Climate: The Final Deadline
AI is our best tool and one of our largest new emitters.
AI can model climate systems with unprecedented resolution, optimise every grid, and accelerate materials science by orders of magnitude. It also consumes staggering amounts of electricity. The next decade decides which of those two facts dominates.
The tools we now have
Better weather forecasting, precision agriculture, discovery of new battery chemistries, and per-building energy optimisation are already deployed at scale. None of them would exist without recent AI.
The load we now have
Training and serving frontier models is one of the fastest-growing sources of demand on power grids in North America and Europe. Without honest carbon accounting, the industry can undo its own contribution.
Adaptation vs mitigation
We have already locked in significant warming. AI's largest climate contribution may be in adaptation: helping vulnerable regions plan, evacuate, insure, and rebuild.
Questions worth arguing about
- ◆Should model training carry a mandatory carbon disclosure?
- ◆Is geoengineering acceptable if AI can model the consequences?
- ◆Who pays for climate adaptation in the places least responsible?