Inside GPT - Large Language Models Demystified

Explore the world of generative pre-trained transformers (GPT) algorithms in this comprehensive conference talk from NDC Oslo 2024. Dive deep into the architecture and inner workings of GPT algorithms and ChatGPT, starting with fundamental concepts of natural language processing such as word embedding, vectorization, and tokenization. Learn how to apply these techniques to train a GPT2 model for generating song lyrics, with demonstrations of the internal processes of word sequence prediction. Examine the power, capabilities, and limitations of larger language models like ChatGPT and GPT4, and understand the impact of hyperparameters such as temperature and frequency penalty on generated output. Discover the concepts of prompt engineering and learn how to leverage Retrieval Augmented Generation (RAG) patterns to create ChatGPT experiences based on custom textual data. Gain valuable insights into harnessing the power of GPT algorithms for your own solutions in this demo-intensive session led by Alan Smith.