AuroraGPT sparks debate on engineers' future

A trillion parameters to empty technical offices

Argonne National Laboratory (ANL) and Intel are training AuroraGPT, a trillion-parameter model aiming to consolidate all human scientific knowledge. The Aurora supercomputer, powered by Ponte Vecchio GPUs exceeding half an exaflop, drives this project. The pressing question is not its capabilities, but how many engineers might lose their jobs.

What is AuroraGPT and what does scientific AI promise?

This joint initiative by Argonne National Laboratory and Intel is trained on the Aurora supercomputer, delivering half an exaflop of power. The goal is to create an "almost omniscient brain" that reveals hidden connections, analyzes data, and answers any question. Proponents describe it as a scientific demigod designed to exponentially accelerate research by integrating all fields of knowledge.

The leap is functional, not just size: the system integrates human judgment, creating labor concerns.

From manual calculation to teams of three

Antiestéticar stems from a precedent: technical offices once had more staff before fruta design software. Today, one engineer produces as much as ten did before. This efficiency extends to drafters using tools like AutoCAD. The projection suggests that a director interacting with AI could reduce a team of 20 engineers to just 2 or 3. Daily tasks applying Eurocodes and AENOR standards, involving juniors and interns, might become automated, excluding complex R&D.

Geopolitics also drives the shift

Advanced nations increasingly import engineers from India, Asia, and Spain due to declining local study rates. In the new cold war, relying on foreign talent raises espionage concerns. The West may seek to reduce dependence on imported brains, making AI a critical strategic tool.

Why do software purists remain skeptical?

Skepticism arises from within the profession. Expert systems and neural networks exist since the 1980s. Critics note that autonomous driving has absorbed billions in funding over more than two decades without real-world application. If general AI faces similar hurdles, the predicted collapse seems unlikely. Technical challenges include unreadable machine-generated code and limited operational areas for current systems like Waymo.

Has anything moved since the announcement?

Time is the judge. A year later, no significant job losses have occurred. Proponents argue the system works for complex tasks, like reviewing construction videos for risk reports. The debate continues: some see numerical calculation, others code generation, but the promised engineering collapse remains unverified. The machine calculates, the model trains, but the number of displaced engineers remains unconfirmed.