3.3. Common AI Workloads
💡 First Principle: "AI workload" just means a category of task AI performs, defined by its input and output. The exam expects you to recognize, from a scenario, which workload is involved — because the workload determines which Azure capability you'd reach for. Learn to classify by asking "what goes in, what comes out?"
Why care? A large block of the concepts domain is scenario-to-workload matching: "extract the total from an invoice," "tell whether this review is positive," "convert this recording to text," "describe what's in this photo." Each maps to a distinct workload. The mental model is a lookup table keyed on input/output, which the subsections below build out.
⚠️ Common Misconception: "Computer vision and image generation are the same workload because both involve images." They're opposites in direction. Computer vision takes an image in and produces information out (labels, detected objects, extracted text). Image generation takes a text prompt in and produces an image out. Different inputs, outputs, and models.
