Quiz – Generative AI Limits

Question 1 of 5

1. Question

Whose 2023 research documented how language models generate plausible-sounding but fabricated information?

Huang and colleagues

Correct Incorrect Correct answer

Bolukbasi and colleagues

Correct Incorrect Correct answer

Brown and colleagues

Correct Incorrect Correct answer

Kingma and Welling

Correct Incorrect Correct answer

Question 2 of 5

2. Question

Whose landmark 2016 research revealed how AI systems inherit and amplify biases from training datasets?

Bolukbasi

Correct Incorrect Correct answer

Huang

Correct Incorrect Correct answer

Devlin

Correct Incorrect Correct answer

Mikolov

Correct Incorrect Correct answer

Question 3 of 5

3. Question

How does the capsule describe a hallucination?

Confident false statements presented as facts

Correct Incorrect Correct answer

A verified citation to a real research paper

Correct Incorrect Correct answer

A refusal to answer a difficult question

Correct Incorrect Correct answer

A grammatical error in the output text

Correct Incorrect Correct answer

Question 4 of 5

4. Question

What effect does training data bias have on model outputs?

It perpetuates stereotypes and unfair associations

Correct Incorrect Correct answer

It guarantees fairer decisions than humans make

Correct Incorrect Correct answer

It removes all errors from the training data

Correct Incorrect Correct answer

It has no measurable impact on the outputs

Correct Incorrect Correct answer

Question 5 of 5

5. Question

Besides hallucinations and bias, which limitation restricts practical deployment of generative AI?

High computational costs and an absence of genuine logical reasoning

Correct Incorrect Correct answer

An inability to generate any text or images

Correct Incorrect Correct answer

A requirement to always cite real sources correctly

Correct Incorrect Correct answer

Complete transparency of every decision process

Correct Incorrect Correct answer