The corpus at the end of the world

A short story about machine learning, the climate crisis, and change

The day we discovered our machine learning models were wrong was the day the earth began to catch fire and didn’t stop.

For years, we’d been fighting brush fires every September and October. One year, they spread into the villages in the foothills. The next, the towns were under threat. The one after that, the fires spread into the cities. Finally, it was everywhere, and we didn’t have a hope of controlling it.

“It’s time to take a break,” my smart assistant said. It was designed to make good choices for me: a long, healthy life achieved through stress management and smart predictions.

Our office sat on a hill overlooking the bay. From our desks, we used to be able to watch sailboats dance in the sparkling water through immaculate floor-to-ceiling windows; now we just saw endless firefighting ships and planes. Our workstations were some of the most powerful in the world, armed with the best machine learning algorithms and processing software: giant artificial brains at our command. We developed intricate models that predicted weather, airflow, and vegetation changes with incredibly accuracy. But try as we might, we couldn’t make our holistic climate change models work perfectly.