With the release of ChatGPT, there has been a paradigm shift in Generative AI and its applications. Despite this groundbreaking perception, the underlying technology has remained relatively consistent over the past five years. What sets ChatGPT apart is the refined integration of traditional methods with advanced data engineering to optimize human preferences, resulting in a sophisticated tool that has revolutionized the industry. This talk will dive into the technical details and the evolution of Large Language Models (LLMs) over the past 4-5 years. At the end, we will explore the potential for a knowledge explosion facilitated by generative models in the coming years. During the talk, I’ll emphasize the tips and tricks for training truly large LLMs, explain FLOPS per politics and show you the timeline of ChatGPT’s development and my experiences as a GPU-poor dancing around the sea of flops. I hope you come scaling your humor like actual Large LLMs.