What Is an LLM?
Knowledge
LLM stands for "Large Language Model." It is software that has been trained on massive amounts of text and has thereby learned to generate human language. ChatGPT, Claude, and Gemini are all based on LLMs.
Think of your phone when you type a message: it suggests the next word. An LLM does essentially the same thing -- but a million times better. It has learned from billions of texts: books, websites, scientific articles, and programming code. This allows it to generate not just individual words but entire paragraphs, stories, or even code.
iImportant to Understand
An LLM is NOT a search engine. Google searches for existing web pages and shows them to you. An LLM generates new text based on patterns it has learned. It essentially "invents" every answer anew.
Understanding
The training of an LLM works like this: the model reads billions of texts and learns statistical patterns in the process. It learns which words are likely to follow each other. When you ask "The capital of France is...", the model has seen this pattern so often that it predicts "Paris" with high probability.
There is a fascinating debate: Is an LLM just a "stochastic parrot" that parrots patterns? Or does it, through its sheer size, actually develop emergent capabilities -- abilities that were not explicitly trained? The truth probably lies somewhere in between. LLMs do not understand language the way humans do, but they can solve astonishingly complex tasks.
What is the main difference between an LLM and a search engine?
*A Prediction Model, Not a Knowledge Model
Remember: an LLM only ever predicts the next most probable word. It does not "know" anything in the human sense -- it recognizes patterns. Once you understand this, you also understand why LLMs sometimes give convincing-sounding but incorrect answers.
Apply
The next time you use ChatGPT or Claude, observe: the answer is generated word by word. The LLM does not "know" at the beginning how the answer will end. It only ever predicts the next most probable word -- and builds up sentence by sentence this way.
Reflect
An LLM has read millions of cooking recipes during training. What can it do best afterwards?