ChatGPT can feel like magic, but underneath it all, it's making highly educated guesses based on patterns. This AI Literacy workshop takes an intentionally oversimplified dive into machine learning, neural networks, and deep learning by turning participants into the computers.

They start by writing traditional programming rules for recognizing a cat, then watch those rules break on wolves, cartoon cats, and other edge cases. This leads into an approachable understanding of supervised learning, where computers learn patterns from labeled examples instead of relying on written rules.

Next, they guess missing words in sentences. Familiar phrases are easy because they've seen them before, while random sentences are nearly impossible without context. That contrast leads into an understanding of self-supervised learning, where large language models (LLMs) learn language patterns by guessing, checking, and adjusting across enormous amounts of text. Depending on the version, they also learn to recognize an unfamiliar writing system from examples or train and tune a simplified LLM with code.

By the end, they'll understand how examples become patterns, how patterns become predictions, and what makes training AI models so expensive. The result may seem intelligent, but it's fundamentally pattern recognition at scale, and still requires human guidance and oversight.

A 60 minute talk or hands on workshop for students from middle school through college and working professionals, with versions for technical and nontechnical audiences.