AI-900 Practice Question: Describe fundamental principles of machine learning 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.

• ✗

  Batch inference is more accurate; real-time is faster but less accurate

  Why it's wrong here

  This statement is false because the accuracy of an inference is determined by the model's training and the quality of its data, not by whether it is executed in real time or as a batch job. Both modes deploy the same model, and if the model parameters and inputs are identical, the output will be identical regardless of timing. Real-time may appear 'less accurate' only if you choose a smaller, faster model to meet latency targets, but that is a deployment trade-off, not an inherent property of the inference mode.

• ✓

  Real-time processes individual requests immediately; batch processes large datasets at scheduled intervals

  Why this is correct

  That is correct: real-time inference (also called online inference) responds to each individual request with a prediction immediately, typically via a REST API, enabling interactive applications like a chatbot or fraud detector. Batch inference, on the other hand, processes a large dataset asynchronously in a scheduled or otherwise triggered job, producing predictions for many records collected together, which suits periodic scoring like daily customer churn analyses. This distinction in latency and workload shape—not accuracy or hardware—is the primary reason you choose one pattern over the other.

• ✗

  Batch requires GPU compute; real-time uses CPU only

  Why it's wrong here

  This is incorrect because the choice of compute (CPU vs GPU) is independent of the inference mode. A batch job can run on GPUs to process huge volumes more quickly, while a real-time endpoint might use a CPU if the model is small and the latency still meets requirements, or a GPU if the model is a deep neural network needing low-latency responses. There is no rule that batch requires GPU or that real-time is CPU-only; the selection depends on model complexity, load, and cost constraints.

• ✗

  Real-time inference is only available in Azure; batch works on-premises too

  Why it's wrong here

  That is false because both real-time and batch inference can be deployed in Azure, on-premises, or at the edge. Azure Machine Learning offers managed online (real-time) and batch endpoints in the cloud, but you can also containerize models and run them on your own infrastructure, such as using Azure Arc or Azure IoT Edge for on-premises/edge real-time inferencing, and you can run scheduled batch scoring scripts locally or in a private Kubernetes cluster. The actual difference remains whether predictions need to be delivered immediately per request or can be computed on a bulk schedule, not the deployment location.

Go deeper

Related to this question