An interpreter at negotiations must translate quickly without losing the meaning of what is said. Similarly, the CAMulator neural network takes a slow climate model and translates it into a fast language, preserving the essentials: air mass, moisture, and energy. Previously, such "translators" either simplified the picture or forgot these laws, and their forecasts broke down. CAMulator is the first to maintain both full resolution and strict physical rules.

But the translator has a weakness: when asked to project a future with a warmer ocean, it makes winters too cold in polar regions. This is because its training did not include interactive sea ice, so it does not know how ice responds to heat. However, it captures El Niño and other important climate phenomena well. If coupled with ocean and ice models, it would become even more accurate. Thus, the fast translator helps scientists peer into the future of climate.