Free PCEI™ – Certified Entry-Level AI Specialist with Python Practice Questions

Sample Practice Questions

Try these 10 sample questions from the PCEI™ – Certified Entry-Level AI Specialist with Python practice bank. Each includes the answer and explanation. Continue with the interactive practice above to review the full question set.

1. 1Which description best captures the purpose of artificial intelligence?

   1. A. Representing information as fixed records for later retrieval
   2. B. Enabling machines to perform tasks associated with human intelligence
   3. C. Transmitting information between computers using network protocols
   4. D. Executing arithmetic instructions without interpreting their results

   Answer: B

   Explanation: IBM defines AI as "technology that enables computers and machines to simulate human learning, comprehension, problem solving, decision making, creativity and autonomy." This concerns intelligent capabilities; storage, networking, and arithmetic alone do not define AI.

   Reference: What Is Artificial Intelligence (AI)? — What is AI?

2. 2A model learns to assign support messages to predefined categories using examples already paired with the correct category. Which learning setup is this?

   1. A. Supervised learning
   2. B. Unsupervised clustering
   3. C. Reward-based reinforcement learning
   4. D. Manually specified rule execution

   Answer: A

   Explanation: Google states that supervised models learn from "data with the known correct results." The paired categories supply those results. Unsupervised clustering lacks supplied targets, reinforcement learning uses action rewards, and fixed rule execution does not learn from these labeled examples.

   Reference: What Is Machine Learning? — Supervised learning

3. 3An open text file f contains a valid JSON document. Which call reads it and returns the corresponding Python object?

   1. A. json.dumps(f)
   2. B. json.dump(f)
   3. C. json.loads(f)
   4. D. json.load(f)

   Answer: D

   Explanation: Python documents json.load as "Deserialize fp to a Python object". load reads a file-like object; loads expects JSON text rather than a file object. dump and dumps serialize Python objects in the opposite direction.

   Reference: Python Standard Library: JSON Encoder and Decoder — Basic Usage, json.load

4. 4Before applying an activation function, a neuron has inputs 2 and 3, corresponding weights 0.5 and 1, and bias -1. What is its weighted sum plus bias?

   1. A. 3
   2. B. 4
   3. C. 5
   4. D. 6

   Answer: A

   Explanation: Google describes taking "the sum of the product of each of its inputs" and a weight, "plus the bias." Thus 2 × 0.5 + 3 × 1 - 1 = 3. Four omits the bias; 5 adds unweighted inputs, and 6 neither applies the stated weights nor subtracts the bias correctly.

   Reference: Neural Networks: Nodes and Hidden Layers — Adding layers to the network

5. 5An AI screening system consistently disadvantages one demographic group even when applicants have comparable task-relevant qualifications. Which ethical concern is most directly illustrated?

   1. A. Loss of confidentiality in stored records
   2. B. Unfair discrimination in model outcomes
   3. C. Insufficient availability of the service
   4. D. Lack of version information for the model

   Answer: B

   Explanation: NIST states that fairness addresses "harmful bias and discrimination". Unequal adverse treatment of comparably qualified groups is a fairness concern. The facts do not describe disclosure, service downtime, or missing version records.

   Reference: Artificial Intelligence Risk Management Framework (AI RMF 1.0) — Section 3.7 Fair – with Harmful Bias Managed, page 17

6. 6A program must convert kilograms to grams using the exact rule grams = kilograms × 1000. No uncertain pattern needs to be learned. Which approach directly meets this requirement?

   1. A. Train an image classifier on labeled weight measurements
   2. B. Cluster historical conversions and select a cluster center
   3. C. Implement the specified arithmetic rule in ordinary code
   4. D. Train a language model to guess each numerical conversion

   Answer: C

   Explanation: Objective 6.1 distinguishes AI problems from those requiring "traditional programming or human judgment". A fixed exact conversion is directly expressed as arithmetic code. Learning categories, clusters, or guessed text adds no needed inference and does not directly implement the exact requirement.

   Reference: PCEI-30-01 Exam Syllabus — Page 11, Objective 6.1 B

7. 7A system excels at identifying defective parts in factory images but has no general ability to learn unrelated intellectual tasks. How should its capability be classified?

   1. A. General AI, because it applies knowledge across unrelated tasks
   2. B. General AI, because its accuracy matches a human inspector
   3. C. Narrow AI, because its abilities are designed for a specific task
   4. D. Narrow AI, because its decisions must be programmed individually

   Answer: C

   Explanation: IBM describes narrow AI as "AI systems designed to perform a specific task or a set of tasks." Specialized image inspection fits this description. High accuracy does not establish general intelligence, and narrow AI can learn rather than require an individually programmed decision for every image.

   Reference: What Is Artificial Intelligence (AI)? — Weak AI vs. Strong AI

8. 8A retailer wants to discover natural groups of shopping behavior without supplying predefined group labels. Which learning approach fits this task?

   1. A. Supervised classification against known categories
   2. B. Supervised regression against numerical targets
   3. C. Unsupervised learning through clustering
   4. D. Reinforcement learning from action rewards

   Answer: C

   Explanation: Google explains that clustering organizes similar data into groups and that "the categories aren't defined by you." Discovering groups without targets is unsupervised learning. Classification and regression learn supplied targets; reinforcement learning requires an action-and-reward setup.

   Reference: What Is Machine Learning? — Unsupervised learning

9. 9A file is read with csv.reader using its default settings. A field contains the characters 42. What type is the corresponding field value returned by the reader?

   1. A. str
   2. B. int
   3. C. float
   4. D. bool

   Answer: A

   Explanation: Python states, "Each row read from the csv file is returned as a list of strings." Default csv.reader does not infer numeric types, so 42 remains text until explicitly converted. The QUOTE_NONNUMERIC exception is not enabled in the question.

   Reference: Python Standard Library: CSV File Reading and Writing — Module Contents, csv.reader

10. 10Why are nonlinear activation functions used between layers of a neural network?

    1. A. To compute loss-based changes to the network's connection weights
    2. B. To preserve an exclusively linear mapping as more layers are added
    3. C. To enable the network to model nonlinear relationships
    4. D. To pass every weighted input sum onward as its unchanged value

    Answer: C

    Explanation: Google states, "Stacking nonlinearities on nonlinearities lets us model very complicated relationships". Nonlinear activations make nonlinear relationships representable. Computing weight changes is a training operation; preserving linearity or passing every sum unchanged fails to add the required nonlinear transformation.

    Reference: Neural Networks: Activation Functions — Neural Networks: Activation Functions, nonlinear transforms