10-202: Introduction to Modern AI (CMU)
Carnegie Mellon University has launched "10-202: Introduction to Modern AI," a comprehensive course designed to teach the underlying mechanics of large language models and machine learning, defining "modern AI" as the technology behind tools like ChatGPT. Crucially, a delayed, free online version of the course, complete with lecture videos and autograded assignments, makes this high-caliber education accessible to a global audience. This offering resonates strongly with the Hacker News community, eager for practical, deep dives into building and understanding contemporary AI systems from the ground up.
The Lowdown
Carnegie Mellon University has launched "10-202: Introduction to Modern AI," a new course designed to demystify the mechanics behind contemporary AI systems, particularly large language models (LLMs). Taught by Zico Kolter, this offering stands out by providing both an in-person CMU experience and a delayed, free online version, making cutting-edge AI education widely accessible. The curriculum aims to equip students with the skills to implement a basic AI chatbot from scratch.
- Course Focus: The curriculum specifically defines "modern AI" as machine learning methods and LLMs, concentrating on the technologies powering systems like ChatGPT, Gemini, and Claude.
- Practical Goal: The primary objective is for students to learn how to implement a basic AI chatbot, write code for open-source LLMs, and train these models from a corpus of data.
- Key Topics Covered: The syllabus includes a brief history of AI, supervised machine learning (linear models, neural networks), large language models (self-attention, transformers, tokenizers, efficient inference), and post-training concepts like fine-tuning, alignment, reasoning models, and AI safety.
- Hands-on Assignments: A significant portion of the course involves developing a minimal AI chatbot through a series of programming assignments, released as Colab and Marimo notebooks.
- Online Accessibility: A free, minimal online version of the course provides lecture videos and autograded assignments, becoming available two weeks after their release to CMU students.
- Prerequisites: Students are expected to have proficiency in Python programming (including object-oriented methods) and basic differential calculus, with an optional benefit from linear algebra and probability knowledge.
- AI Usage Policy: The course encourages using AI assistants for homework as a learning tool but advises against over-reliance, particularly noting that in-class evaluations (quizzes and exams) are strictly closed-book and closed-notes.
This course appears to be a comprehensive and practical dive into the core technologies powering today's AI, presented in an accessible format for a broad audience keen on understanding and building these complex systems.