AI has made rapid progress in recent years, leading some to call for a research pause and even government regulation. However, AI systems are not yet able to help decision-makers understand causation or uncertainty, and the need for large datasets to power them can create risks around security, privacy, legality and ethics. While AI can be useful for making predictions in low-stakes situations, when the stakes are higher, we need to understand what is driving an outcome and how confident we are in our knowledge of the world.
