Which task is a generative AI task? The short answer: writing a poem based on a given theme is a generative AI task, because it involves the AI creating brand-new content rather than analyzing, sorting, or labeling something that already exists.
This is a common question on AI foundations exams and quizzes, and the confusion usually comes from mixing up generative tasks with discriminative tasks — two very different jobs an AI model can do.
The Simple Test: Does It Create or Does It Judge?
A generative AI task produces new content: text, images, audio, video, or code that didn’t exist before the model created it.
A discriminative (or analytical) AI task does the opposite — it looks at existing data and returns a label, a score, a category, or a decision. It doesn’t create anything new; it classifies or evaluates what’s already there.
Ask yourself one question about any task: “Is the output something new, or is it a judgment about something that already exists?” If it’s new content, it’s generative. If it’s a label, rank, or answer pulled from fixed options, it’s not.
Examples of Generative AI Tasks
- Writing a poem based on a given theme
- Generating an original short story from a prompt
- Creating an image from a text description
- Composing music in a specific style
- Writing code to complete a described function
- Drafting an email or article from a set of bullet points
- Generating a synthetic voice reading a script
Each of these has the model producing something original that didn’t exist before the prompt was given.
Examples of Tasks That Are NOT Generative AI
These are common answer choices that trip people up on quizzes, because they involve AI — just not generative AI:
- Identifying the main topic of a news article — this is classification, not generation.
- Calculating a shipping cost or numerical value — this is computation, not content creation.
- Detecting spam in an email — this is a discriminative task; the model sorts input into “spam” or “not spam.”
- Recognizing a face in a photo — this identifies something that already exists rather than creating new content.
- Predicting whether a customer will churn — this returns a probability or label, not new content.
All of these are legitimate AI tasks, and some even use similar underlying model architectures. But none of them produce new content, which is the defining feature of generative AI.
Why This Question Comes Up So Often in AI Courses
Courses like Oracle’s OCI AI Foundations, Google’s AI fundamentals training, and general machine learning introductions all draw a clear line between generative and discriminative AI early on, because it’s foundational to understanding how modern AI tools are categorized.
Generative AI tasks carry a different kind of risk than discriminative ones, too. A misclassified email sits quietly in the wrong folder. A generative model that confidently produces an incorrect answer, on the other hand, can hand a user a well-written but wrong response — which is part of why understanding this distinction matters beyond just passing an exam. The same applies to other certification-style questions — like knowing which criteria cannot be used to create a custom segment in Google Analytics — where the exam tests a specific technical distinction rather than general knowledge.
Quick Reference Table
| Task | Generative? | Why |
|---|---|---|
| Writing a poem on a theme | Yes | Creates new text |
| Generating an image from a prompt | Yes | Creates new visual content |
| Detecting spam email | No | Classifies existing input |
| Identifying a news article’s topic | No | Labels existing content |
| Composing original music | Yes | Creates new audio content |
| Calculating a numeric score | No | Computation, not creation |
| Writing code from a description | Yes | Creates new code |
| Recognizing an object in an image | No | Identifies existing content |
Frequently Asked Questions
Which task is a generative AI task?
Writing a poem based on a given theme is the classic example, since the AI produces original text that didn’t exist before the prompt.
What makes a task generative versus discriminative?
A generative task creates new content (text, images, audio, code). A discriminative task analyzes existing content and returns a label, score, or classification.
Is detecting spam a generative AI task?
No. Spam detection is a discriminative task — the model classifies an existing email as spam or not spam rather than creating new content.
Is image recognition a generative AI task?
No. Recognizing what’s in an image is discriminative, since the model is identifying something that already exists rather than generating new content.
Are chatbots like ChatGPT generative AI?
Yes. Text-based conversational AI tools generate new responses to prompts, making text generation one of the most common generative AI applications.
Final Thoughts
Which task is a generative AI task? Any task where the AI produces new content — writing, images, music, code, or audio — rather than analyzing or labeling existing data. Keeping the “create vs. judge” test in mind makes it easy to sort generative tasks from discriminative ones, whether you’re studying for a certification exam or just trying to understand how different AI tools actually work.
Read More: https://flyingvoices.org/how-to-use-freedoor2-4-6-8-on-mac/