"AI art" isn't made by AI, it's built with theft, and it has serious creative limits

Introduction

"Air," does the subject on Vogue right now. At least in the circles that I run, it is a complex subject. The technology itself is really just an evolution of the same machine learning that shows you banner ads on websites or decides which YouTube videos to recommend. But because it's called AI, artificial intelligence, people get suckered into all kinds of very bad arguments about the nature of creativity and what really counts as plagiarism.

Human Minds and Binary Code

The TL;DR though is this; number one, human minds do not run on binary code. And number two, these machine learning models are not AI; they are machine learning models. They are built using untold gigabytes of training data consisting of photos and artwork that was used without permission. And however you want to slice that, that is a form of theft.

Creative Limitations

Because they are binary machine learning, they inherit the biases and limitations of the data that they're trained on. That puts hard limits on how creative it is even possible to be using these tools, and it introduces a massive trend toward homogenization and the lowest common denominator.

Historical Precedent

There's also a much longer historical precedent for what these tools represent. AI art is a form of automation, and you can find a much longer discussion of what that means and what the consequences are over on my main channel.

Keyword

Machine Learning
AI
Artificial Intelligence
Plagiarism
Training Data
Creative Limitations
Automation

FAQ

Q: What is AI art?
A: AI art refers to artwork generated using machine learning models, often mislabeled as "artificial intelligence."

Q: How are machine learning models built?
A: They are built using extensive amounts of training data, which often include photos and artwork used without permission.

Q: Why is using training data without permission problematic?
A: Using this data without authorization is a form of theft and raises ethical concerns regarding exploitation of original creators' work.

Q: Do machine learning models possess true creativity?
A: No, their "creativity" is limited by the biases and constraints of their training data, leading to homogenized and lowest common denominator outputs.

Q: What historical context exists for AI art as a form of automation?
A: AI art is part of a broader history of automation, a topic discussed extensively elsewhere for its implications and consequences.