AWS AIF-C01 Practice Test

AIF-C01 practice test: what's covered

• AI Practitioner Foundational

Free AIF-C01 sample questions

1. A product manager is explaining to executives where generative AI sits relative to other terms. How should the relationship between AI, machine learning, deep learning, and generative AI be described?

   • A. AI is the broadest field, machine learning is a subset that learns from data, deep learning is ML using neural networks, and generative AI uses deep learning to create new content
   • B. Generative AI is the broadest field and contains artificial intelligence, which contains deep learning and then machine learning at the core
   • C. Machine learning and artificial intelligence are unrelated disciplines, while deep learning and generative AI are interchangeable terms for the same thing
   • D. Deep learning is the broadest umbrella term and machine learning is a marketing label that has no technical meaning in modern systems
   Show answer

   Correct: A. The correct nesting is AI ⊃ ML ⊃ deep learning, with generative AI being an application of deep learning that produces new text, images, or audio. The other options are incorrect because generative AI is not the broadest field, machine learning and AI are not unrelated, and deep learning is not the umbrella term, whereas AI is the overarching discipline.

2. A retailer wants to group customers into segments without any predefined labels so the marketing team can discover natural buying patterns. Which type of machine learning fits this goal?

   • A. Unsupervised learning, because it finds structure and clusters in unlabeled data
   • B. Supervised learning, because it requires labeled target values for every customer record
   • C. Reinforcement learning, because it optimizes a cumulative reward through trial and error
   • D. Transfer learning, because it copies labels directly from a different pretrained dataset
   Show answer

   Correct: A. Customer segmentation with no labels is a classic unsupervised clustering task. Supervised learning is incorrect because it needs labeled targets, reinforcement learning is incorrect because there is no agent acting in an environment for rewards, and transfer learning is incorrect because it adapts a pretrained model rather than discovering clusters in unlabeled data.

3. An engineer notices that building the model takes days on GPU clusters but serving predictions to users returns in milliseconds on smaller hardware. How are these two phases best characterized?

   • A. Training builds the model from data and is compute-intensive, while inference uses the trained model to make predictions and is comparatively fast
   • B. Training and inference use identical hardware and cost, so they can always be run on the same instance without any optimization
   • C. Inference is the slow phase that learns patterns, while training is the fast phase that only applies the learned weights
   • D. Training only stores raw data unchanged, while inference is the step that actually creates the model parameters from scratch
   Show answer

   Correct: A. Training learns parameters from data and is resource-heavy, whereas inference applies the finished model and is optimized for low latency. The reversed option is incorrect because inference does not learn, the identical-hardware claim is incorrect because the workloads differ, and training does not merely store raw data unchanged.

4. A startup is scoping its first generative AI pilots and wants examples that genuinely fit the technology's strengths. Which set of use cases is most appropriate for generative AI?

   • A. Chat assistants, document summarization, code generation, and image creation
   • B. Guaranteeing perfectly accurate financial calculations with no possibility of error
   • C. Providing legally binding medical diagnoses without any human review whatsoever
   • D. Replacing deterministic database transactions for transferring money between accounts
   Show answer

   Correct: A. Generative AI excels at producing language, code, and media, making assistants, summarization, and image generation strong fits. The other options are incorrect because generative models are probabilistic and can hallucinate, so they are unsuitable for guaranteed-exact math, unreviewed medical diagnoses, or deterministic money-transfer transactions that require strict correctness.

5. A team wants to call several foundation models such as Anthropic Claude, Meta Llama, and Amazon Titan through one API without provisioning any servers. Which AWS service is designed for this?

   • A. Amazon Bedrock, a fully managed service offering foundation models from multiple providers through a single API
   • B. Amazon EC2, where the team must manually install and patch each model on self-managed virtual machines
   • C. Amazon Rekognition, a managed computer vision service for detecting objects and faces in images
   • D. Amazon Comprehend, a managed natural language service for sentiment and entity extraction only
   Show answer

   Correct: A. Amazon Bedrock provides serverless access to multiple foundation models behind one API with pay-per-token pricing. EC2 is incorrect because it requires self-managing infrastructure, Rekognition is incorrect because it only does vision, and Comprehend is incorrect because it is limited to NLP tasks rather than offering general foundation models.

Last updated: 2026-10-19

AIF-C01 practice test FAQ

Related practice tests

• AWS CLF-C02 practice test
• AWS MLA-C01 practice test
• Google Cloud Generative AI Leader practice test
• AWS SAA-C03 practice test
• NCA-GENL practice test
• Databricks Generative AI Engineer Associate practice test