Applied Causality in Computer Vision

The webinar was jointly organised by IAS and the Graduate Students’ Clubs of MAE, SCSE and EEE.

You probably have heard that correlation is not causality or seeing is illusion. There are multiple ways to seek causality, rather than just seeing and measuring the correlation between data. In this talk, Professor Hanwang Zhang talked about applied causality in computer vision related tasks, and different ways in which we can seek causality. He began the talk by explaining the need of causality. As humans, we often think in terms of cause and effect – if we understand why something happened, we can change our behavior to improve future outcomes. He gave multiple examples to emphasize this point – from Kepler’s laws of planetary motion to Newton’s law of attraction – why learning from incomplete data does not derive causality. He then defined basic terminologies, such as intelligence, AI, difference between causality and statistics, reversed causation, confounder, survivorship bias, etc., which helped the attendees to delve deeper into the upcoming concepts. The varied examples he discussed in his talk are aimed towards how we can think causally and how to capture causality in real life.

Purohit Shantanu | MAE Graduate Students’ Club