Teaching an AI system about a subject. There are two ways AI systems learn. The "supervised" method enters thousands of examples (data parameters) with an explanation of what each one means. The "unsupervised" approach uses parameters without explanations and lets the AI system recognize patterns.
Training Takes a Lot of Computer Power
The training phase in today's large language models uses an enormous amount of computer power, and new datacenters are being built just to accommodate these requirements. See
AI datacenter,
supervised learning and
AI training vs. inference.