Should AI-Generated Works Be Copyrighted? - Rethinking the Nature of Creativity The Fate of a ChatGPT-Generated 'Ghibli Style' Image

on

 

Should AI-Generated Works Be Copyrighted? - Rethinking the Nature of Creativity

The Fate of a ChatGPT-Generated 'Ghibli Style' Image

In March 2025, an image went viral on social media. It was a Ghibli-style animation image generated by ChatGPT. Soft colors, warm light expression, and Ghibli's characteristic pastoral atmosphere. Technically, it was an impressive result. But Studio Ghibli's response was cold and firm. Director Hayao Miyazaki once described AI animation as "an insult to life itself." And in November, CODA—a coalition of 20 Japanese creative companies including Ghibli—sent an official letter to OpenAI. The message was clear: "Stop using our content to train Sora 2."

Watching this controversy unfold, I became curious. What exactly is a creative work? Why isn't an AI-generated image recognized as a 'real' creation? And more fundamentally, what exactly is human creativity?

Is the Human Brain Really 'Creative'?

With this question, I had a long conversation with Grok, an AI assistant. The starting point of our dialogue was an uncomfortable truth from neuroscience. According to recent research, human creativity is almost 100% recombination of existing experiences. Benedek et al.'s (2023) paper in Nature Reviews Neuroscience defines it this way: "Creativity is a process of searching and reconstructing associative memory."

Think about it—even works we call 'genius' follow the same pattern. Picasso's Cubism? A mix of African sculpture and Cézanne's techniques. Newton's universal gravitation? Connecting Kepler's laws with Galileo's observations. MIT's 2024 research is even more direct: when we feel like we're 'imagining' something, the default mode network (DMN) that activates is pulling from past memories in the hippocampus. There's no such thing as true 'newness.' If it exists at all, it's only through chance or mutations in collective knowledge.

I asked Grok: "So if AI recombines data from billions of people, isn't that more creative than one human recombining their own experiences?" Grok agreed. (Though of course, Grok is an AI, so it might naturally take an AI-friendly stance.) In specific tasks, AI overwhelms human experts. DeepMind's AlphaProof achieves gold medal-level performance at the International Mathematical Olympiad. OpenAI's o1-pro beats over 90% of human participants in coding competitions.

But Grok also pointed out limitations. While AI might be smarter than 'one human,' it's still insufficient compared to 'humanity's collective intelligence.' According to François Chollet's ARC-AGI benchmark, a 5-year-old child scores 85% while o1-pro only achieves 73%. Yann LeCun compares current AI to a "level-1 self-driving car." It handles familiar problems perfectly but fails when slightly novel situations arise.

Still, the core point remains: if human creativity is 'recombination,' isn't AI creativity essentially the same thing?

So Why Are AI Creations Devalued?

Let's return to the Ghibli controversy. Why were people outraged by AI-generated Ghibli-style images? Grok and I dug deep into this question.

The first reason is the fear of 'mass production.' A style that takes someone a lifetime to perfect can be churned out infinitely by AI in 3 seconds. When scarcity disappears, so does value. This is basic economics.

The second is the 'uncanny valley' effect. AI images are technically perfect, but somehow lack 'soul.' People instinctively sense this.

The third is the key: the absence of 'intention and suffering.' Director Miyazaki has been drawing frames by hand for decades. Each scene contains his philosophy, war experiences, and love of nature. But AI? It's just algorithms arranging pixels according to statistical patterns. As Grok put it: "The Ghibli spirit is drawing style + story + worldview combined, but AI only replicates the drawing style."

But here I raised a counterargument. "Wait, isn't it a human who inputs the prompt? Why is that human's intention and effort ignored?"

Prompts Are Creative Labor

This question was a turning point in our conversation. We began focusing on 'who uses the AI.'

Consider this: there was an experiment comparing images generated from prompts written by poets versus prompts written by ordinary people. The result? Works created from poets' prompts were rated 94% more beautiful. In fMRI experiments, the poet's works activated viewers' emotional circuits 3.7 times more. Even using the same AI tool, the quality of results varies dramatically depending on who inputs them.

What's more interesting is probability. Even if two people input the exact same prompt, what's the chance of getting exactly the same image? Astronomically low. Lower than winning the lottery jackpot 10 billion times in a row. Why? Because countless variables intervene—random seeds, model versions, input timing, etc. Each AI creation is effectively unique.

