Free PECB Certified Artificial Intelligence Professional exam Practice Questions

Sample Practice Questions

Try these 10 sample questions from the PECB Certified Artificial Intelligence Professional exam practice bank. Each includes the answer and explanation. Continue with the interactive practice above to review the full question set.

1. 1A vision model is built exclusively to identify faulty welds. Which capability category describes its intended scope?

   1. A.Narrow artificial intelligenceCorrect answer: A
   2. B.Artificial general intelligence
   3. C.Artificial superintelligence
   4. D.Human-level general reasoning

   Explanation: IBM states, "Weak AI: Also known as “narrow AI,” defines AI systems designed to perform a specific task or a set of tasks." The model's capability is confined to weld inspection. General intelligence requires broad transfer across tasks; superintelligence exceeds human general capability, and human-level general reasoning is not established by one specialized model.

   Reference: IBM — What is artificial intelligence? — Weak AI vs. Strong AI

2. 2An analyst wants to inspect the distribution of continuous sensor readings across numeric intervals. Which visualization directly shows the count in each interval?

   1. A.HistogramCorrect answer: A
   2. B.Pie chart
   3. C.Scatter plot
   4. D.Network graph

   Explanation: The source states, "Histograms: This graph plots a distribution of numbers using a bar chart" A histogram groups numeric values into intervals. Pie charts encode shares, scatter plots paired measurements, and network graphs entity relationships.

   Reference: IBM — Data visualization — Types of data visualizations

3. 3A model predicts delivery time in minutes from distance and traffic measurements. What supervised task does this define?

   1. A.RegressionCorrect answer: A
   2. B.Clustering
   3. C.Association discovery
   4. D.Dimensionality reduction

   Explanation: The source states, "Regression models predict continuous values" Delivery duration is continuous. Clustering groups unlabeled observations, association finds co-occurrence, and dimensionality reduction compresses features.

   Reference: IBM — Machine learning — Supervised learning

4. 4In a neural unit, which parameter controls how strongly a particular input affects the weighted sum?

   1. A.Its weightCorrect answer: A
   2. B.Its bias
   3. C.Its optimizer learning rate
   4. D.Its activation threshold

   Explanation: The source states, "Weights act like dials that control how strongly each input feature influences the decision". A weight multiplies its particular input. Bias shifts the unit offset; learning rate controls update size; an activation threshold governs response rather than the coefficient of that input.

   Reference: IBM — Neural networks — How do neural networks work?

5. 5A vision system assigns one predefined category to an entire photograph without locating individual objects. Which output task is being performed?

   1. A.Image classificationCorrect answer: A
   2. B.Object detection
   3. C.Instance segmentation
   4. D.Object tracking

   Explanation: The source states, "Image classification is a core computer vision task that categorizes images into predefined groups or classes." Classification supplies the image label. Detection adds localization, instance segmentation supplies per-instance pixel boundaries, and tracking follows objects over time.

   Reference: IBM — Computer vision — Image classification

6. 6An AI agent may send external emails. Which control best limits unauthorized high-impact actions while preserving supervised use?

   1. A.Validate each email against a fixed output schema.
   2. B.Scan the email text for confidential information.
   3. C.Require user confirmation and constrain extension permissions.Correct answer: C
   4. D.Check the email's factual claims against retrieved sources.

   Explanation: PECB includes "extension permission control, user confirmation protocols, and logging of agent actions". Confirmation and permission limits govern authority to send. Schema checking validates format, scanning targets disclosure and grounding checks facts; those controls alone do not authorize an external action.

   Reference: PECB CAIP Candidate Handbook — p. 12 Knowledge statements, item 6

7. 7An organization has more AI risks than it can treat at once. What should primarily guide allocation of risk-management resources?

   1. A.Assessed risk level and potential impactCorrect answer: A
   2. B.Model training cost and hardware investment
   3. C.The number of parameters in each model
   4. D.The number of users requesting new features

   Explanation: NIST states, "Policies and resources should be prioritized based on the assessed risk level and potential impact of an AI system." Risk and impact guide treatment priority. Training investment, parameter count and feature demand may inform other management decisions but do not determine the level and consequences of the assessed risks.

   Reference: NIST — Artificial Intelligence Risk Management Framework (AI RMF 1.0) — §1.2.3 Risk Prioritization

8. 8Which capability would distinguish artificial general intelligence from a specialized speech transcription system?

   1. A.Transcribing recordings in one trained domain
   2. B.Applying learned knowledge across a wide range of tasksCorrect answer: B
   3. C.Running a transcription service continuously
   4. D.Processing speech faster on a larger GPU

   Explanation: IBM states, "possess the ability to understand, learn and apply knowledge across a wide range of tasks" Broad cross-task application is the distinguishing capability. Domain transcription remains specialized; continuous service and faster hardware concern operation and speed, not generality.

   Reference: IBM — What is artificial intelligence? — Weak AI vs. Strong AI — Strong AI

9. 9A study records both temperature and energy use for each observation. Which chart best reveals how those two measured variables vary together?

   1. A.Stacked area chart
   2. B.Scatter plotCorrect answer: B
   3. C.Pie chart
   4. D.Gantt chart

   Explanation: The source states, "Scatter plots: These visuals are beneficial in reveling the relationship between two variables" Each pair can be positioned on two numeric axes. Stacked areas emphasize cumulative time series, pie charts composition and Gantt charts schedules.

   Reference: IBM — Data visualization — Types of data visualizations

10. 10A training dataset pairs each input with an independently supplied correct outcome. Which learning paradigm uses that ground truth directly?

    1. A.Unsupervised learning
    2. B.Supervised learningCorrect answer: B
    3. C.Reinforcement learning
    4. D.Unsupervised association mining

    Explanation: The source states, "In conventional supervised learning, that ground truth is provided by labeled data." Supplied correct outcomes supervise prediction. The unsupervised alternatives lack external target labels; reinforcement learning uses action-dependent rewards rather than these independent correct outcomes.

    Reference: IBM — Machine learning — Supervised learning