Intro to Generative AI and LLMs
Free Oracle Cloud Infrastructure 2025 AI Foundations Associate practice — 6 questions on Intro to Generative AI and LLMs, with explanations. No sign-up. Full 12-question mixed test →
A retail company has two AI initiatives: (1) a model that predicts whether a customer will churn based on historical account data, and (2) a model that writes personalized marketing emails from a short brief. Which statement correctly classifies these two models?
Churn prediction assigns an existing label (churn/no-churn) to input data, which is the defining behavior of a discriminative model. Writing a new marketing email from a brief produces novel content, which is the defining behavior of a generative model.
In the context of large language models, what does a token embedding represent?
Embeddings map tokens into dense numerical vectors positioned in a continuous space such that semantically similar tokens end up close together, which is what enables the model to reason about meaning and similarity.
You are configuring an OCI Generative AI chat model to draft highly consistent, deterministic legal document summaries where creative variation is undesirable. Which decoding parameter adjustment BEST achieves this goal?
A low temperature makes token sampling favor the most probable next token, producing more deterministic, consistent output — exactly what is needed for repeatable legal summaries.
A healthcare provider wants its chatbot to answer questions using the latest internal clinical guidelines, which are updated weekly. The provider does not want to retrain the model each time content changes and needs verifiable source citations. Which approach BEST meets these requirements?
RAG retrieves the most current guideline content from an external knowledge store at query time and grounds the response in those retrieved passages, avoiding retraining and enabling traceable source citations.
A company wants to build an AI assistant that can autonomously check a customer's order status by calling an internal REST API, decide whether to escalate the issue based on the API response, and then compose a reply email — all without a human specifying each step. Which capability BEST describes what is needed?
AI agents extend an LLM with the ability to call external tools (like an API), reason over intermediate results, plan next steps, and take autonomous multi-step actions — exactly matching the described escalation-and-reply workflow.
An LLM-powered chatbot deployed for a law firm frequently fabricates case citations that sound plausible but do not exist. Which action would MOST effectively reduce this hallucination behavior while preserving the chatbot's ability to answer legal questions?
Hallucinated citations occur because the model generates plausible-sounding text without factual grounding. RAG anchors responses in a verified external source, and instructing the model to cite only retrieved content directly addresses the root cause.
Ready for the real thing?
The full course has two full-length practice tests, video lessons for every exam domain, hands-on labs and detailed answer explanations.
Start my full course on Udemy →