Quiz: AI and Creativity

Test your understanding of Chapter 11's key concepts. Try to answer each question before looking at the explanation.

Question 1. What do diffusion models, GANs, and autoregressive models all have in common?

  • (A) They all use the same neural network architecture
  • (B) They all learn from massive datasets of human-created content
  • (C) They all generate text as their primary output
  • (D) They all require real-time internet access to generate content

Answer: (B). Despite using fundamentally different technical approaches, all three families of generative AI learn patterns from large datasets of human-created content — text, images, music, or video. This shared characteristic is central to debates about training data ethics, copyright, and what it means for AI to "create."

Question 2. A diffusion model generates an image by:

  • (A) Copying and pasting parts of images from its training data
  • (B) Learning to progressively remove noise from random static, guided by a text prompt
  • (C) Drawing the image pixel by pixel from left to right, like reading a page
  • (D) Having two competing networks — one that generates and one that evaluates

Answer: (B). Diffusion models work by learning the reverse of a noise-addition process. During training, they learn what images "should look like" by seeing how noise is added to real images and learning to reverse it. During generation, they start with random noise and progressively denoise it into a coherent image, guided by the text prompt. Option D describes GANs, not diffusion models.

Question 3. Margaret Boden's framework identifies three types of creativity. Which type would be hardest for current AI to demonstrate?

  • (A) Exploratory creativity (finding new possibilities within an established style)
  • (B) Combinational creativity (connecting ideas from different domains)
  • (C) Transformational creativity (fundamentally changing the rules of a domain)
  • (D) All three types are equally easy for AI

Answer: (C). Transformational creativity involves breaking existing rules and inventing entirely new frameworks — like Picasso inventing Cubism or the Beatles creating a new genre. Current AI systems learn from existing patterns and recombine them; they don't independently conceive of fundamentally new artistic paradigms. Exploratory and combinational creativity are more achievable because they work within or across existing patterns.

Question 4. The "training data provenance problem" refers to:

  • (A) The difficulty of creating enough training data for generative AI models
  • (B) The unresolved questions about consent, compensation, and rights related to the human-created content used to train generative AI
  • (C) The technical challenge of storing large datasets
  • (D) The problem of AI models forgetting their training data over time

Answer: (B). The training data provenance problem is about the ethical and legal questions surrounding the use of human-created content (art, writing, music, photographs) to train generative AI models, often without the creators' knowledge, consent, or compensation. It's at the center of major lawsuits and policy debates.

Question 5. According to the chapter, the U.S. Copyright Office's current position on purely AI-generated works (without meaningful human creative input) is:

  • (A) They can be copyrighted by the person who wrote the prompt
  • (B) They can be copyrighted by the AI company
  • (C) They cannot be copyrighted — they belong to the public domain
  • (D) Copyright law doesn't apply to digital content

Answer: (C). The U.S. Copyright Office has ruled that works require "human authorship" for copyright protection. Purely AI-generated content without meaningful human creative input cannot receive copyright protection. However, works where a human exercised significant creative judgment (in selection, arrangement, or modification of AI outputs) may qualify for partial protection. The exact threshold is evolving.

Question 6. "Curatorial creativity" refers to:

  • (A) The creativity of museum curators specifically
  • (B) The creative act of selecting, combining, contextualizing, and giving meaning to AI-generated options
  • (C) The AI's ability to curate content for users
  • (D) The process of organizing training data before model training

Answer: (B). Curatorial creativity is the concept that in a world where AI can generate unlimited raw creative material, the distinctively human creative contribution shifts toward selection, combination, contextualization, and meaning-making. Just as a museum curator's creative act lies in choosing and arranging works rather than painting them, a human working with AI exercises creativity through judgment about what to use, how, and why.

Question 7. A freelance illustrator says AI image generators are "stealing" from artists. An AI company says training on public images is "fair use, like humans learning from art." What concept from the chapter is most relevant to evaluating these competing claims?

  • (A) The authorship gradient
  • (B) Curatorial creativity
  • (C) The training data provenance problem
  • (D) Diffusion model architecture

Answer: (C). The training data provenance problem directly addresses the tension between creators' rights over their work and AI companies' use of that work for training. Both the illustrator's and the company's arguments engage with this problem — the illustrator emphasizes consent and compensation, while the company emphasizes transformative use. The chapter acknowledges that both arguments have validity, which is what makes this genuinely hard.

Question 8. What is the "authorship gradient"?

  • (A) The technical quality scale from bad AI art to good AI art
  • (B) The spectrum from fully AI-generated content to fully human-created content, along which any AI-assisted work can be located
  • (C) The process by which AI models learn to generate increasingly realistic content
  • (D) The career path from amateur to professional artist

Answer: (B). The authorship gradient recognizes that AI-assisted creation isn't binary — it's a spectrum. A one-word prompt producing an AI image sits near one end; a fully hand-painted artwork sits near the other; and the vast middle ground includes various degrees of human-AI collaboration. Locating a specific work on this gradient is more useful than trying to draw a sharp line between "human-authored" and "AI-authored."

Question 9. The chapter describes generative AI's impact on creative industries as involving both "displacement" and "democratization." Which statement best captures the relationship between these two effects?

  • (A) Displacement is happening now, but democratization will happen later
  • (B) They are different effects of the same technology, occurring simultaneously, often in tension with each other
  • (C) Democratization is just a marketing term — only displacement is real
  • (D) They affect the same people in the same way

Answer: (B). The chapter emphasizes that displacement (creative professionals losing work) and democratization (non-professionals gaining creative tools) are simultaneous effects of the same technology. They're in tension because the same tool that empowers a small business owner to create their own graphics displaces the freelance designer who previously did that work. The people being displaced and the people being empowered are typically different people.

Question 10. When Priya used an AI tool to brainstorm, outline, suggest transitions, check facts (some of which were wrong), and provide feedback on her essay — but wrote the core arguments and research herself — this example primarily illustrates:

  • (A) Academic dishonesty that should be punished
  • (B) The complex middle ground of AI-assisted creation, where authorship is distributed along a gradient
  • (C) That AI is better at writing essays than humans
  • (D) That AI tools should be banned in education

Answer: (B). Priya's experience illustrates the authorship gradient in practice. She used AI as a tool for specific tasks (brainstorming, suggesting structure, providing feedback) while retaining creative and intellectual agency for the core work (thesis development, argumentation, research verification). The example also illustrates the lack of consensus — different professors had different policies about what constitutes acceptable AI use. This ambiguity is characteristic of the current moment.

Scoring Guide

  • 9–10 correct: Excellent command of generative AI concepts, creativity frameworks, and copyright issues. You can navigate these debates with nuance and precision.
  • 7–8 correct: Strong understanding with a few gaps. Review the areas where you missed questions — particularly the distinction between different generative approaches and the training data provenance problem.
  • 5–6 correct: Developing understanding. Re-read sections 11.2 (how AI creates) and 11.4 (authorship and copyright) carefully.
  • Below 5: Return to the chapter and re-read it, paying special attention to the key terms and the summary. The concepts in this chapter build on Chapter 5's discussion of LLMs, so reviewing that chapter may also help.