AI-901 - Identify AI concepts and capabilities (42% of the exam) - Section 1.2
Identify AI model components and configurations, including how generative models work, selecting a model by capability, and deployment and configuration parameters.
How generative AI models work at a conceptual level (tokens, next-token prediction, training versus inference); choosing a model by its stated capabilities, such as a language model versus a multimodal model versus an embedding or image-generation model; and the deployment and configuration parameters that shape a deployed model's behaviour, including temperature and maximum tokens. Model selection by capability and parameter configuration are genuine single-best-answer topics.
generative AI modelstokensmodel selection by capabilitymultimodal modelembedding modelmodel deployment optionstemperatureconfiguration parameters
Practice question for this objective
lock_openFree sampleIdentify AI concepts and capabilitiesmedium
During a briefing on how generative models produce output, an engineer wants to explain what happens at inference time when a deployed language model answers a prompt. Which statement best describes how the model generates its reply?
AIt retrains its own internal weights on the user's prompt before it returns any generated text.
BIt searches a fixed database of stored answers and returns the closest matching record it finds.
CIt predicts the next token repeatedly, each time using the prompt and the tokens it has already produced.check_circle Correct
DIt converts the prompt into one numeric vector and then returns that vector as its final reply.
At inference a generative language model produces its reply by predicting one token at a time, conditioned on the prompt and the tokens already generated. Inference is autoregressive: the model repeatedly predicts the next token from the prompt and its own prior output, so it neither retrains, nor looks up stored answers, nor simply emits a vector.
Why A is wrong: This is tempting because prompts do influence the output, but inference does not change the model's weights; training and inference are separate, so no retraining happens per prompt.
Why B is wrong: This sounds like retrieval, which makes it plausible, but a generative model composes new text by predicting tokens rather than looking up a stored answer, so the description is wrong.
Why C is correct: Correct: at inference the model generates one token at a time, feeding the prompt plus its own earlier tokens back in to predict each following token until the reply is complete.
Why D is wrong: Producing a vector is what an embedding model does; it is tempting because vectors are involved internally, but a language model returns generated text, not a raw vector.
Exam traps in Identify AI concepts and capabilities
Answers that look right on this material and are not. Each one is a distractor from a different question in the AI-901 bank for this domain.
The maximum response tokens, which sets how long a reply may be
errorWhy it is wrong: Raising the maximum response tokens lets replies run longer, which is easy to confuse with variety, but length is not the same as randomness and it will not change how predictable the wording is.
Temperature, which governs how random or varied the wording of the reply becomes.
errorWhy it is wrong: Temperature is a genuine generation setting, so it is tempting, but it shapes the style and randomness of the reply rather than capping its length.
It searches a fixed database of stored answers and returns the closest saved reply
errorWhy it is wrong: Retrieval systems do look up stored text, which makes this familiar, but a generative model composes new text token by token rather than returning a pre-written stored answer.