AI certifications AI fundamentals practice questions
10 free questions with answers and explanations. Try each one, then tap “Show answer”.
- ML basics
1. Training a model with labelled examples is
- A. supervised learning
- B. rule-based programming
- C. reinforcement learning
- D. unsupervised learning
Show answer
Answer: A. supervised learning
Unsupervised learning finds patterns in unlabelled data.
- Evaluation
2. A classifier's confusion matrix: TP = 42, FP = 13, FN = 25, TN = 74. What is its precision?
- A. 76.4%
- B. 75.3%
- C. 85.1%
- D. 62.7%
Show answer
Answer: A. 76.4%
Precision = TP ÷ (TP + FP) = 42 ÷ 55 = 76.4%.
- ML basics
3. A model that performs very well on training data but poorly on new data is
- A. overfitting
- B. well-generalised
- C. underfitting
- D. unsupervised
Show answer
Answer: A. overfitting
More data or regularisation can help.
- Responsible AI
4. Which is one of Microsoft's six responsible AI principles?
- A. Secrecy
- B. Speed
- C. Fairness
- D. Profitability
Show answer
Answer: C. Fairness
The others: reliability and safety, privacy and security, inclusiveness, transparency, accountability.
- Evaluation
5. A classifier's confusion matrix: TP = 90, FP = 17, FN = 22, TN = 143. What is its accuracy?
- A. 84.1%
- B. 80.4%
- C. 85.7%
- D. 89.4%
Show answer
Answer: C. 85.7%
Accuracy = (TP + TN) ÷ total = 233 ÷ 272 = 85.7%.
- ML basics
6. Predicting a house price is an example of
- A. classification
- B. object detection
- C. regression
- D. clustering
Show answer
Answer: C. regression
Regression predicts a number.
- Evaluation
7. Data used only to estimate performance on unseen data after training is the
- A. training set
- B. label set
- C. test set
- D. feature store
Show answer
Answer: C. test set
A validation set helps tune the model during development.
- Generative AI
8. Raising the 'temperature' setting of an LLM makes its output
- A. more accurate always
- B. more random and creative
- C. more deterministic
- D. shorter only
Show answer
Answer: B. more random and creative
Lower temperature gives more predictable answers.
- ML basics
9. Learning by trial and error with rewards and penalties is
- A. unsupervised learning
- B. supervised learning
- C. reinforcement learning
- D. transfer learning
Show answer
Answer: C. reinforcement learning
Used in game playing and robotics.
- AWS services
10. Which AWS service gives access to foundation models through an API?
- A. Amazon Rekognition
- B. Amazon Polly
- C. AWS Glue
- D. Amazon Bedrock
Show answer
Answer: D. Amazon Bedrock
SageMaker is for building and training your own models.
Get fresh questions every day, focused on your weak topics, plus timed mock tests. Free to start.
Practise free in the web app Practise on TelegramAbout the AI certifications on Mock Night · Original practice questions; not affiliated with the exam board.