Panel Session: Methodological Trade-offs in ML for climate change

Machine learning models continue to grow in parameter count. While in the 2000s they were in the hundreds, in the 2010s we moved to tens of millions, in the 2020s we started with billions, and now we are presumably reaching trillions of parameters. The exponential growth in model complexity is opening up important and broad points of discussion, related, among other things, to performance, carbon footprint, training and inference time, explainability, operational costs, and ethical considerations. Under the theme of our workshop, Roots and Routes: A Dialogue on Machine Learning Methods for Climate Impact, we will hear our experts' opinions on the trade-offs between the development and deployment of small and large models, and on the implications of their use for solving problems related to climate change, biodiversity, and population impacts.