# How LLMS Work
* To understand this we need to understand
* Transformer
* Attention
* Temperature: Temparature of llm models make word prediction creative
* Attention
“`
The cat jumped out of building, after it has seen a bird.“`
* Word Embeddings: These are all the words, these words are represented as vector in high dimensional space
* Prompt:
* You are an helpful assistant
* I want to learn about operating systems
* Foundational Models: They have been trained on large volumes of data from internet and various other sources
* GPT => Generative Pretrained Transformer
* BERT
* Foundational Models are very huge, but we have lots of smaller models designed for a specific purpose which can be run locally (on-premises or behind enterprise firewalls)
* What we are trying to build
* RAG
* AGENT
