Conditional probability measures the likelihood of an event happening based on the knowledge that another event has occurred. It's calculated by dividing the probability of both events occurring by the probability of the given event. This concept is essential in AI for updating predictions and making decisions based on new information, especially in models like Bayesian networks.