AI-900 Practice Question: Describe fundamental principles of machine learning on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Having users actively participate in model training by rating AI responses

  Why it's wrong here

  Rating AI responses is the essence of reinforcement learning from human feedback (RLHF): raters provide preference signals that become a reward model, and the base model is fine-tuned to maximize that reward. Active learning, in contrast, is a data-selection strategy that decides which unlabeled examples are most valuable to send to annotators before training. The rating loop improves model behavior, whereas active learning improves labeling efficiency; they operate at different stages of the ML workflow.

• ✓

  Strategically selecting the most informative examples for human labelling to maximise learning efficiency

  Why this is correct

  Active learning is a human-in-the-loop labeling strategy in which the model itself selects the most informative unlabeled instances—typically those with the highest predictive uncertainty, query-by-committee disagreement, or expected model change—for a human to label. By targeting examples that would most reduce model error, it achieves high accuracy with far fewer labeled samples than random sampling. This maximizes learning efficiency because each annotation contributes more discriminative information, lowering annotation cost while preserving model performance.

• ✗

  A training approach where the model actively searches the internet for additional training data

  Why it's wrong here

  Searching the internet or scraping external websites is a raw data acquisition technique that expands the unlabeled corpus; it neither prioritises which existing examples to annotate nor necessarily improves model performance. Active learning presupposes a pool of unlabeled data and chooses the next samples to label, rather than discovering new data sources. Uncontrolled web harvesting can also pull out-of-distribution or noisy content, whereas active learning deliberately samples from the target task distribution to improve per-label value.

• ✗

  Continuous model training that runs actively in the background as new data arrives

  Why it's wrong here

  Continuously retraining or updating a model in the background as new batches or streams of data arrive is online or continuous learning, an architectural pattern for keeping models current. It does not inherently reduce labeling effort, because the model still needs labeled data from each incoming stream. Active learning is specifically a labeling strategy for batch annotation workflows, where a finite set of unlabeled examples is ranked by informativeness and only the top examples are sent for human annotation. The two can be combined—active learning can choose which streaming data to label before online updates—but they remain distinct concepts.

Go deeper

Related to this question