Generative AI Leader Fundamentals of Generative AI Practice Question
Which TWO statements are true about generative AI models?
⚠ Common exam trap
Google Cloud often tests the misconception that generative AI models are deterministic and always produce the same output for the same input, when in fact they are probabilistic by design, especially at non-zero temperature settings.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
They are typically pre-trained on large datasets.
Option A is correct because generative AI models such as large language models and diffusion models undergo a pre-training phase on massive, broad datasets (e.g., web text, images) before any fine-tuning or alignment, which is what gives them general-purpose capabilities. Option D is correct because these models learn the underlying probability distribution of the training data and can sample from it to produce novel outputs—new sentences, images, or code—that were not present verbatim in the training set. Option B is incorrect because generative models are inherently stochastic; sampling steps such as temperature, top-k, or top-p introduce randomness, so they are not deterministic by design. Option C is incorrect for the same reason: with non-zero temperature or random sampling, the same input can yield different outputs across runs. Option E is incorrect because training data is essential—without a corpus to learn from, the model would have no parameters or distribution to generate from.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
They are typically pre-trained on large datasets.
Why this is correct
Pre-training on large datasets is the defining characteristic of generative models: self-supervised learning over billions of tokens or images builds the broad statistical patterns later fine-tuned for specific tasks. This satisfies the stem's requirement for a true statement about how such models are typically built.
- ✗
They are deterministic by design.
Why it's wrong here
Generative models sample from learned probability distributions, so identical prompts can yield different outputs; temperature and top-p control this randomness. Determinism describes rule-based or fixed-weight inference pipelines, not generative architectures. This statement is tempting because deterministic outputs are desirable for reproducibility, but that requires setting temperature to zero, not a design property.
- ✗
They always produce the same output for the same input.
Why it's wrong here
Sampling-based decoding means the same prompt can produce different completions across calls; only greedy decoding with temperature zero approaches repeatability, and even then floating-point and batching effects can vary output. This tempts because deterministic behaviour is expected from conventional software, but generative models are probabilistic by construction.
- ✓
They can generate new content not seen in training.
Why this is correct
Generative models produce novel outputs by sampling from learned probability distributions rather than retrieving stored examples, so responses are synthesised combinations not present verbatim in training data. This directly satisfies the stem's requirement for a true statement about generative capability.
- ✗
They require no data for training.
Why it's wrong here
Generative models are trained on large corpora and learn parameters from that data; without training data they cannot produce meaningful output. Pretraining, fine-tuning and RLHF all consume data. The statement tempts because foundation models require no additional data from the deploying organisation, but that is zero-shot inference, not zero training data.
Go deeper
Related to this question
About these practice questions
One of 1,008 original Generative AI Leader practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This Generative AI Leader practice question is part of Courseiva's free Google Cloud certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the Generative AI Leader exam.