Generative AI Leader Free Practice Questions — Page 2
Question 11
A customer service team wants to use generative AI to improve the quality and consistency of their email responses to customer inquiries. They need a solution that can guide the AI to adopt a helpful, empathetic tone while adhering to company policies. Which prompting technique should they use?
A. Role prompting that instructs the AI to act as an experienced customer service representative with corporate knowledge.
B. Few-shot prompting that provides examples of good and bad customer service emails.
C. Prompt chaining that engages the AI in a conversation to gather the necessary information before generating the email response.
D. One-shot prompting that provides a single example of a good customer service email.
Show Answer
Correct Answer: A
Explanation: Role prompting is the best fit because it explicitly instructs the model to behave as an experienced customer service representative, shaping its tone (helpful and empathetic) while following company policies and organizational context. Few-shot prompting can improve consistency through examples, but the primary requirement is to establish the desired role, behavior, and policy adherence.
Sources:
https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/safety-system-instructions
Question 12
What are core hardware components of the infrastructure layer in the generative AI landscape?
A. User interfaces
B. TPUs and GPUs
C. Tools and services for building AI models
D. Pre-trained models
Show Answer
Correct Answer: B
Explanation: The infrastructure layer of the generative AI stack consists of the underlying compute hardware used to train and run models. GPUs (Graphics Processing Units) and TPUs (Tensor Processing Units) are specialized accelerators designed for the large-scale matrix computations required by deep learning. The other options belong to higher layers: user interfaces are application-facing, tools/services are development platforms, and pre-trained models are model-layer assets.
Question 13
A company wants to build a model to classify customer reviews as positive, negative, or neutral. They have collected a dataset of thousands of customer reviews, and each review has been manually tagged with the corresponding sentiment: positive, negative, or neutral. What machine learning should the company use?
A. Reinforcement learning
B. Deep learning
C. Unsupervised learning
D. Supervised learning
Show Answer
Correct Answer: D
Explanation: Because the dataset consists of labeled examples (customer reviews with manually assigned sentiment labels), this is a supervised learning classification task. The model learns to map inputs to known target labels (positive, negative, or neutral). Reinforcement learning is for learning through rewards, unsupervised learning uses unlabeled data, and deep learning is a model family rather than the learning paradigm required by the question.
Question 14
A marketing team wants to use a generative AI model to create product descriptions for their new line of eco-friendly water bottles. They provide a brief prompt stating, “Write a product description for our new water bottle.” The model generates a generic, lackluster description that is factually accurate but lacks engaging language and doesn’t highlight the environmental benefits that are key to their brand. What should the marketing team do to overcome this limitation of the generated product description?
A. Lower the temperature setting of the model to produce more consistent results.
B. Increase the token count for the model to allow for longer descriptions.
C. Add details to the prompt about the audience, tone, and keywords.
D. Train the model on a dataset of marketing materials from other eco-friendly brands.
Show Answer
Correct Answer: C
Explanation: The issue is an underspecified prompt. Improving the prompt with details about the target audience, desired tone, brand values, environmental benefits, and important keywords gives the model the context needed to generate a more engaging and relevant product description. Lowering temperature mainly affects variability, increasing token count affects maximum length, and training a new model is unnecessary for this use case.
Question 15
A team is using a generative AI model to automatically generate short summaries of customer feedback. They need to ensure that these summaries are concise and easy to digest. What model setting should they adjust?
A. Temperature
B. Output length
C. Top-p (nucleus sampling)
D. Safety settings
Show Answer
Correct Answer: B
Explanation: Adjust the output length (often controlled by a maximum output tokens parameter) to make summaries more concise. Temperature affects randomness, top-p affects token sampling diversity, and safety settings govern content filtering rather than summary length.
Question 16
A national bank is overwhelmed by customer inquiries across multiple channels and needs an AI-powered solution to provide seamless, consistent support, empower customer support agents, and improve service quality. What Google Cloud product should the bank use?
A. Gemini for Google Workspace
B. Google Contact Center as a Service
C. Vertex AI Search
D. Gemini for Google Cloud
Show Answer
Correct Answer: B
Explanation: Google Contact Center as a Service (CCaaS) is the Google Cloud solution built for omnichannel customer support. It provides AI-powered virtual agents, agent assist, consistent experiences across channels, and tools to improve service quality and agent productivity. The other options target productivity apps (Gemini for Workspace), cloud administration (Gemini for Google Cloud), or enterprise search rather than a full contact center platform.
Question 17
What does Model Garden enable a company to do?
A. Train new models from scratch using large datasets.
B. Evaluate the performance of different models using various metrics.
C. Discover, customize, and deploy existing models from Google and its partners.
D. Manage different versions of a model including the code, data, and parameters used to train it.
Show Answer
Correct Answer: C
Explanation: Vertex AI Model Garden is a catalog of foundation and open models from Google and partners that allows users to discover, customize (such as fine-tuning where supported), and deploy models. Training entirely new models from scratch, benchmarking models in general, or managing model lineage and versions are different capabilities.
Question 18
A learning and development team wants to quickly create a new hire training video with a custom avatar and voiceover that matches their company’s branding and key messaging. They did not receive any money to spend on the production. What should they do?
A. Generate the video frames with Imagen.
B. Create a video with Google Vids.
C. Train a model with Vertex AI and produce a video.
D. Prompt the Gemini app to create a video.
Show Answer
Correct Answer: B
Explanation: Google Vids is designed to quickly create training and presentation videos using built-in AI features, including AI-assisted scripting, voiceovers, avatars, and video generation workflows, without requiring custom model training or production budgets. Imagen generates images rather than complete videos, Vertex AI custom model training is unnecessary and costly for this use case, and the Gemini app is not the primary tool for producing branded training videos.
Question 19
What is a key advantage of using Google’s custom-designed TPUs?
A. TPUs increase the storage capacity and data retrieval speeds within Google Cloud data centers.
B. TPUs are lightweight processors intended for deployment on edge devices.
C. TPUs are specialized AI processors that excel at parallel processing for machine learning workloads.
D. TPUs are primarily designed to improve the general processing speed of virtual machines in the cloud.
Show Answer
Correct Answer: C
Explanation: TPUs (Tensor Processing Units) are custom ASICs designed specifically to accelerate machine learning, particularly tensor and matrix operations. Their architecture is optimized for highly parallel computation, making them significantly more efficient for many AI workloads than general-purpose processors. They are not primarily for storage, edge-device deployment, or general VM acceleration.
Question 20
A global news company is using a large language model to automatically generate summaries of news articles for their website. The model’s summary of an international summit was accurate until it hallucinated by stating a detail that did not occur. How should the company overcome this hallucination?
A. Fine-tune the model on a larger dataset of news articles.
B. Use grounding to base the model output on the source articles.
C. Implement stricter safety settings to filter out potentially controversial topics.
D. Increase the temperature setting of the model to encourage more diverse outputs.
Show Answer
Correct Answer: B
Explanation: Grounding the model on the source articles constrains generation to the provided factual content, which is the standard approach to reducing hallucinations in summarization. Fine-tuning may improve performance but does not reliably eliminate hallucinations, stricter safety settings are unrelated to factual accuracy, and increasing temperature generally increases output variability rather than factual consistency.
$19
Get all 73 questions with detailed answers and explanations
Instant download HTML + PDF delivered the moment payment clears.
Secure Stripe checkout we never see or store your card details.
7-day refund if files are defective see our refund policy.