Free GenAI Leader practice questions

The gen AI concept hierarchy: AI is the umbrella field, machine learning is a subset of AI that learns from data instead of following hand-written rules, and generative AI is a subset of ML that produces new content such as text, images, code, or video.

Option (a) states the nesting correctly: AI ⊃ ML ⊃ gen AI. Option (b) inverts the hierarchy, making ML a child of gen AI. Option (c) wrongly treats the fields as unrelated when gen AI literally depends on ML techniques like neural networks. Option (d) conflates a specific capability (generating content) with the entire AI discipline, which also covers non-generative tasks like prediction and classification.

1. Start with AI: any technique that lets machines perform tasks normally requiring human intelligence.
2. Narrow to ML: systems that learn patterns from data rather than being explicitly programmed.
3. Narrow again to generative AI: ML models (typically foundation models) that create new content rather than only predicting or classifying.

Exam tip: Remember the nesting AI > ML > generative AI; gen AI is the content-creating subset of machine learning.

Google's AI-first approach means AI has been embedded in its research priorities and products for many years, from Search to Translate to Photos, long before the current wave of generative AI.

Because Google has treated AI as central to its business for years — investing in research (such as the transformer architecture that underpins modern gen AI), building AI into consumer products used by billions, and designing custom AI hardware — its generative AI offerings inherit that depth. For a business leader, the practical takeaway is that Google's gen AI solutions are not a bolt-on: they are the product of long-term commitment and continued innovation. The other options mistake a strategic posture for a customer mandate, a product-line restriction, or a first-to-market claim, none of which describe what AI-first means.

1. Recall that "AI-first" describes Google's own long-term strategy of putting AI at the center of its research and products.
2. Connect that history to the present: years of AI investment translate into mature, cutting-edge gen AI solutions today.
3. Eliminate options that describe customer requirements, product discontinuations, or unrelated first-mover claims.

Exam tip: AI-first means Google's long-standing, company-wide commitment to AI is what powers the depth and pace of its gen AI innovation.

Hallucination is one of the most cited limitations of foundation models: the model produces confident, fluent text that is factually incorrect because it predicts likely words rather than verifying facts.

The model asserted a wrong launch date with confidence, which is the textbook definition of a hallucination. It is not a refusal (so not a data-dependency framing), not a prompting-technique issue (prompt chaining), and drift monitoring is an operational practice rather than a limitation. Recognizing hallucinations matters because remedies such as grounding and retrieval-augmented generation (RAG) exist specifically to anchor answers in verifiable sources.

1. Identify the behavior: fluent, confident output that contradicts documented fact.
2. Match that behavior to the limitation vocabulary: this is a hallucination.
3. Recall the recommended remedies: grounding and RAG tie responses to trusted data so fabricated facts are reduced.

Exam tip: Confident but factually wrong output is a hallucination — the primary remedy is grounding the model in verifiable data.

Gen AI solutions fall into recognizable types — text generation, image generation, code generation, and personalized user experiences — and the first step in choosing a solution is matching the business need to the right type.

The need here is producing written content (product descriptions) at scale, which is a text generation use case. A text generation solution built on a foundation model can draft consistent, on-brand copy for thousands of items far faster than manual writing. Image generation and code generation solve different output modalities (visuals and software, respectively), and workload monitoring is a security tool rather than a gen AI solution type, so none of the other options fit the stated need.

1. Identify what the business actually needs to produce — in this case, written product descriptions.
2. Map that output to a gen AI solution type: written content maps to text generation.
3. Rule out types that produce a different kind of output, such as images or code.
4. Confirm the choice against the business requirement of scaling to thousands of items, which text generation handles well.

Exam tip: Match the output the business needs (text, image, code, or personalization) to the corresponding gen AI solution type.

A foundation model is a large ML model pre-trained on broad, diverse data that serves as a reusable base ('foundation') for many downstream applications, from chat to summarization to code generation.

Option (a) captures the two defining traits: large-scale pre-training on vast data, and adaptability to many tasks. Option (b) is the opposite pattern — the narrow, single-task models that foundation models were designed to move beyond. Option (c) confuses a model with data infrastructure. Option (d) describes symbolic, rules-based AI, which predates and differs fundamentally from learned models.

1. Identify the defining trait: pre-trained once on vast, broad data.
2. Identify the second trait: adaptable to many downstream tasks rather than one.
3. Map to Google examples: Gemini (multimodal), Imagen (images), Veo (video) are all foundation models.

Exam tip: Foundation model = pre-trained at scale + adaptable to many tasks; Gemini, Imagen, and Veo are Google examples.