Question: Explain why end-to-end machine translation is an example of deep learning’s move toward rich-output learning, and contrast it with tasks that predict a single simple value.

Sample answer: End-to-end translation is a rich-output task because the model learns to map one structured input, such as a sentence in a source language, directly to another structured output sentence in a target language. Training uses paired examples of source text and target text. This illustrates that deep learning can produce complex outputs like full sentences, not just one number or one class label.

Key points:

• Directly maps source-language text to target-language text
• Trains on paired input-output examples
• Produces structured outputs such as sentences
• Contrasts with predicting a single scalar or label

Rubric: The student should clearly state that translation uses labeled input-output pairs and maps one sequence to another, and should contrast this kind of rich structured output with simpler outputs such as a single number or class.