Key AI and LLM Concepts Explained: Weights, Tokens, and Output Controls

Large Language Models (LLMs) are mathematical models built from billions of learned parameters called weights, trained on vast amounts of text and image data using Transformer architecture. Due to their size, running LLMs locally requires sufficient storage and computing power, with tools like Ollama and LM Studio enabling local deployment. LLMs function as next-word predictors, assigning probability scores to possible outputs, with parameters like Temperature, Top K, and Top P allowing users to control the randomness and diversity of generated responses. Text input is broken into smaller units called tokens through a process known as tokenization, with each token assigned a numerical value and converted into an embedding. A model's context window defines the maximum number of tokens it can process at one time, effectively acting as its short-term memory.