Large Language Model
A large language model (LLM) is an AI system trained on massive collections of text to predict the next token, which gives it a general ability to write, summarise, translate and reason without being programmed for each task.
No jargon, no math
In plain words, what is it?
A large language model is a next-word engine scaled up until prediction becomes a general skill. During training it reads enormous volumes of text and repeatedly adjusts its internal weights to better predict the next token (a word fragment) given the tokens before it. Those weights end up storing statistical patterns of language and the world: grammar, facts, the typical way a contract or an email unfolds. At run time nothing is programmed: you feed tokens in, the model scores every possible next token, picks one, appends it, and repeats. The result reads like understanding because predicting well over enough text forces the model to capture structure. Where it breaks: knowledge freezes at the training cutoff, there is no memory between separate conversations, and when patterns run ahead of facts you get confident guessing, the famous hallucination. Scale fixed fluency, not truth: verify the numbers it cites.
Terms that live nearby.
Transformer
The transformer is the neural network architecture, published by Google researchers in 2017, behind every modern language model: it processes all tokens of a text at once and uses attention to let each token weigh the relevance of every other token.
ReadAttention Mechanism
Attention is the mechanism inside a transformer that lets each token weigh how much every other token matters before being processed, so the model can track what 'it' refers to or which date belongs to which clause.
ReadToken
A token is the chunk of text a language model actually reads and writes: on average about three quarters of an English word, converted to a number from a fixed vocabulary before the model can process it.
Read§ 03
Want this running in your stack?
We design and ship AI systems with permissions, logs and human approval. It starts with a 30-minute scoping call.