Educating future engineers about LLMs: A scalable workshop

URL SCAN: Educating future engineers about LLMs: A scalable workshop
FIRST LINE: # Computer Science > Human-Computer Interaction

The Dissection

This paper is not merely teaching engineering students to use LLMs. It is institutionalizing a new subordinate role: humans as prompt designers, verifiers, and supervisors for machine-produced plans.

The workshop converts displacement into curriculum. Students are trained to adapt their behavior as models improve—from issuing step-by-step instructions to providing higher-level guidance. That strategic shift is real, but it signals rising machine abstraction, not durable human indispensability. The Spot demonstration adds physical theater: the model produces the plan, the robot executes it, and the human remains in the loop as a temporary trust and safety layer.

The Core Fallacy

The paper treats verification and collaboration as enduring engineering advantages. Under the Discontinuity Thesis, that is the central error.

Verification is itself cognitive work. Once models become sufficiently reliable, the system will automate much of the checking, comparison, planning, and exception handling that this workshop presents as the human contribution. The paper confuses participation in an AI workflow with economic necessity within that workflow.

Prompting at a higher level is not sovereignty. It is evidence that the human is moving upward in the interface while the machine absorbs more of the underlying labor. The students are being prepared to operate an AI-owned production system, not to own or control it.

Hidden Assumptions

• Human engineers will remain necessary as verifiers even after verification tools improve.
• Skills learned in grid navigation transfer cleanly to complex, open-ended engineering work.
• Automated scoring reliably measures competence rather than performance inside a narrow toy environment.
• Self-reported learning and engagement demonstrate durable labor-market value.
• A physical robot demonstration creates a lasting moat rather than merely exposing temporary hardware and institutional lag.
• Human-AI collaboration will remain a stable equilibrium instead of collapsing toward fewer supervisors overseeing larger machine systems.
• Teaching responsible use can preserve broad engineering employment after the underlying cognitive tasks become cheaper to automate.
• Education can scale human adaptation faster than AI can erase the tasks being adapted to.

Social Function

This is transition management, ideological anesthetic, prestige signaling, and partial truth.

The partial truth is that engineers currently need to understand model behavior, prompt design, output verification, and physical execution constraints. The anesthetic is presenting those transitional skills as a durable professional settlement. The workshop gives institutions a respectable narrative for the erosion of authorship: students are not being displaced; they are becoming “collaborators.”

Its real function is to train prospective Servitors for an AI-centered engineering stack. It makes the transition more orderly and more psychologically acceptable. It does not preserve the post-WWII employment circuit.

The Verdict

The paper documents the early adaptation of engineering education to cognitive automation; it does not refute the Discontinuity Thesis. It scales the pedagogy of human supervision while leaving ownership, model capability, and productive control untouched.

The workshop may produce competent operators during the lag phase. Its deeper promise is terminal. As AI absorbs planning and verification, the human role it trains will be compressed from engineer, to supervisor, to exception handler, and finally to a small residual layer of accountability around systems controlled by Sovereigns.