Master CompTIA AI Fundamentals with Confidence

1. A new employee is trying to understand the relationship between the terms artificial intelligence (AI), machine learning (ML), and generative AI. Which statement BEST describes how these three concepts relate to one another?

   • A. Machine learning is the broadest field, and both AI and generative AI are smaller specialties contained within it.
   • B. Artificial intelligence is the broad field, machine learning is a subset of AI, and generative AI is a subset of machine learning.
   • C. The three terms are interchangeable and describe exactly the same technology.
   • D. Generative AI is the broad field, and artificial intelligence and machine learning are types of generative AI tools.

   Answer: B. AI is the broad umbrella concept of machines performing tasks that appear intelligent; machine learning is a subset of AI in which systems learn patterns from data rather than following only hand-coded rules; generative AI is a further subset of ML focused on creating new content. Option A inverts the hierarchy, C is wrong because the terms describe different scopes, and D wrongly places the narrowest term (generative AI) at the top.

   Exam domain: AI concepts and how AI works

2. A retailer uses two different systems. System 1 analyzes historical sales data to predict which customers are likely to cancel their subscriptions. System 2 produces brand-new product descriptions and marketing images from a short text prompt. Which classification correctly describes each system?

   • A. Both systems are examples of generative AI because they both use data.
   • B. System 1 is best described as predictive machine learning, while System 2 is best described as generative AI.
   • C. System 1 is generative AI, while System 2 is traditional rule-based software.
   • D. Neither system involves artificial intelligence because they are automated business tools.

   Answer: B. System 1 learns patterns from past data to make a prediction (a churn likelihood), which is classic predictive machine learning. System 2 creates original content (text and images), the defining behavior of generative AI. Option A overgeneralizes since using data does not make a tool generative, C reverses the categories, and D is wrong because both are AI-based capabilities.

   Exam domain: AI concepts and how AI works

3. Which characteristic MOST clearly distinguishes a machine learning approach from traditional, explicitly programmed software?

   • A. Machine learning systems never make mistakes once they are deployed.
   • B. Machine learning systems learn patterns from examples in data rather than relying solely on rules written by a programmer.
   • C. Machine learning systems run only in the cloud, while traditional software runs only on local computers.
   • D. Machine learning systems do not require any data to function.

   Answer: B. The defining trait of machine learning is that behavior is learned from data and examples instead of being fully specified by hand-coded rules. Option A is false because ML systems can and do make errors, C describes an unrelated deployment detail that is also factually untrue, and D is the opposite of the truth, as ML depends heavily on data.

   Exam domain: AI concepts and how AI works

4. At a conceptual level, how does a large language model (LLM) generate the text in its response to a user?

   • A. It searches a built-in database of pre-written answers and returns the closest exact match.
   • B. It predicts the next token from learned patterns and prior words.
   • C. It connects live to the author of every source document to confirm each sentence before replying.
   • D. It randomly selects complete sentences with no relationship to the user's prompt.

   Answer: B. An LLM works by repeatedly predicting a likely next token given the prompt and the text generated so far, assembling a response piece by piece. Option A describes a lookup table rather than a generative model, C is not how LLMs operate, and D ignores that outputs are conditioned on the prompt and learned patterns.

   Exam domain: AI concepts and how AI works

5. A marketing coordinator needs to generate a dozen rough first-draft social media captions to brainstorm ideas before a human edits and approves the best ones. Why is this task a GOOD fit for a generative AI chatbot?

   • A. AI guarantees that every caption it produces will be factually verified and legally cleared
   • B. The task involves producing many low-stakes creative drafts quickly, with a human reviewing the output before use
   • C. AI removes the need for any human involvement in the marketing process
   • D. AI can legally bind the company to whatever promises appear in the captions

   Answer: B. Generative AI excels at quickly producing volume drafts for brainstorming and creative ideation when the stakes are low and a human reviews before publishing. Options A and D overstate AI's reliability and authority (AI does not verify facts or carry legal authority), and C is wrong because human review remains essential.

   Exam domain: When to use AI (appropriate use)

6. Which of the following tasks is the LEAST appropriate use of a general-purpose AI chatbot?

   • A. Drafting a friendly out-of-office email reply
   • B. Summarizing the key points of a long article you provide
   • C. Making a final, unreviewed decision to deny a claim
   • D. Suggesting synonyms to vary the wording of a paragraph

   Answer: C. High-stakes decisions that affect people's rights, finances, or well-being should not be made autonomously by AI without human oversight, because AI can be biased or wrong and lacks accountability. The other tasks are low-stakes drafting and summarization activities where AI assistance is appropriate, especially with light human review.

   Exam domain: When to use AI (appropriate use)

7. A small accounting firm wants to use a public AI chatbot to draft a client's tax return by pasting in the client's full name, Social Security number, and income details. Beyond accuracy concerns, what is the PRIMARY reason this is a poor fit?

   • A. Chatbots can only process text written in English
   • B. AI chatbots are physically incapable of producing numbers
   • C. Tax returns are too short for an AI to be useful
   • D. Pasting sensitive personal data into a public tool risks exposure

   Answer: D. Entering confidential personal data such as Social Security numbers into a public AI tool can expose that data and violate privacy and data-protection obligations, making it an inappropriate use. The other options are factually false: chatbots handle multiple languages and numbers, and length is not the issue here.

   Exam domain: When to use AI (appropriate use)

8. Which of the following is a well-recognized STRENGTH of generative AI chatbots?

   • A. Personally taking full legal responsibility for the content it creates
   • B. Guaranteeing the output contains no bias
   • C. Always producing 100% factually accurate answers
   • D. Quickly generating drafts and ideas across many topics in seconds

   Answer: D. A core strength of generative AI is rapidly producing drafts, ideas, and text across a wide range of topics. The other options describe things AI cannot guarantee: it can be inaccurate, can reflect bias from training data, and cannot bear legal accountability.

   Exam domain: When to use AI (appropriate use)

9. A company trains an AI resume-screening tool using ten years of its own past hiring decisions, during which it rarely hired women for engineering roles. The tool now tends to rank male applicants higher. What is the most likely cause of this outcome?

   • A. The model is hallucinating applicant qualifications that do not exist
   • B. The training data reflected historical bias, which the model learned and reproduced
   • C. The applicants intentionally entered misleading information in their resumes
   • D. The model lacked enough computing power to evaluate female candidates

   Answer: B. AI systems learn patterns from their training data, so biased historical data leads to biased predictions, a phenomenon often called data or historical bias. Hallucination (A) refers to fabricated content, not skewed rankings from real data. Applicant behavior (C) and compute limits (D) do not explain a consistent, systematic preference traced to the historical hiring record.

   Exam domain: Responsible and ethical AI

10. A healthcare organization wants to reduce the chance that its AI risk-prediction model treats one demographic group less favorably than another. Which approach BEST addresses fairness before the model is deployed?

    • A. Increase the model's response speed so results are returned faster
    • B. Audit the training data and model outputs across different demographic groups for disparities
    • C. Hide the demographic fields from end users so they cannot see the predictions
    • D. Add a disclaimer telling users the model may occasionally be wrong

    Answer: B. Proactively examining training data and testing outputs across groups helps detect and correct unfair disparities before deployment. Faster responses (A) and disclaimers (D) do nothing to reduce bias, and merely hiding fields from users (C) conceals the problem rather than fixing the underlying disparity in the model.

    Exam domain: Responsible and ethical AI

Want more? Finish the quiz above to unlock 10 more free questions, or open the full CompTIA AI Fundamentals practice app.