The Robot Era: Skilled Workers' Dilemma

It is pointed out that there are limitations to the current mainstream large language models (LLMs) for evolving into Artificial General Intelligence (AGI). This is because LLMs learn knowledge based on language data, but lack the ability to understand and predict the laws of the physical world. To overcome these limitations, the importance of 'world models' is being highlighted.

A world model refers to the ability to understand the external world by combining physical laws and various sensory information, and to predict and infer situations without data. This is considered an essential element for AI to interact with and solve problems in real environments.

LLMs can be compared to having learned an entire library, acquiring symbolic knowledge through vast textual data. World models, on the other hand, learn the principles of the physical world by directly interacting with it. This signifies the advent of an era where the real world is increasingly simulated and digitized like a game.

This shift suggests that AI will gain the ability to react in real-time and make decisions in physical environments, not just process information. It indicates that AI's influence could extend to areas involving human physical labor.

To advance physical AI, securing motion data from skilled workers is crucial. This is a global phenomenon, not limited to Korea, with major countries like the U.S., China, and India fiercely competing to collect this motion data.

Related institutions, such as industrial ministries in various countries, are strongly pushing to secure skilled workers' motion data as an essential step for implementing physical AI. This is because it is recognized as a key factor for future industrial competitiveness.

Tacit knowledge, in philosophy, refers to a type of knowledge that cannot be explained in words or writing. For example, a grandmother's cooking touch falls into the realm of tacit knowledge that cannot be fully passed on through recipes alone. This is a difficult-to-explain type of knowledge, with limitations in textbook-like manualization.

In industrial settings, the skills acquired by highly skilled workers through physical experience also possess these characteristics of tacit knowledge. They cannot be perfectly conveyed through words or writing and are internalized through long experience and repetition.

In the past, the outcomes of knowledge work that were considered special can now be easily replicated and implemented by AI and computers. Conversely, physical labor, such as cleaning or dishwashing, which was valued less, faces significant difficulties in being implemented by physical AI. This shows a reversal in the valuation of labor.

In the AI era, the value of skilled physical labor, especially techniques belonging to the realm of tacit knowledge, is relatively increasing. As AI rapidly replaces knowledge work, the importance of complex task performance capabilities in the physical world grows.

A dilemma arises in the process of digitizing the movements of highly skilled workers. For instance, high-level movements like welding on curved surfaces in shipbuilding fall into the realm of tacit knowledge for skilled workers, making them very difficult to explain in words. However, to implement physical AI, these highly skilled worker movements must be turned into data to teach robots.

This is akin to a skilled worker showing how to ride a bicycle by doing it themselves, rather than explaining it, essentially transferring their skills to a robot. Workers are concerned that if their skills are digitized and applied to robots, their jobs could ultimately be threatened.

Furthermore, comprehensive data is gathered by video-recording any deformation in materials.

Major countries worldwide are staking their livelihoods on securing this motion data. China, in particular, is actively moving, for example, by creating dedicated studios to gather workers and intensively collect motion data.

Collecting motion data from skilled workers is not a simple task that ends with a single scan. It is difficult to define what constitutes a standard movement, and movements vary from person to person and situation to situation. Even a single action like slicing an onion presents an infinite number of complex possibilities.

Therefore, skilled workers must provide refined movements to maximize data efficiency. This means the active cooperation of skilled workers is essential to obtain high-quality data necessary for AI learning.

Skilled workers are reluctant to provide their motion data to AI. This is because it is akin to handing over their lifetime-accumulated skills and knowledge to an external entity in an instant. They face the risk of losing their jobs once the data is provided, making cooperation difficult.

Director Hong Ki-bin likens this to Esau selling his birthright to Jacob for a bowl of stew in the Bible, explaining that it is a matter of life and death for workers. A situation similar to the labor-management conflicts over process monitoring in 20th-century industrial sites could recur.

Director Hong Ki-bin proposes social consensus and the introduction of a new compensation system, 'data commons,' to resolve this dilemma. Instead of a short-term compensation method where motion data is sold once and then finished, he suggests managing workers' motion data like a commons (shared asset), guaranteeing workers a continuous stream of income whenever capital utilizes it.

This implies the need for a new social contract between labor, businesses, and the state. It advocates for establishing a sustainable system where everyone can share the benefits of social change brought about by technological advancement.

For companies like Hyundai Motor, for example, how to persuade workers to transfer their work methods to robots is a key factor for successful robot adoption. However, labor unions perceive robot adoption as a threat to their survival and cannot readily allow it.

This stalemate between labor and management can delay AI transformation and intensify social conflict. It could result in not fully leveraging the opportunities for technological advancement.

China tends to move rapidly with technology adoption without social obstacles. In this situation, if Korea fails to achieve social consensus on robot technology adoption and makes no progress, there is concern that its manufacturing base could be encroached upon by China.

Korea should learn from the example of Europe losing its electric vehicle market to China. It is warned that if Korea also fails to achieve swift social consensus and progress on robot technology adoption, it could lose competitiveness and fall behind in future markets.

There has been insufficient discussion on the large-scale paradigm shift that IT technology has brought to entire industries. IT has gone beyond merely a specific technological field to bring about revolutionary changes in the very way industries and societies are organized.

This causes a clash between traditional ways of thinking and IT-based solution thinking, demanding a fundamental shift in how we understand and solve problems in the world.

The 20th-century mindset tended to circumvent or overcome physical and social limitations. In contrast, the IT mindset has the characteristic of continuously creating solutions within the IT system when a problem arises. It believes that if a problem occurs, another IT solution can be created, considering there are no limits.

This IT mindset brings about a fundamental difference in problem-solving approaches, which also influences the speed and direction of technological development.

This is why electric vehicles, too, should be interpreted as an extension of IT devices, not merely a change in propulsion.

Only by understanding this holistic change can we accurately predict and respond to the future trends of industry and society.

While a military-like push might be beneficial for short-term results, long-term agreement with skilled workers is essential.

Such an agreement model can protect workers' rights while leading to results that benefit everyone through technological advancement. In other words, we must seek ways for humans and AI to coexist and create greater synergy.

For democracy to have vitality and resilience, it must not be rigid in its ideology. We need to approach practically and concretely where democracy can be efficient and where social consensus can be built. It is important to consider how social consensus and efficiency can be interconnected.

Thinking flexibly to solve real-world problems without being confined by ideology is the path to sustainable democratic development.

The superiority of democratic methods was proven during the overcoming of the Great Depression in the 1930s. At that time, the New Deal policies showed that they could be as effective as, or even more effective than, other systems like communism or fascism in overcoming a crisis.

This can be seen as a historical example proving that democracy can have powerful practicality in solving real problems, beyond just being an ideology.