AI conversations become easier to follow when a few terms are clear. The distinctions below help you ask what a tool actually does and what evidence you need before relying on it.
What is an AI model?
A model is a mathematical system used to produce predictions or other outputs. In machine learning, its parameters are learned from data. A model is one part of an AI application; the application may also include search, tools, and access controls.
Algorithms and training
An algorithm is a procedure for carrying out a task. Training is the process of adjusting a model using data. The training method, data, and evaluation all influence the resulting system.
Supervised learning
Supervised learning uses examples paired with target answers or labels. A system might learn to classify documents from previously labeled examples. Performance on new material still needs testing.
Large language models and tokens
A large language model works with language represented as tokens, which can be words or parts of words. Generating fluent language is different from establishing that a statement is true.
Context windows and fine-tuning
The context window limits the information available in a request. Fine-tuning is additional training of an existing model. Supplying a document in a prompt is different from training the model on that document.
