Summary

Large Language Models (LLMs) are perhaps the largest step forward in natural language processing in recent years. LLMs combine almost inconceivable amounts of textual data, the latest developments in Transformer deep neural networks, and self-supervised machine learning approaches to learn billions of parameters. The end result is a class of systems that is unprecedented in its capability to generate text and interact with users in a way that feels natural.

Over the past 2 to 3 years, a large number of LLMs have been trained by different industry and academic teams and optimized for specific tasks and architectures, such as unidirectional and bidirectional transformers. In this live course, you will learn about the fundamental concepts underlying LLMs and the pros and cons of each approach, and analyze specific models in some detail. Our goal is to provide you with the conceptual framework necessary to understand the latest developments in this area and to quickly evaluate which model might be the right solution for your own specific problem. Practical examples using ChatGPT and the OpenAI API will be used to give attendees a hands-on understanding of the power and limitations of this class of systems.


Program

  • Language Models

    • Basic Principles

    • Statistical Models

    • Encoder-Decoder

    • Transformer Models

  • Large Language Models

    • ChatGPT Architecture

    • BERT Architecture

    • LLAMA Architecture

    • Model Comparison

  • Embeddings

    • Understanding Embeddings

    • Question Answering

    • Recommendations

    • Long Texts

  • Applications Outside NLP

    • CODEX Model

    • DALL-E

    • BloombergGPT

    • BlockGPT


Resources