Tools and Other Things

Learning Objects using Gen AI

When statistical pattern-matching is described as "understanding" and probability sampling as "choosing," an entire theory of mind is brought along for the ride, and with it riding shotgun, assumptions about agency, responsibility, and moral status that the technology doesn't warrant.

Yes, metaphors are often necessary for accessibility. However, there is a cost, so why not make it visible: the gap between what the language implies and what the system actually does. As the Mechanistic Mode in these apps states: "There is no moral agent in the machine. Only a product and its producers."

Here are some examples of learning objects, created using Anthropic’s Claude that I’ve used in AI workshops with students.

LLM Trainining Journey An interactive, step-by-step walkthrough of how Large Language Models are built from raw internet data to deployed chatbot. Users navigate through 8 stages: data collection, preprocessing, tokenization, pretraining, attention mechanisms, supervised fine-tuning, reinforcement learning from human feedback, and deployment.

LLM Inference Journey An interactive exploration of what happens when you send a message to an AI chatbot and receive a response. Users navigate through 8 stages: prompt input, tokenization, embedding lookup, attention computation, the prefill phase, the decode phase, KV caching, and output sampling.