Four hands-on interactive stations that teach core AI concepts by doing. 3–5 minutes each. No technical background required.
A plain-language, five-stage visual walkthrough of how an AI model gets made: data collection, pre-training, post-training, evaluation, and deployment. Step through each stage to see what actually happens and what it reveals about how AI fundamentally works — no choices to make, just the big picture.
The deeper dive through the same five phases. Choose an approach at each step to see how real labs differ in practice, with peer-reviewed or lab-published sources. A concluding section highlights the two decisions that most clearly move environmental impact (grid choice and model size) without pretending to a precision the disclosure landscape doesn’t support.
LLM stands for Large Language Model — the AI behind tools like ChatGPT, Claude, and Gemini. An LLM doesn't look up answers; it predicts the most likely next word, one token at a time. Watch tokens appear live, then adjust the temperature dial to see how the same prompt produces more conventional or more surprising completions. Temperature 0 is fully deterministic; higher temperatures sample from wider probability distributions. (For most major foundation models, temperature is adjustable only via the API, not the chat interface.)
Every AI product is built on three distinct layers: the model, the harness, and the app. Sort twelve products and concepts into the layer each one belongs to, then get instant feedback. The same Claude model powers both claude.ai and Microsoft Copilot — the difference lives in the harness and the app.
Built for nonprofit leaders, social impact professionals, educators, and anyone curious about how generative AI actually works. Works well as a self-guided exploration or as a facilitated teaching tool in workshops and training sessions.