Resource type
Use type
Project name
Newest Unit:
In response to the Everyday AI (EdAI) community’s interests and needs regarding Large Language Models (LLMs), we have developed a new stand alone unit on LLMs that address 1) how they work, 2) how to use them, and 3) implications for society and security. The unit comprises 5 lessons and two optional capstone projects. Lessons accommodate different grade levels to include students whose access to LLMs is restricted because they are 13 or under.
Unit Overview:
This new unit introduces LLMs through a series of critical investigations into the relationships between prompting and generated output as both introductory and culminating lessons. The heart of the unit centers unplugged games and simulations of foundational mechanisms or algorithms that make LLMs work, including word embeddings, attention mechanisms, fine-tuning, and reinforcement learning. The unit includes two optional final projects through which students can 1) debate the benefits and risks of using LLMs and 2) create a public service announcement (PSA) about LLMs for their community.
Essential Questions:
- What’s under the hood? How do LLMs work?
- What are good prompting frameworks for LLMs?
- What are good rationales and methods for using LLMs in real-life (for students and teachers) with considerations for safety and ethics?
-
What are the implications of LLMs for safety and ethics?
Lessons & Projects:
- Lesson 4.0: Intro to LLMs
- Lesson 4.1: Pre-training
- Lesson 4.2: Reinforcement Learning
- Lesson 4.3: Prompting
- Lesson 4.4: Investigating LLMs
- Capstone Projects
Background
Our goal was to develop a new unit that addressed a gap in existing Large Language Model (LLM) curricula. Specifically, we sought to move beyond instruction on how to use LLMs, and focus on how LLMs work empowering learners with a deeper understanding of the underlying mechanisms of LLMs. We are not reinventing the wheel; where our work aligns with established LLM curricula or resources, we provide direct links, fostering a connected learning experience with exposure to a variety of educational resources on LLMs.
As of the time of this post, we are still in the early stages of development. This means that – unlike all other content in the DAILy 2.0 curriculum – the materials in this new unit have not yet been piloted with youth. However, we have piloted and refined these materials through feedback from the Everyday AI community, which includes both experienced EdAI educators and AI literacy experts.
NOTICE: These materials have not yet been piloted with youth. If you plan to implement these materials in an educational context, we recommend additional preparation or a “run-through” to ensure the activities and materials are well suited for your learners.
We welcome your feedback.
If you pilot these materials in an educational context, please feel free to reach out to share your questions and/or experiences. You can email the creator, Katherine (Kate) Moore, [email protected].
Future Directions
We are interested in piloting these materials in the future and, as of the time of publication, are seeking funding to support that work. If readers have interest in participating in this future work (e.g., piloting materials, supporting the research in other ways) please reach out Katherine (Kate) Moore, [email protected].