Modern AI systems need training data--great huge piles of it--to make their models work. Heaps of information. As much as the AI makers can get their hands on, and as fast as possible. These massive amounts of raw inputs ensure that an AI system can answer a question about which actor was in such-and-such a TV show, but also play a vital role in how AI makers ensure their generative AI models are actually able to churn out useful information in coherent human-like sentences when users put questions to them.