AI-900 Practice Question: Describe features of generative AI workloads on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Extracting and copying text from scanned images using OCR

  Why it's wrong here

  OCR is an extraction task: it detects and transcribes characters that already exist in an image, mapping pixel patterns to known glyphs. This is recognition, not generation, because the model never creates new content—it simply reproduces source text. Text generation, by contrast, produces original sequences of tokens from a prompt, such as writing a new paragraph or drafting an email.

• ✓

  Creating new coherent text from prompts for writing, code, summaries, and conversational AI

  Why this is correct

  Text generation is the core capability of large language models (LLMs) in Azure AI services: the model uses autoregressive token prediction to produce new, coherent text that was not present in the input. This enables writing assistance, code completion, summarization, and conversational AI, where the output is semantically novel and directly conditioned on the prompt. It is the only option that aligns with the definition of text generation as a generative rather than extractive or transformative workload.

• ✗

  Converting speech audio into a written transcript

  Why it's wrong here

  Converting speech into a written transcript is a speech-to-text (speech recognition) workload, not text generation. The system transduces an audio signal into a sequence of words that were spoken, so the output is bound to the existing acoustic content and contains no novel information. Azure AI Speech provides this capability separately from generative language models, which synthesize original text from a prompt rather than transcribe an input signal.

• ✗

  Formatting existing text by adding headings, bullets, and correct punctuation

  Why it's wrong here

  Adding headings, bullets, or punctuation to existing text is a formatting or editing operation that preserves the original content and only changes its structure or presentation. This can be implemented deterministically with rules or templates, and while an LLM might perform it as one type of edit, it is not the defining characteristic of text generation. Text generation requires producing new text content—words, sentences, or code—that did not exist before the prompt was issued.

Go deeper

Related to this question