Introduction to Generative AI
An introductory generative AI course taught by Professor Hung-yi Lee of National Taiwan University. It explains large model training and alignment techniques from a principles perspective, has no prerequisites, and is suitable for learners with no prior background.
Timeline (last 90 days)
At the opening of the course, Hung-yi Lee clarified that the course is not about teaching how to use ChatGPT, but about the full picture of generative AI, including where models come from, how they are trained, and future possibilities.
According to an integrated version by a Bilibili creator, the course is divided into three major modules: the fundamentals section (Python, environment setup, prompt engineering, Transformer, pretraining, SFT, RLHF, etc.), the advanced section (RAG, Agent, LangChain, fine-tuning and private deployment), and the hands-on section (enterprise-level projects such as medical Q&A, e-commerce customer service, digital humans, and educational teaching assistants).
Assignment difficulty is marked with triangles: one triangle is basically a giveaway question, two triangles require time to get full marks, and three triangles involve training models and require at least several hours.
All Facts (3)
At the opening of the course, Hung-yi Lee clarified that the course is not about teaching how to use ChatGPT, but about the full picture of generative AI, including where models come from, how they are trained, and future possibilities.
50%UnverifiedAccording to an integrated version by a Bilibili creator, the course is divided into three major modules: the fundamentals section (Python, environment setup, prompt engineering, Transformer, pretraining, SFT, RLHF, etc.), the advanced section (RAG, Agent, LangChain, fine-tuning and private deployment), and the hands-on section (enterprise-level projects such as medical Q&A, e-commerce customer service, digital humans, and educational teaching assistants).
50%UnverifiedAssignment difficulty is marked with triangles: one triangle is basically a giveaway question, two triangles require time to get full marks, and three triangles involve training models and require at least several hours.
50%