What are parameters in Large Language Models, and how do they help models learn language patterns? How are parameters updated during training, and why do different models have different parameter counts? What relationship exists between model size, performance, and computing requirements?