of risk events, thereby helping data users reduce, disperse or transfer risks. However, there is still no big database or AI algorithm that can predict the probability of success of an innovation in the future.

AI will never be able to predict why plasma TVs and rear-projection TVs failed to compete with LCD TVs, nor can it tell us why DeepSeek suddenly emerged, the traditional taxi industry lost to Uber and Uber lost to Didi (a Chinese online ride-hailing company) in the Chinese market.

If our world only had risks (such as natural disasters caused by weather and earthquakes) and no uncertainties, predictions based on big data and AI technology might be possible. But in the real world, in addition to risks, there are countless uncertainties. Dealing with uncertainty requires human autonomy and unique thinking ability. If we mistakenly believe that our world is full of risks and try to deal with uncertainty by dealing with risks, the result will definitely be the stifling of human creativity.

The concepts govern people’s behavior

For a long time, the basic view of mainstream economics is that interests completely dominate human behavior, and rationality can almost explain everything. But as Scottish economist David Hume pointed out more than 200 years ago, human behavior is not only dominated by interests, but also by the concepts they hold. 

Moreover, people’s understanding of interests is often perceived through concepts as well. In other words, what one thinks is in their interests is related to the concepts they hold. Therefore, even if people only act according to their own interests, their behavior is still affected by concepts.

Because concepts influence people’s behavior, it is often difficult to make a clear distinction between cause and effect. The same phenomenon can have completely different consequences, depending on how people understand it.

 For example, the Great Depression of the 1930s led to the Social Democratic Party taking power in Sweden, while in Germany, the Nazis came to power. The East Asian financial crisis in 1997 led to the decentralization of government power in some countries, while the financial crisis in 2008 made the governments of some countries more centralized. In the face of financial crises, the government’s response policies depend on whether those in power believe in Keynesian economics or Hayek’s business cycle theory.

Since concepts govern human behavior, it is impossible to accurately predict the future from past data. No matter how powerful an AI model is, no one could have predicted the sudden collapse of the Soviet Union in 1991, because it is impossible to predict that human concepts will change so quickly. ChatGPT did not predict that Trump would retake office in 2024, and DeepSeek also did not predict that the Trump 2.0 administration would immediately issue such a shocking tariff policy.

Therefore, understanding the future is always a matter for humans. No AI model can accurately predict the future of humans simply by using big data. If AI wants to succeed in this regard, it must control the human mind, which is an impossible mission. However, humans can indeed be brainwashed sometimes.

From the perspective of evolution

The stability of species comes from heredity, while evolution comes from variation. Variation refers to the mistakes made by genes during replication. Once this mistake is positive and has higher adaptability, nature or humans will favor it, and new genes will replace old genes. With such variation, species evolve. If genes do not mutate, species will not evolve.

However, heredity may be predictable, but mutation is unpredictable. This is true for the evolution of species and the progress of human society. Likewise, innovation is the mutation of social genes. Without innovation, human society will not progress. Among others, thinkers and entrepreneurs are just the power of mutation in our society. Big data and AI may reveal what the future holds if we follow the rules. However, they have no way to predict what kind of innovation will appear in the future.

Innovation is the emergence of unique ideas and concepts and their implementation. If this new concept gains recognition from society, it will gradually dominate societal development. If it is not recognized and is killed at the outset, our concepts will not change, and society will not progress.

Tobacco originated in America and was introduced to China in the late 16th century. Yunnan Province, located in southern China, has a tobacco planting history of over 100 years. However, the reason why Yunnan Province nowadays has become China’s “Tobacco Kingdom” is not only due to climatic conditions, but also related to the evolution of tobacco varieties. 

At the beginning, Nanyang Brothers Tobacco Company introduced a high-quality American tobacco variety called “Golden Yuan” in Yunnan. In 1962, a tobacco farmer in Yunnan Province discovered a unique and different “Golden Yuan” with large and gorgeous flowers. Then, the farmer sent the plant to the Yunnan Academy of Agricultural Sciences for research. Experts found that it was a mutant “Golden Yuan”. As a result, “Big Golden Yuan” was cultivated and spread soon, leading to changes in the tobacco varieties in Yunnan Province.

The emergence of the “Big Gold Yuan” is obviously not something that can be deduced by the AI model. On the contrary, if humans strictly follow the conclusions given by the AI model, the “Big Gold Yuan” may not even survive.

Similarly, innovation cannot be calculated by AI. Valuable innovation must be discovered and screened by humans in the process of practice. Moreover, the process of discovery often happens by accident, not by conscious search. Complete reliance on big data and AI may ultimately stifle innovation and the source of human progress.

Extension of the “Lucas Critique”

There is a famous theory in macroeconomics — the “Lucas Critique”, which means that any economic model based on empirical data cannot be used for policy-making, because the implementation of the policy will change the model itself that derives the policy.

I would like to extend the Lucas Critique as follows: any market economy-based empirical laws cannot be used for policy making, because the implementation of such policies will change the basis of the behavior of the empirical subject. 

For instance, suppose that through the calculation of big data, the AI model tells us that the enterprises with an annual output of 10 million tons in the steel industry are the most efficient. Suppose the government stipulates that enterprises with a yearly production capacity of less than 10 million tons are not allowed to invest. In that case, enterprises with an annual production capacity of 10 million tons cannot be the most efficient, as this is no longer a result of competition. 

Similarly, most of the large number of entrepreneurs who emerged in China when the reform and opening up just started in the 1980s had never attended college. If a policy is formulated based on this big data to allow only those who have not attended college to start businesses, then many potential entrepreneurs will lose the opportunity to enter the market.

Big data and AI may indeed replace a large number of non-creative jobs, but they can never replace people’s innovative spirit, as long as human society does not enter a static state, namely, without change, innovation or progress. If ever humans still need an innovative spirit, we cannot simply rely on big data and AI.

Finally, two points needed to be pointed out:

First, this article does not completely oppose big data and artificial intelligence, but the idea of ​​relying entirely on them and firmly believing that they will replace humans in the future.

Second, this article does not mention the unique incentive mechanism problem in the human economic system. The above statement only proves that even without considering incentive, it is impossible for big data and AI to completely replace human labor and thought. If the incentive factor is taken into account, it is even more unrealistic for big data and AI to replace humans.

[Kaitlyn Diana edited this piece.]

The views expressed in this article are the author’s own and do not necessarily reflect Fair Observer’s editorial policy.