Introduction to AI Computing: Generative Modeling

KAIST AIC.20000: Introduction to AI Computing (Fall 2026)

Lecture #1: Generative Modeling

Instructor: Minhyuk Sung (mhsung [at] kaist.ac.kr)

TA: Yunjae Jeong (frogjj [at] kaist.ac.kr)

Welcome to Introduction to AI Computing! 👋

In this lecture, you will explore the fundamentals of generative modeling through two activities:

1. In-class hands-on session: Implement a small diffusion model from scratch on MNIST.
2. Mini-project: Create creative and unique visuals using Stable Diffusion and showcase your results in a class-wide contest.

Note: Both activities are designed to be completed on Google Colab.

Lecture Schedule

Week	Date	Activity	Materials
1	Sep 3	In-class hands-on session	hands-on/
2	Sep 9	Hands-on session submission due (23:59 KST)	hands-on/
	Sep 10	Visual Content Contest	project/
	Sep 10	Mini-project submission due (23:59 KST)	project/

Repository Structure

AIC200-Project-Diffusion/
├── hands-on/                # In-class hands-on session
│   ├── diffusion.ipynb      # Implement and train a small diffusion model on MNIST
│   └── README.md
└── project/                 # Mini-project and visual content contest
    ├── dreambooth.ipynb     # Project example: Fine-tune Stable Diffusion v1.5 with DreamBooth
    └── README.md

Before starting each activity, please carefully read the README.md in the corresponding directory for detailed instructions and guidelines.

• Hands-on session: Implement and train a small diffusion model on MNIST to learn the fundamentals of diffusion models.
• Mini-project: Create creative and unique visuals using Stable Diffusion and showcase your results in a class-wide contest.

Grading

This lecture accounts for 20 points (20% of the total course grade).

Component	Points
In-class hands-on session	5pt
Mini-project	15pt
└ Write-up	5pt
└ Visual Content Contest	10pt
Total	20pt

Bonus points: Top-ranked finalists in the Visual Content Contest may receive up to +2 bonus points. However, the total score for this lecture cannot exceed 20 points.

Details for each item, including the bonus-point policy, are described in the README.md of the corresponding directory.