Por dentro das LLMs como ChatGPT - O Curso Completo

Learn about the AI ​​Agent Engineer Training: https://asimov.academy/ia-formacao/?u... This material was previously only available to subscribers of the AI ​​Applications with Python track from Asimov, but we've decided to make it completely free for you here on YouTube. In this course, we'll explore LLMs, the models behind the artificial intelligence boom we've experienced in the last two years. I'll explain how these models are created, the three training stages needed to transform raw data from across the internet and "compress" it within the parameters of a neural network, turning it into an assistant capable of answering questions on any subject, performing daily tasks… and much more. TIMESTAMPS: 00:00 Introduction 02:39 Lesson 1 - Transformers and LLMs 05:32 Lesson 2 - First stage - Processing the entire internet 10:50 Lesson 3 - Second stage - Tokenization 17:40 Lesson 4 - Third stage - Performing pretraining 28:44 Lesson 5 - The properties of a BASE model 38:33 Lesson 6 - Fourth stage - Supervised Fine Tuning 47:28 Lesson 7 - Dealing with Hallucinations and using Tools 54:36 Lesson 8 - Models need tokens to think 59:42 Lesson 9 - Fifth stage - Reinforcement Learning 01:06:07 Lesson 10 - DeepSeek, AILeaderboard and closure