I‘ve been working in AI field for some time and recently, I’ve begun to view Large Language Models more like databases.
In this article I will explain why (as simple as possible).
What is database?
A database is essentially a system for storing data using an index (such as 1, 2, 3) paired with corresponding values (like A, B, C). By referencing the index, you can quickly retrieve the specific value you’re looking for. Simple as that.
What is Large Language Model?
A Large Language Model (LLM) is an advanced AI tool that, through training on vast amounts of data, can understand patterns in language and generate text-based responses. It is based on decoder-only architecture — how it looks then?
This diagram illustrates a simplified architecture of a transformer:
It consists of an input layer, multiple hidden neural network layers, and an output layer.
The input can be your query (prompt), which passes through the neural network neurons (determined by weights) and ultimately produces an output in the form of a word.
