MIT IAP Seminar 2023: Introduction to Text-to-Image Generation for K-12 Education

Instructors

Safinah Ali

Prerna Ravi

Katherine Moore

Faculty Sponsor

Hal Abelson

Cynthia Breazeal

Time and Dates

10-11:30 AM EST on Tuesdays & Thursdays [from Jan 10, 2023 to Jan 26, 2023(6 remote sessions)]

https://mit.zoom.us/j/95419500439

https://forms.gle/S2WzejmkUwUyqoMM9

Pre-requisites

Open to anyone affiliated with MIT, Harvard and Wellesley College. We strongly encourage members of MIT RAISE (Responsible AI for Social Empowerment and Education) to attend this seminar. While this seminar will discuss the current advancements and tools in text-to-image generation, its central focus will be on using these tools to create curriculum for K-12 AI Education.

Description

There is currently a proliferation of digital platforms to perform text-to-image generation. These platforms are breaking new ground in AI tools that let anyone, even beginners, easily create images with professional quality appearance. This seminar will explore these text-to-image generation platforms with an emphasis on opportunities in K-12 education. In this seminar, students will review recent technological development that has led to the rapid advancement of text-to-image generation, explore the components of text-to-image generation - including transformers, latent space, and diffusion - and discuss ethical and societal implications of this technology. Final projects will prototype learning activities for a target K-12 age group including learning goals, age appropriate tool introduction, and assessment. The main goal of the seminar will be to create curriculum using this image generation capability that’s aligned with our approach of constructionism and computational action. We would also like to stimulate collaborative development between the Personal Robotics Group(PRG) at the MIT Media Lab, App Inventor at MIT CSAIL, and the MIT Scheller Teacher Education Program (STEP) in working towards creating curricula and prototypes.

Format | Schedule

The seminar will comprise six sessions that will be held remotely via zoom. The schedule is summarized below:

Date Topic Description Session Video Slides Assignments
01/10/2023 Introduction: Exploring generative AI models and their relevance in K-12 AI Literacy - What are language inference and visual generative AI models and how do they work
- Examples of generative media
- Motivation for teaching this to K-12 students: New media and possibilities for creation and computational creativity and their ethical considerations		 Session 1 video Session 1 slide deck Assignment 1
01/12/2023 Experimenting with the latest generative AI platforms Survey and experiment with some existing platforms: DALL-E 2, Stable Diffusion, Dreamstudio, Midjourney, DrawThings, NightCafe and complete two activities using them: Creative AI Storytelling and Self-portraits Session 2 video Session 2 slide deck Assignment 2
01/17/2023 Reflection on using generative AI platforms and technology behind their models Reflect on your experience of creating images for the activities above and learn more about the technology used by the above tools: Transformers, Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), prompt engineering, diffusion Session 3 video Session 3 reflection slide deck
Session 3 technical slide deck		 Assignment 3
01/19/2023 K-12 generative AI literacy Hands-on workshop to investigate existing creative AI curriculum in K-12 education Session 4 video Session 4 slide deck Assignment 4
01/24/2023 Draft projects and proposals for creative AI literacy Discuss project proposal and provide peer feedback:
- Target age group
- Learning goals
- Tools used
- Learning activities
- Assessment		 N/A N/A -
01/26/2023 Presentations for prepared AI literacy proposals Class presentations Session 6 video Session 6 slide deck Class feedback form

1) Slack Invite Link for the Seminar: MIT_IAP23_Image_Generation: we will be using this space for content discussion during IAP and beyond.
2) (Recommended Read!!) Excellent article from MIT Technology Review summarizing the advances in this field: Generative AI is changing everything. But what’s left when the hype is gone?
3) How do Diffusion models work: https://towardsdatascience.com/diffusion-models-made-easy-8414298ce4da
4) Greg Rutkowski: This artist is dominating AI-generated art. And he’s not happy about it. https://www.technologyreview.com/2022/09/16/1059598/this-artist-is-dominating-ai-generated-art-and-hes-not-happy-about-it/
5) Diffusion models explained: https://www.youtube.com/watch?v=yTAMrHVG1ew
6) Illustrated Stable Diffusion (a bit more technical) https://jalammar.github.io/illustrated-stable-diffusion/
7) Diffusion Models: A Practical Guide: https://scale.com/guides/diffusion-models-guide
8) CLIP (connecting text and images): https://openai.com/blog/clip/
9) Paper: High Resolution Image Synthesis with Latent Diffusion Models (more technical but a great resource): https://arxiv.org/pdf/2112.10752.pdf
10) If you want to play with idioms extension that Parker demo’d, follow the Adding Extensions instructions here: https://en.scratch-wiki.info/wiki/Extension and look for DallE-dioms

More to come soon!

Contact

If you have any questions regarding the seminar, please email one of us: