What is a Large Language Model?
An LLM is not a search engine with a human-like mind. It is a model trained to predict likely next tokens from context. That simple idea explains much of its capability and many of its failures.
The next-token idea
A language model receives a sequence of tokens and estimates a probability distribution for what token should come next. A token may be a whole word, part of a word, punctuation, or another unit used by the tokenizer.
During generation, the system selects a token, appends it to the context, and repeats the process. A useful answer emerges from many small predictions made in sequence.
Why this can look intelligent
Training exposes a model to large collections of text and other data. It learns statistical structure: grammar, facts that frequently appear together, styles, patterns of explanation, and relationships between concepts.
Those learned patterns can support summarisation, translation, coding, classification and reasoning-like behaviour. But fluent output is not proof that a claim is true or that the model has checked a source.
What the model does not guarantee
- It does not automatically know whether a statement is true.
- It may produce plausible details when information is missing.
- It can reflect errors, gaps and biases in training data.
- Its answer depends on the prompt, context, model version and generation settings.
Sign in to access the lab and homework
The lesson explanation is free to read. Create a free learner account to use the practical exercise, submit homework and track your progress.
CREATE FREE ACCOUNT →