Individual or Instrument - GenAI's Current Place in the Art and Intellectual Property Worlds

Class: PHIL-282
Author: Arthur I. Miller
Title: The Artist in the Machine: The World of AI-Powered Creativity (Cambridge, MA: MIT Press, 2020).

GenAI is Changing Creative Fields: Widespread tools like Midjourney and DALL-E are transforming how art is made and consumed.
A Double-Edged Sword:
- Pros: Makes art creation more accessible to those without traditional skills or technology; makes art appreciation more affordable.
- Cons: Devalues the labor of human artists, who now compete with computers.
The Big Question: Is GenAI an independent creative entity or just an instrument/tool used by humans?.
- This matters because it will shape how we value, regulate, and assign ownership to AI-created art.

Miller's Basic Definition: Creativity is "the production of new knowledge from already existing knowledge" through problem-solving.
Two Types of Creativity:
- Little-c creativity: Minor, day-to-day discoveries or combining familiar ideas in new ways.
- Big-C creativity: Transformational, rule-breaking innovation. This is the book's main focus.
Frameworks for Understanding Creativity:
- Graham Wallas's Four Stages: (1) Conscious thought (focusing on a problem), (2) Unconscious thought (letting it simmer), (3) Illumination ("aha!" moment), and (4) Verification (refining the idea).
- Margaret Boden's Three Forms: Combinatorial & Exploratory (like little-c) and Transformational (like Big-C).
- Miller's Criteria for Big-C Creativity: An entity must demonstrate intent, imagination, and unpredictability. GenAI currently struggles with unpredictability because it is designed to find existing patterns, not create truly novel ones.

Core of the Book: These sections analyze different GenAI models in visual art, music, and literature.
Example: GANs (Generative Adversarial Networks):
- How they work: Two dueling networks—a "generator" creates images and a "discriminator" judges them against a training dataset. The generator learns from feedback to create more realistic images.
- Notable GAN: The Painting Fool creates emotional art based on news articles it reads. Its creator, Simon Colton, believes we shouldn't compare human and computer art but respect GenAI creations for what they are.
Professional Consensus? There is none. Opinions fall into three camps:
1. Affirm: GenAI is creative, either like humans or in its own unique way.
2. Deny: GenAI is not creative. Some think it never will be (creativity is uniquely human); others think it might become creative in the future but is currently just a tool.
3. Defer: It's incomparable. The public must decide if GenAI art is "creative".
Miller's Final Take: For now, "AI is more like a child that needs to be painstakingly taught" before it can create on its own.

Lack of Ethical Discussion: The book is criticized for not addressing the ethical implications of GenAI, such as trained bias that can reinforce stereotypes.
The Ownership Problem: This is a major issue the book fails to explore in depth.
- Distributive Justice: The principle that creators should be fairly credited and compensated for their work.
- Training Data: Many GenAI models are trained on artists' work without permission (a distributive injustice), like the ImageNet database. This can be considered a violation of intellectual property rights.
- Legal Precedent: US courts have ruled that works made by GenAI cannot be copyrighted because they lack a human author (Thaler v. Perlmutter).

Beyond "Artist vs. Tool": We should focus on creating frameworks for ethical human-AI interaction in art.
Proposed Model: The Tribrid
- Instead of seeing the GenAI as a standalone creator or a simple tool, we should view the creative entity as a tribrid: (1) the Programmer, (2) the GenAI model, and (3) the User.
- Each part is essential: The programmer provides imagination (by building the model), the user provides intent (through prompts), and the GenAI provides execution.
Distributing Credit: Since both the programmer and user make creative contributions (imagination and intent), they should share rights and compensation for the final product, according to distributive justice. The GenAI, lacking true unpredictability, remains a tool in this model and cannot yet claim ownership.

Summary version

Main Question: Is GenAI a creative artist or just a tool for humans? This shapes how we value, own, and regulate its art.
Creativity Defined: Producing new knowledge from existing knowledge.
- Little-c: Everyday, minor creativity.
- Big-C (Focus of book): Transformational, rule-breaking innovation.
Miller's Criteria for "Big-C" Creativity: An entity must show intent, imagination, and unpredictability. GenAI currently fails at unpredictability because it's designed to find existing patterns in data, not invent truly new ones.

Book's Content: Analyzes GenAI models in visual art, music, and literature.
Example (GANs): A "generator" network creates images and a "discriminator" network judges them, providing feedback that helps the generator improve.
Professional Consensus: There is no consensus on whether GenAI is creative.
- Affirm: It is creative, either like humans or in its own way.
- Deny: It is not creative and is just a tool, though it might become creative in the future.
- Defer: It's incomparable; the public must decide.
Miller's Conclusion: For now, GenAI is like a child that needs to be "painstakingly taught" before it can create on its own.

Key Problems: The book is criticized for not deeply exploring major ethical issues, especially ownership and bias.
Distributive Justice: The idea that creators should be fairly credited and compensated for their work.
The Training Data Problem: GenAI models are often trained on artists' work without their permission, which is a violation of their intellectual property rights and a distributive injustice.
Copyright Law (US): The courts have ruled that work made by GenAI cannot be copyrighted because it lacks a human author (Thaler v. Perlmutter).

The Path Forward: Focus should be on frameworks for ethical human-AI interaction in art.
The Tribrid Model: View the creative entity not as a single artist or tool, but as an integrated tribrid:
1. The Programmer (Provides imagination by creating the model).
2. The GenAI Model (Provides execution).
3. The User (Provides intent through prompts).
Distributing Credit: Since both the programmer and user make creative contributions (imagination and intent), they should share the rights and compensation for the final artwork, in line with distributive justice.