A predictive model is a statistical or machine learning model trained to estimate the likelihood of specific future outcomes based on patterns found in historical or input data. Rather than describing what has already happened, its purpose is to forecast what's likely to happen next, whether that's a customer's probability of churning, a patient's risk of a medical condition, or the next word in a sentence.
ChatGPT and other language models are, at their core, a specialized kind of predictive model: given the text so far, they predict the probability distribution over what word is most likely to come next, then select from that distribution to generate a response. The same underlying concept, learning from past data to estimate future probabilities, applies just as directly to more traditional predictive uses like sales forecasting or fraud detection.
