'AI learns from our masters': Posco's vision for the AI-powered steel mill - Heraldcorp.com

TEXT START: At Posco, whether or not someone holds the title of master craftsman, there are people throughout the company who have accumulated remarkable know-how and experienced all kinds of unexpected situations.

The Dissection

This is a corporate transition narrative disguised as a human-centered AX interview. Its central operation is to convert veteran judgment—scarce, embodied labor—into machine-readable training data, then connect separate models into an integrated controller for the mill.

The article presents extraction as preservation: AI supposedly receives human intelligence so the company can do its “core work” better. The actual direction is harsher. Human expertise is being compiled into capital that can make decisions repeatedly, cheaply, and across shifts.

The text quietly discloses the end state. AI already controls meaningful galvanizing decisions, vision systems combined with worker know-how automate steelmaking, robots perform dangerous inspections, and an energy agent coordinates plant-wide tradeoffs. Human approval is a current validation bottleneck, not proof that human judgment remains structurally necessary.

The Core Fallacy

It confuses the survival of knowledge with the survival of the people who possess it. Once expertise is codified, its scarcity falls. The master becomes a data source, evaluator, and temporary exception handler; the trained system becomes the repeatable production asset.

It also treats firm-level productivity as if it were worker-level security. Under P1, superior cognitive automation expands. Under P2, no stable human-only economic domain can be preserved when an integrated mill can coordinate processes more cheaply and consistently. Under P3, the result is a smaller number of people supervising a larger automated asset. The most capable workers may accelerate their own displacement by making the model better.

Hidden Assumptions

• Human sign-off will remain permanently necessary rather than serving as a lag defense during verification, liability, and institutional adaptation.
• Tacit knowledge cannot be generalized from sensor data, operating records, outcomes, and repeated recalibration.
• Better AI performance will create enough new human work to offset the labor it removes.
• Productivity gains will benefit workers rather than primarily accruing to owners and controllers of the mill.
• “People as the core competitive advantage” means people retain bargaining power, although the mechanism turns their expertise into transferable corporate capital.
• Physical AI is merely assistance, rather than the convergence of perception, reasoning, control, and machinery into a labor-substituting production system.
• Keeping humans in the loop is equivalent to keeping them economically indispensable. It is not.

Social Function

Primary classification: transition management wrapped in ideological anesthetic, with a substantial partial truth.

The partial truth is real: domain expertise, clean data, verification, safety, and cross-process integration are necessary to deploy industrial AI. The anesthetic is the framing. Workers are told their accumulated judgment is being honored because it is being harvested. Executives can present displacement as augmentation while the organization measures automation through productivity and profit and avoids the distribution question: who remains necessary, who owns the resulting intelligence, and what happens to those made redundant?

The “masters” are legitimizing figures. Their presence softens the fact that their knowledge is being detached from their bodies and embedded in a system designed to operate with fewer bodies. This is not sentimental preservation. It is transition management for converting labor into infrastructure.

The Verdict

Operationally credible; socially evasive. Posco is not showing that humans remain central to the steel mill. It is showing how human competence is converted into an autonomous, integrated production asset. The energy brain is the decisive clue: once AI moves from automating tasks to coordinating the whole facility, augmentation becomes system-level substitution.

The steel mill may become more productive, safer, and more profitable. That does not preserve the postwar employment bargain. The masters are the teachers and temporary scaffolding. The trained machine is the successor. Human approval, recalibration, and intervention are lag defenses—hospice care for a labor role already being engineered out.