All Large Language Models (LLMs) You Should Know in 2023 | by Terence Shin | Jul, 2023

Intuitive explanations of the most popular LLMs

In my last article, we dived into the world of machine learning models, understanding their working principles and how they fit into various practical applications.

Today, we’ll venture into something that has quite literally taken over the entire tech space, large language models. Specifically, we’re going to go through several of the most influential language models in use as of 2023.

With that said, let’s dive into it?

Before we dive in, large language models can be generally classified into three categories based on their architecture:

1. Transformer-based models
2. RNN-based models
3. Other innovative architectures

These models leverage the power of attention mechanisms to process language data. Popular transformer-based models include GPT-4, BERT, RoBERTa, and T5

GPT-4

GPT-4 uses the transformer architecture with a particular emphasis on the self-attention mechanism to capture the contextual relationship between words in a sentence irrespective of their positions. Its “masked” training methodology allows the model to generate highly coherent and contextually relevant text.

• Pro: Highly skilled at generating coherent and contextually relevant text.
• Con: As a generative model, it may create plausible-sounding but factually incorrect or misleading information.
• Useful for: Text generation tasks, conversation agents, content creation.

BERT

BERT uses bidirectional transformers, meaning it processes input data from both left-to-right and right-to-left. This bidirectional context gives BERT a deeper understanding of the meaning of each word in a sentence and how they relate to each other, greatly enhancing its performance on tasks like question answering and sentiment analysis.