Fundamentals of AI and ML
This domain builds the vocabulary the rest of the exam depends on. You will learn how artificial intelligence, machine learning, and deep learning nest inside one another, the main learning paradigms, the difference between training and inference, and where core AWS machine learning services fit. Roughly one in five scored questions comes from this domain, so a firm grasp here pays off across the whole test.
AI, machine learning, and deep learning
Artificial intelligence (AI) is the broad goal of getting computers to perform tasks that normally require human intelligence, such as understanding language or recognizing images. Machine learning (ML) is a subset of AI in which systems learn patterns directly from data instead of being explicitly programmed with rules. Deep learning is a further subset of ML that uses multi-layer neural networks, loosely inspired by interconnected neurons in the brain, to model very complex patterns. Remembering the nesting order, deep learning inside ML inside AI, prevents a common category mistake on the exam.
Types of machine learning
Supervised learning trains on labeled examples, where each input is paired with its correct output, so the model can predict labels for new data. Classification predicts a discrete category (spam or not spam), while regression predicts a continuous number (a house price). Unsupervised learning works on unlabeled data to find structure, with clustering grouping similar records together. Reinforcement learning trains an agent to take actions that maximize a reward signal over time. Matching a business scenario to the right paradigm, especially supervised classification versus regression versus unsupervised clustering, is a frequently tested skill.
The ML lifecycle: training, evaluation, and inference
A model is built by feeding it training data, during which its internal parameters (weights) are adjusted to reduce error. To judge how well it will perform on data it has never seen, practitioners hold out a separate test set. Overfitting occurs when a model memorizes the training data, including its noise, and then performs poorly on new data; underfitting is the opposite, where the model is too simple to capture the pattern. Inference is the phase where a finished, trained model generates predictions on new inputs in production.
Features, labels, and data quality
Features are the input variables the model reads, and the label is the target value it learns to predict. The quality, quantity, and representativeness of the training data largely determine how good a model can be, which is why data preparation often consumes most of a project's effort. Poor or biased data produces poor or biased models, a theme that reappears in the Responsible AI and Governance domains.
When to use AI/ML and where AWS fits
Machine learning shines when patterns are complex, changing, or hard to write down as explicit rules, and when enough representative data exists. A simple, stable problem may be better served by a traditional deterministic rule-based system, which is cheaper and more transparent. On AWS, Amazon SageMaker AI is the managed platform for building, training, tuning, and deploying custom models across this lifecycle, while higher-level services such as Amazon Rekognition, Comprehend, and Textract provide ready-made AI capabilities through an API so teams do not have to train models from scratch.