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.

• ✗

  Transparency is easy for all AI models because they use simple mathematical formulas

  Why it's wrong here

  This is incorrect because while classical models like linear regression or short decision trees rely on transparent formulas, modern AI leverages deep neural networks with millions of parameters arranged in layers of nonlinear transformations, none of which have straightforward real-world interpretations. The sheer scale and non-linearity mean that local approximations are required even for simple questions about model behavior, and "simple formulas" alone cannot explain the emergent, high-dimensional decision surfaces that these systems construct. Thus, interpretability is a distinct and difficult engineering problem, not an inherent property of all mathematical models.

• ✓

  Deep learning models are 'black boxes' — high performance but difficult to explain because of millions of interacting parameters

  Why this is correct

  This is correct. Deep learning architectures like convolutional and recurrent networks learn millions of connection weights through non-linear activation functions, and the resulting feature representations are distributed and hierarchical, with no single parameter responsible for a prediction. Even with weight matrices and activations exposed, the internal logic is not human-readable, which makes it challenging to provide meaningful explanations for individual predictions, a central obstacle to achieving full transparency.

• ✗

  Transparency only matters for AI systems used in consumer products

  Why it's wrong here

  This statement is incorrect because AI transparency is a critical requirement across all deployment domains, including healthcare, finance, criminal justice, and public administration, where decisions have significant societal and legal consequences. In consumer products, lack of transparency might lead to distrust, but in high-stakes sectors, opaque models can cause irreversible harm, making regulatory and ethical transparency obligations even more stringent. Transparency is not a feature that can be toggled based on audience; it is a fundamental principle of responsible AI governance.

• ✗

  Transparency is fully solved by showing the training data to stakeholders

  Why it's wrong here

  This is incorrect because transparency in AI requires more than access to the data; it demands a clear understanding of the learned relationship between inputs and outputs, including feature importance, decision boundaries, and potential biases. Merely sharing training data reveals nothing about how the model weighs conflicting evidence or why it generalized to new cases; it could even expose the organization to privacy and security risks. Full transparency encompasses model cards, documentation, algorithmic impact assessments, and interpretable machine-learning techniques, not just dataset access.

Go deeper

Related to this question