AI and ML Core Terminology and Paradigms: 10 practice questions

Question 1 of 10

Which description best defines artificial intelligence?

1. AA deployment process that applies a trained model to new records
2. BA training artifact that stores patterns learned from labeled examples and later applies them to new records
3. CA recipe that maps input attributes to a target during training
4. DThe broad field of creating or imitating machines that perform tasks requiring human-like intelligence

Question 2 of 10

What distinguishes machine learning within the broader field of AI?

1. AIt enables computers to learn from data instead of programming every decision rule explicitly
2. BIt creates models without algorithms, examples, or a training process
3. CIt limits intelligent systems to interpreting images through fixed schemas
4. DIt requires autonomous agents to make goal-directed decisions and take purposeful action on every prediction

Question 3 of 10

Which statement correctly describes deep learning?

1. AIt is a specialized subset of ML that uses neural networks with multiple layers
2. BIt is the act of applying any trained model to new data for a prediction
3. CIt is a fairness method that balances prediction behavior across stakeholder groups
4. DIt is the umbrella field containing AI, machine learning, and computer vision

Question 4 of 10

How do an ML algorithm and an ML model differ?

1. AThe algorithm performs inference only; the model discovers patterns before any training occurs
2. BThe algorithm is the learning recipe; the model is the artifact containing patterns produced by training
3. CThe algorithm measures group fairness; the model removes bias from the source dataset
4. DThe algorithm stores learned patterns; the model supplies the labeled examples used for training

Question 5 of 10

Which sequence correctly relates training and inference?

1. ATraining discovers patterns from example data and produces a model; inference applies that model to new data
2. BInference discovers patterns from example data and produces a model; training applies it to unknown targets
3. CTraining applies an existing prediction to new data; inference converts the result into a learning algorithm
4. DInference labels the training examples first; training then removes the need for a deployed model

Question 6 of 10

Which comparison between bias and fairness is accurate?

1. ABias is poor training-data accuracy; fairness is strong evaluation-data accuracy after deployment
2. BBias measures evaluation speed across models; fairness measures the size of each training dataset
3. CBias means a model has multiple layers; fairness means those layers use the same algorithm
4. DBias is an imbalance across groups in data or predictions; fairness considers effects on stakeholder groups

Question 7 of 10

A model performs well on training examples but poorly on evaluation examples. What does this indicate?

1. AThe model is overfitting because it memorizes seen data and fails to generalize
2. BThe model is underfitting because it captures the training relationship too accurately
3. CThe model is fair because evaluation performance differs from training performance
4. DThe model is performing inference because both datasets contain known target answers

Question 8 of 10

A manufacturer must detect tiny defects in camera images and separately analyze the meaning of technicians' written reports. Which mapping fits both requirements?

1. AUse NLP for the camera images and computer vision for the written reports
2. BUse computer vision for the camera images and NLP for the written reports
3. CUse model fit for the camera images and inference for the written reports
4. DUse fairness for the camera images and training for the written reports

Question 9 of 10

A support system must first draft new response text from a prompt. In a later phase, it must reason over a resolution goal, decide which approved tool to call, and take the action. Which classification is most precise?

1. AThe drafting phase is agentic AI; the decision and action phase is computer vision
2. BThe drafting phase is deep learning; the decision and action phase is model fit
3. CThe drafting phase is inference; the decision and action phase is training data
4. DThe drafting phase is GenAI; the goal-directed decision and action phase is agentic AI

Question 10 of 10

A team needs a pretrained model that can summarize and translate text, while the surrounding application uses prompts to create new text. Which statement distinguishes the model from the application behavior?

1. AThe pretrained language model is an LLM, and creating new prompted text is GenAI behavior
2. BThe pretrained language model is training, and creating new text is labeled-data behavior
3. CThe pretrained language model is an algorithm, and creating new text is model-fit behavior
4. DThe pretrained language model is computer vision, and creating new text is fairness behavior