The recent wave of images generated in the ‘Ghibli style,’ fueled by OpenAI’s new feature built on the GPT-4o model, has reignited tension between technological progress and the copyright of human creators. The saga raises fundamental questions: can an artistic ‘style’ be legally protected? And does training AI systems on copyrighted works, without creators’ consent, amount to fair use or copyright infringement?
Before Passover, social media feeds filled up with images created “in the style” of the renowned Japanese animation studio Ghibli, using the popular GPT-4o model’s new feature. What began as entertainment turned into a trend that swept up millions of users worldwide, who created images “in the style” of Studio Ghibli – sparking a heated debate about copyright, the boundaries between copying and inspiration, and whether training artificial intelligence on protected works constitutes fair use or copyright infringement.
As is well known, Hayao Miyazaki, the 84-year-old Japanese animator, is among the world’s most celebrated animators, known for works such as “Spirited Away,” “Ponyo,” “Princess Mononoke,” and “My Neighbor Totoro.” In Israel, he is also known as one of the creators of the television series “Heidi,” remembered by children of the 1970s and 1980s as “Marco.” Miyazaki’s style is unique and highly inspirational, employing small details while creating a complex and captivating fantasy world.
It is interesting to note that Miyazaki’s views on artificial intelligence, as expressed nearly a decade ago, are unfavorable – he even called AI-generated animation “an insult to life itself,” emphasizing that artificial intelligence is incapable of understanding genuine human emotions.
His attitude resembles that of the well-known singer Nick Cave, when a fan sent him a song created using artificial intelligence. Cave’s reply was:
“…what ChatGPT is, in this instance, is replication as travesty. ChatGPT may be able to write a speech or an essay or a sermon or an obituary, but it cannot create a genuine song… Songs arise out of suffering, by which I mean they are predicated upon the complex, internal human struggle of creation, and as far as I know, algorithms don’t feel. Data doesn’t suffer… ChatGPT’s melancholy role is that it is destined to imitate and can never have an authentic human experience, no matter how devalued and inconsequential the human experience may in time become.
What makes a great song great is not its close resemblance to a recognizable work. Writing a good song is not mimicry, or replication, or pastiche, it is the opposite. It is an act of self-murdering that destroys all one has strived to produce in the past…. This is what we, the mortal humans, can offer, that AI can only mimic, the transcendent journey of the artist that forever grapples with his or her own shortcomings. This is where human genius resides, deeply embedded within, and reaching beyond those limitations….
Thank you for the song, but with all the love and respect in the world, this song is bullshit, a grotesque mockery of what it is to be human, and, well, I don’t much like it.”
This anti-AI stance of Miyazaki’s amplifies the ethical questions surrounding the use of his works to train artificial intelligence and to generate content in his distinctive style.
From a copyright law perspective, the law protects individual works, while a general style or artistic movement is not protected by copyright. For example, there is no protection for cubism, surrealism, or musical works in the bossa nova or chill-out style.
However, this issue raises for discussion the question of using protected works to train models. Training an AI system requires extensive datasets relevant to the desired output. In many cases, this content includes copyrighted material – including images, texts, works, and musical recordings – used to train the model.
Copyright law generally restricts the right to use content created by another person, raising the question of creators’ or rights-holders’ consent to the use of their works for machine-learning purposes. Against these restrictions, proponents of such use believe that free access to information is necessary for the technology, which learns in a manner similar to humans, and that this will benefit society as a whole and contribute to the advancement of progress. The issue raises complex questions regarding the proper balance between these competing considerations.
Recently, a number of lawsuits have been filed against various companies, many of them seeking class-action status, over the use of protected content – including images, code, literary works, and musical works – for the purpose of training AI models.
The legal rulings in these important and intriguing lawsuits will have far-reaching implications on a question of enormous economic significance, and will contribute to clarifying the issue and shaping the rules of what is permitted and prohibited regarding the use of works for model training purposes.
When we look at the outputs of artificial intelligence, we do not dwell on their sources, their owners, or the extensive work that went into producing, collecting, sorting, or labeling them. Today, in many cases, we are already talking about billions or even trillions of parameters, and protected works (including images, music, etc.) harvested from the internet without permission. Without these images, OpenAI would not have had software that it now commercializes for paid use at all.
The focus on the role of the technology and on AI outputs obscures the discussion related to the substantial human contribution required to create it. This contributes to the establishment of techno-utopian narratives around the products, as if they were created out of nothing.
However, as scholar Kate Crawford noted: once we stop recognizing information as having value, ignore its sources and contexts, and treat it only in terms of its scale, suddenly its uniqueness is lost and it becomes merely “infrastructure.” Technologies that replicate human creativity flatten, in a sense, the human creative process, so that the content generated by AI models renders, in a sense, the human contribution invisible.
It is doubtful that artificial intelligence will replace the human creative process, but it is already changing the creative process and shining a spotlight on how the creative process based on traditional tools differs from the creative process using generative AI tools. It appears that we are approaching new frontiers in copyright law, and legislators, courts, and policymakers will be required to grapple with these fascinating issues at the interface between copyright law and the world of artificial intelligence.
Another related issue is the environmental implications of using artificial intelligence. According to a report on research conducted at the University of California, no less than 216 million liters of water were consumed within just a few days during the peak of the image-creation trend, water required to cool the massive computing centers powering AI systems. There is no doubt that environmental impacts and the issue of energy will become an increasingly central topic as the technology develops.
