John Cronin: AI can answer, but can't judge

Cronin's column draws a straight line from the slide rule and FORTRAN of his college days through engineering software like StruCalc, Forte and ClearCalcs, to the automated systems on a modern ship's bridge. In every case, he says, the machine produced an answer, but a person stayed responsible for interpreting it. The turning point he keeps returning to is Coecles Harbor, where an intern's software found a wave-action solution in under an hour that turned out to be absurd — proof, for Cronin, that speed without understanding is dangerous.

He notes that maritime navigation already has safeguards built in: even with Automatic Radar Plotting Aid calculating the closest point of approach, a mariner seeking a Coast Guard license must still understand the underlying math. AI, he suggests, does not come with that kind of defined purpose and understood limits. He points out that even the companies developing these systems acknowledge significant questions about their capabilities and limitations, and that some are turning to philosophers for help.

The stakes, as Cronin frames them, may hinge less on how intelligent machines become and more on whether people stay serious about preserving the judgment to know when a machine's answer is — and is not — the answer. For engineers, mariners and anyone who acts on automated output, that is a test of human discipline as much as technical capability.