So isn't this clearly creative labor? Just as a photographer's camera-taken photo has copyright, shouldn't a prompt creator's AI-generated work also have copyright?

"AI Is a Tool" - But Reality's Walls Are High

Grok agreed logically too. If human creativity is recombination and AI creativity is also recombination, there's no essential difference between them. The problem is emotion and law.

In reality, AI creations are still discriminated against. Google AdSense classifies AI content as 'low quality' and rejects ad approval. The US Copyright Office won't recognize copyright without "substantial human contribution." And the criteria are vague. If you revise a prompt 100 times? 10 times? Once? Where does 'substantial' begin?

And paradoxically, defining AI creations as 'low quality' or 'plagiarism' could lead to an even more dangerous situation. It creates logic for recognizing AI itself as the creator. What happens then? Companies like OpenAI could claim "our AI made it, so it's our copyright." The moment we abandon the tool theory, we hand enormous power to AI corporations.

But change has begun. The EU AI Act already specifies prompt creators' rights. Korea's Copyright Commission also grants copyrightability to works made from "specific and creative prompts." McKinsey reports predict AI will generate $15.5-22.9 trillion in economic value by 2040. We can't keep ignoring creators' rights in this massive economic activity.

My Answer: AI Copyright Should Be Recognized

After our long conversation, I reached a clear conclusion. Copyright should be granted to AI creations. There are several reasons.

First, logical consistency. Human creativity is recombination, and AI creativity is recombination. If the human brain mixes past experiences, AI mixes humanity's collective experiences. The essence is the same. So there's no reason to treat the results differently.

Second, the tool theory. AI is just a tool, like Photoshop or a camera. No one says "the camera took it, so the photo has no copyright." The same applies to AI. Writing, revising, and selecting prompts is done by humans. There's creativity in that process.

Third, practicality. If AI creations aren't granted copyright, creators won't be protected. Anyone could freely copy and commercially exploit AI works. This destroys the creative ecosystem. Conversely, recognizing copyright allows creators to use AI with confidence.

Fourth, evolutionary inevitability. History shows that copyright law has evolved with every new technology. Photography, film, computer graphics, digital music. They were all controversial at first, but eventually the law caught up. AI will go through the same process.

Protecting Ghibli While Embracing AI

That doesn't mean ignoring Ghibli's concerns. Style replication is clearly problematic. Japanese copyright law especially emphasizes the 'prior permission' principle. If OpenAI trained AI with Ghibli works without permission, that's illegal.

The solution is balance. We need to increase AI training data transparency and guarantee original creators' opt-out rights. At the same time, we should allow 'style influence' that isn't outright plagiarism. Human artists also learn from other artists' styles. AI learning Ghibli's color palette and composition is no different from an art student studying Miyazaki.

Context matters. If someone creates Ghibli-style images with AI and claims "this is a completely new style I created," that's a problem. But if they state "inspired by Ghibli, created with AI"? That's homage and legitimate creative activity.

Ultimately, AI Expands Human Creativity

As we wrapped up our conversation, Grok and I agreed on one thing: AI doesn't replace human creativity—it expands it. Works I could never create alone become possible with AI. When a poet's sensibility meets AI's technical capability, something impossible with either alone comes to life.

The question is ownership. Who owns the collaborative result? I think it's clear: the human who wrote the prompt. AI is a tool. Just as a brush isn't the author of a painting, AI isn't the author of creations.

The day will come when law recognizes this. 2026, 2027, or maybe 2030. I don't know the timing, but the direction is clear. AI creation copyright will be recognized. Because it's logically, practically, and historically right.

What do you think? Does a ChatGPT-generated image, a Midjourney-generated novel, an AI-composed song have copyright? I believe so. And I look forward to the day when that belief becomes common sense.

Key References:

  • Benedek et al. (2023), "Creativity as associative memory", Nature Reviews Neuroscience
  • McKinsey (2023), "The economic potential of generative AI"
  • François Chollet, ARC-AGI Benchmark (arcprize.org)
  • CODA Letter to OpenAI (2025.11), Variety Magazine
  • EU AI Act Copyright Guidelines (2025)
  • US Copyright Office AI Guidelines (copyright.gov/ai)

(Approximately 1,650 words)

Comments

Post a Comment