AI Systems

A convincing caption does not prove a cause. Follow an invented video-annotation task to see what visual reasoning needs to establish and what remains unknown.

A probability is not permission to act. Use a fictional mixed-issue support ticket to examine classification, uncertainty and meaningful human review.

Follow a fictional equipment-loan question through chunk boundaries, embeddings, vector similarity and reranking—then evaluate retrieval separately from the final answer.

Follow a fictional delivery-status prompt to distinguish tokens, Transformer architecture, pretraining, inference and multimodal input from the evidence an answer needs.

Follow a fictional request classifier to separate AI, machine learning, data, training, parameters and inference—and understand what changes when a model generates a reply.

An AI chatbot responds to a message. An AI agent pursues a goal through a sequence of decisions, observations and actions. The distinction is not intelligence in the abstract; it is the operating loop around the model.