AI-900 Practice Question: Describe Artificial Intelligence workloads and considerations

Answer analysis

Option-by-option breakdown

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

• ✗

  An AI model is software; an AI system includes the hardware it runs on

  Why it's wrong here

  Reducing the AI-system boundary to just the hardware is misleading because hardware is only one physical dependency and not the defining feature. An AI model is indeed a software artifact with learned parameters, but a responsible-AI view of a system also includes data provenance, feature engineering, deployment scripts, user interfaces, monitoring, and the humans who make decisions based on outputs. Hardware alone cannot account for harms that arise from biased training data, unclear UX, or misuse by operators.

• ✓

  An AI model is the prediction function; an AI system includes all surrounding pipelines, interfaces, and human processes

  Why this is correct

  An AI model is fundamentally a mathematical function that maps inputs to outputs—for instance, a neural network's weights and architecture. An AI system is the entire sociotechnical envelopment: the data pipelines that feed it, the application or API that exposes it, the human workflows that interpret its results, and the governance procedures that review its behavior. This system-level perspective matters for responsible AI because the same model can lead to very different real-world consequences depending on how it is integrated, monitored, and acted upon by people.

• ✗

  AI systems are more accurate than individual models because they combine multiple models

  Why it's wrong here

  The claim confuses an accuracy-enhancing technique with a definitional boundary. While ensembles can combine multiple models to improve prediction accuracy, an AI system does not require more than one model, nor is it defined by model count. A single model wrapped with input validation, output interpretation, and human oversight is already an AI system. The model-versus-system distinction is about scope of responsibility and impact, not about whether multiple models are stitched together to get better numbers.

• ✗

  An AI model runs offline; an AI system requires internet connectivity

  Why it's wrong here

  Connectivity is an infrastructure detail, not a conceptual boundary. An AI model may run fully offline on an edge device, and an AI system may also operate entirely within a private, disconnected network—neither situation changes the fact that the model is just the predictive component while the system includes surrounding workflows and interfaces. Conversely, a cloud-deployed model with an API and human-in-the-loop review is an AI system regardless of whether the internet is used. The real separation is between a raw prediction function and the broader sociotechnical context in which it operates.

Go deeper

Related to this question