The Machines Modeling the Water Crisis Don't Understand a Drop of It

The most-cited advisor in Geneva this week never enters the hall and does not care whether it is right. The Concordat's allocation debate is being steered, clause by clause, by a bank of near-sentient forecasting systems that can project reservoir drawdown and cross-border demand with a fidelity no human team could match — and that understand, in any meaningful sense, nothing at all. "It models thirst without ever having been thirsty," said Dr. Wei-Lin Tan of the Singapore Institute, invited to remind delegates what they are leaning on. The systems produce fluent, confident scenarios; they mourn no shortage and fear no queue; their apparent judgment is, as the field's consensus still holds, extraordinarily complex programming and not one gram of experience. Tan's warning to the ministers was pointed: trust the arithmetic, distrust the fluency. "The map is superb," he said. "It has simply never been thirsty, and it never will be. Do not mistake its calm for wisdom." The delegates nodded, took the projections, and went on deciding — as only the thirsty can — what the numbers were allowed to mean.