AI Fundamentals and Terminology
AI literacy starts with clean vocabulary. Artificial intelligence is the broad field; machine learning is the subset where systems learn patterns from data rather than following hand-coded rules; deep learning is ML using multi-layer neural networks. Know the difference between narrow AI (task-specific, everything in production today) and general AI (hypothetical). Distinguish supervised learning (labeled data, e.g., classification and regression), unsupervised (clustering, dimensionality reduction), self-supervised (predicting parts of the input, the basis of modern LLMs), and reinforcement learning (an agent learning from reward signals).
Separate training (fitting parameters) from inference (using a trained model). Parameters are learned weights; hyperparameters are set by humans before training (learning rate, batch size). A feature is an input variable; a label is the target. Model, algorithm, and architecture are not synonyms: an algorithm is the procedure, an architecture is the network structure, and a model is the trained artifact.
Traps: questions swap "unsupervised" for "self-supervised," or claim deep learning always beats simpler methods. Also watch for confusing AI with automation (rule-based, no learning) and for definitions that treat generative AI as a learning paradigm rather than a capability.
Rule of thumb: if a human supplied the answer key, it is supervised; if the data supplied its own structure, it is unsupervised; if the model predicts hidden parts of its own input, it is self-supervised. Ask "what is the label?" - if there is none, it is not supervised.