A broad readiness check, not an exam simulation — 10 scenario questions covering the core AI/ML concepts behind AWS AI Practitioner domain 1.
AI/ML readiness check — 10 questions
1. A retail company wants a system that groups customers into segments based on purchasing behavior, without any predefined segment labels. Which ML approach is this?
2. A robotics team trains a warehouse robot to pick items by giving it a positive signal when it succeeds and a negative signal when it drops an item, repeated over many attempts. Which ML approach is this?
3. A model performs very well on its training data but performs poorly on new, unseen data. What is this problem called, and what is a standard fix?
4. Which of the following best describes the difference between training and inference?
5. A production fraud-detection model's accuracy has quietly declined six months after launch, even though no code changed. What is the most likely cause, and what practice would have caught it earlier?
6. Which statement correctly distinguishes generative AI from traditional discriminative ML models?
7. A hiring-model team wants to check whether their model's predictions differ significantly across gender or ethnicity groups, even though overall accuracy looks good. What are they evaluating for?
8. Which of the following correctly orders the standard ML development lifecycle stages?
9. A dataset of product reviews has no star ratings or sentiment tags attached — just raw text. What kind of data is this, and which ML approach can use it directly for training?
10. What specifically makes a Large Language Model (LLM) a 'generative' AI system rather than a traditional discriminative classifier?