AI Fundamentals and Terminology
Core distinctions
Start with the boundaries between AI, machine learning, and deep learning: AI is the broad goal of machine intelligence, ML is the subset where systems learn patterns from data rather than following hand-written rules, and deep learning is ML using multi-layer neural networks. Be able to classify a system correctly (a rules-based chatbot is AI but not ML).
Learning paradigms and task types
Know supervised learning (labeled data; classification and regression), unsupervised learning (clustering, dimensionality reduction), self-supervised learning (labels generated from the data itself, the basis of modern language models), and reinforcement learning (an agent, actions, rewards, and a policy). Distinguish classification from regression by output type, not by algorithm name.
Vocabulary traps
Questions often test precision: a model is the learned artifact, an algorithm is the procedure, training fits parameters, inference applies them. Distinguish parameters (learned weights) from hyperparameters (set before training). Know that features are model inputs and labels are targets. Understand terms like ground truth, overfitting, generalization, and benchmark.
Worked example
A system that flags fraudulent transactions from past labeled fraud cases is supervised classification. If it instead groups transactions by similarity with no labels, it is unsupervised clustering. If it learns by receiving penalties for missed fraud over time, it is reinforcement learning. Ask: are labels present? Is the output a category or a number? Is feedback a reward?
Rule of thumb
When a term is ambiguous, anchor on data: what goes in, what comes out, and where the labels come from.