GenAI capabilities, limits and value: AIF-C01 study notes and practice questions
What to know about GenAI capabilities, limits and value for the AIF-C01 exam, from the Fundamentals of Generative AI domain (24% of the exam guide), then 2 free questions to test it. Part of the AIF-C01 study guide.
Weigh performance on your task, cost, speed (latency), context window size, modalities (text, images), customization options and compliance requirements.
Measuring business value
Look at outcomes: conversion rate, average revenue per user, customer satisfaction, time saved, cost per interaction, cross-domain performance and efficiency. Technical scores alone don’t show business value.
Common traps
Using generative AI where exact results are required without checks, such as financial calculations.
Measuring success with model size or token counts instead of business outcomes.
Assuming the model knows recent or private facts. It only knows its training data unless you provide context.
A travel firm's customer chatbot tells a guest, in fluent and confident language, about a spa at a hotel that does not exist.
No one checked the reply before it was sent.
Which generative AI limitation does this error describe?
Choose one answer.
AA. Overfitting, where a model memorizes its training examples too closely
BB. Latency, the time a model takes to return a response
CC. Data drift, a gradual change in the data the model sees in production
DD. Hallucination, output that sounds plausible but is false or unsupported
Show the answer and explanation
Answer: D
A. Wrong: Overfitting is a training problem in which a model learns its training examples too closely and then performs poorly on new inputs. It does not describe a confident, invented answer to a customer question.
B. Wrong: Latency measures how long a response takes to arrive. The problem in this chatbot is the content of the reply, not how quickly it came back.
C. Wrong: Data drift is a change over time in the inputs a deployed model receives, which can lower its accuracy. It describes a pattern across many inputs, not one fabricated fact.
D. Right: Correct. A hallucination is output that reads as fluent and certain but is false or not supported by any source. Grounding answers in trusted data and checking responses reduce the risk.
Explanation
Generative models predict likely text, so they can produce fluent statements that are not true. AWS's overview of generative AI notes that these systems can sometimes produce inaccurate or misleading information. Grounding answers in trusted data and checking responses reduce the risk.
A marketing agency is weighing generative AI for first drafts of client campaigns.
The partners want quick turnaround and low cost per draft, but every draft will still go through human review.
Which TWO statements describe genuine advantages of generative AI for this work? (Choose two.)
Choose two answers.
AA. It creates new draft copy quickly from a short written prompt
BB. It always returns factually accurate claims that need no review
CC. It adapts to new writing tasks without building a separate model for each
DD. It returns identical wording every time the same prompt is run
EE. It explains exactly why each sentence was written the way it was
Show the answer and explanation
Answers: A, C
A. Right: Correct. Generative models produce new content from a prompt in seconds, at a fraction of the time and cost of drafting by hand. This speed is the main reason agencies use them for first drafts.
B. Wrong: Generative AI can state false facts with confidence, so claims still need checking. Accuracy cannot be assumed from fluent output.
C. Right: Correct. Foundation models are general purpose, so one model can draft copy, summarize a brief and suggest headlines. The agency does not need a separate model built for each task.
D. Wrong: Generative AI is usually nondeterministic, so the same prompt can produce different drafts. Consistent wording needs settings and review, not a guarantee from the model.
E. Wrong: Large models have limited interpretability, so it is hard to explain why they wrote a given sentence. This is a drawback to plan around, not an advantage.
Explanation
Generative AI's advantages include fast content creation and general-purpose models that adapt to many tasks. Its drawbacks include inaccurate output, nondeterministic responses and limited explainability, so the agency should keep human review in its workflow.
Want more on GenAI capabilities, limits and value? The 5 full-length practice sets have 16 more GenAI capabilities, limits and value questions, 325 in all, none of them in the free exam. One payment of $14.99, no subscription. See the practice sets →