These materials are designed to power roughly 8 hours of class time focused on demystifying LLM technology for people who have no software background whatsoever. There are four broad segments:
-
Technical details about LLMs. (lecture driven)
- How they are constructed.
- How they are trained and that process works.
- Strengths and limitations that stem from these decisions.
-
Case Studies. (discussion driven)
- Actual examples of LLMs that have been deployed, and what happened as a result.
- How things can go poorly.
- How things can go well.
- How to mitigate errors.
-
The current corporate and academic landscape related to LLM technology. (research driven)
- Who is performing research?
- What kinds of LLM products and services are there?
- Who is providing which kinds of LLM services?
- Impact and politics around open and closed models.
-
Using LLMs. (exercise driven)
- Prompt strategies
- Lots of practice
- Files are named to indicate the order in which they should be delivered.
- Any markdown files are lecture notes, written with the expectation that you'll have a (digital) blackboard available where you can illustrate many of the concepts in the lecture.
- If you won't have that, you may wish to create some diagrams or slides to accompany the lesson plans.
- Periodically there are "micro" and "mini" exercises embedded in the markdown and Python files.
- Micro exercises should take about 1-2 minutes to complete.
- Mini exercises should take about 5 minutes to complete.
- There are also a variety of "full" exercises.
- Exercises in this repo describe their own timing expectations.
These materials were created by Tyler Bettilyon and Teb's Lab. You can support the creation of more free, open source, public domain educational materials by sharing them with others subscribing to our newsletter, or signing your team up for one of our corporate training classes, or signing yourself up for one of our open enrollment classes (when available).