What is an LLM?

Large Language Models

Large language models are trained on large amounts of text data, typically using a technique called “unsupervised learning.” This means that there is no explicit information given on what to learn, but rather left to discover patterns in the data by itself.

• Popular LLM architectures include
  • GPT (Generative Pretrained Transformer)
  • BERT (Bidirectional Encoder Representations from Transformers)
  • T5 (Text-to-Text Transfer Transformer)
• Characteristics
  • Large-scale.
  • Deep learning architecture, they are trained based on deep learning techniques like artificial neural networks, which allow them to learn complex language.
  • Pre-training and re-trained on specific tasks, a trained LLM is fine-tuned to perform specific tasks.
  • Contextual, they don’t just use the word, but also the context to better understand its meaning. This helps in creating a more human-like language.
  • Generative, they can generate new text.
  • High-quality output, it makes it difficult to distinguish from human-written text.

LLM Optimization

Optimizing Large Language Models (LLMs)

General LLM Concepts

General Intelligence

• AI is any system that exhibits behavior that can be interpreted as human intelligence.