Free AWS Certified AI Practitioner (AIF-C01) Practice Questions

Sample AIF-C01 Practice Questions

Browse all 10 free AWS Certified AI Practitioner practice questions below.

1. A company is explaining AI terminology to its leadership team. Which statement correctly describes the relationship between artificial intelligence, machine learning, and deep learning?

   • Deep learning is a subset of artificial intelligence that does not involve machine learning.
   • Machine learning is a subset of deep learning, which is a subset of artificial intelligence.
   • Artificial intelligence, machine learning, and deep learning are three independent fields.
   • Deep learning is a subset of machine learning, which is a subset of artificial intelligence.

2. A team is reviewing the cost of a generative AI application. The provider bills based on tokens. What is a token in this context?

   • A single API request made to the foundation model.
   • An authentication credential used to access the model endpoint.
   • A fixed-size block of memory allocated to the model during inference.
   • A unit of text the model processes, roughly a word fragment.

3. A company's HR assistant gives outdated answers about a policy that changed last month. The policy documents live in a company library and are revised several times a year. Which approach keeps answers current without a project for every revision?

   • Fine-tune the model each time a policy document changes.
   • Increase the model's temperature so it produces more varied answers.
   • Use Retrieval Augmented Generation so answers are drawn from the current documents.
   • Include the entire policy handbook in every prompt the assistant receives.

4. A company's recruitment model advances candidates from some universities far more often than others with comparable experience. Which responsible AI feature is most directly engaged?

   • Safety, because rejected candidates may suffer financial harm.
   • Veracity, because the model's outputs may not be factually accurate.
   • Fairness, because outcomes differ systematically across an identifiable group.
   • Robustness, because the model behaves unpredictably on unusual applications.

5. A retailer wants to group its customers into segments without any predefined categories, so the model must discover the groupings from the data itself. Which type of machine learning fits this requirement?

   • Unsupervised learning, because the model finds structure in unlabelled data.
   • Supervised learning, because the model learns from examples with known outcomes.
   • Reinforcement learning, because the model improves through trial and feedback.
   • Transfer learning, because the model reuses knowledge from a related task.

6. A company has 50,000 historical loan applications, each labelled with whether the borrower defaulted. It wants a model that predicts default for new applications. Which type of machine learning applies?

   • Semi-supervised learning, because only some applications have outcomes recorded.
   • Unsupervised learning, because the model discovers structure in the application data.
   • Supervised learning, because the model learns from labelled examples with known outcomes.
   • Reinforcement learning, because the model improves as more loans are issued.

7. A team is preparing a large document library for a retrieval system. Before creating embeddings, they split each document into smaller sections. What is this step called?

   • Chunking
   • Tokenization
   • Distillation
   • Quantization

8. A company wants a managed AWS capability that handles document ingestion, chunking, embedding and retrieval so its assistant can answer from company documents. Which capability provides this?

   • Amazon Comprehend
   • Amazon Bedrock Knowledge Bases
   • Amazon Bedrock Guardrails
   • Amazon Textract

9. A company is describing the business applications of AI agents to its leadership. Which description is accurate?

   • Agents eliminate the need for human oversight of consequential actions.
   • Agents replace the need for foundation models in an application.
   • Agents complete multi-step tasks by using tools to interact with other systems.
   • Agents guarantee deterministic outcomes for every request.

10. A logistics company wants a system that learns the best sequence of routing decisions by trying options and receiving feedback on the outcome of each attempt. Which learning approach fits?

    • Supervised learning
    • Reinforcement learning
    • Unsupervised learning
    • Transfer learning