Plenary IV: From Statistical To Causal Ai

In machine learning, we use data to automatically find dependences in the world, with the goal of predicting future observations. Most machine learning methods build on statistics, but one can also try to go beyond this, assaying causal structures underlying statistical dependences. Can causal knowledge help us produce intelligent behavior? We argue that causality can play a central role in addressing some of the hard open problems of machine learning, due to the fact that causal models are more robust to changes that occur in real world datasets. We discuss implications of causal models for machine learning tasks, focusing on an assumption of ‘independent mechanisms’, and discuss an application in the field of exoplanet discovery.