Training vs Inference in LLMs

Every large language model has two very different lives: the phase where it learns, and the phase where it works. These are called training and inference, and while they're related, they're almost opposite in how they operate, what they cost, and what they're trying to...