The Earth system is unique and highly complex, presenting a daunting challenge to researchers that seek to model and understand it. Noting that existing approaches seem unable to arrive at reliable predictions for the implications of CO2 emissions, in this video, MARKUS REICHSTEIN proposes that new methodologies incorporating machine learning and artificial intelligence be brought to bear on the problem. * Identifying notable parallels between conceptual challenges in the Earth system and successful applications of machine learning, Reichstein is careful to foreground problematic aspects of AI, arguing that the best methodological approach may well be hybrid, involving more traditional modeling alongside the data centered approach. This LT Publication is divided into the following chapters: 0:00 Question 2:05 Method 3:40 Findings 5:28 Relevance 6:17 Outlook