Interpretability: Understanding how AI models think [video]