When OpenAI released a new image generation model, many users quickly realized they could transform their photos into whatever art style they wished. One style in particular gained popularity: Studio Ghibli’s animation style. Even OpenAI founder Sam Altman changed his X icon to a “Ghiblified” image of himself.
Studio Ghibli films are cult classic films that bring joy, sorrow and reflection to all ages. From “My Neighbour Totoro“, “Castle In The Sky” and “Spirited Away“, all carries their own charm and soul that the studio lovingly brings to the screen through it’s simple yet detailed art style. With the movie’s popularity, it’s no wonder people wish to image themselves in the whimsical worlds presented on screen. Enter ChatGPT and it’s ability to generate or transform almost anything you can image. Simply provide an image with a short description and the AI will reimagine your image as if it was drawn in the iconic Ghibli style.
Since ChatGPT isn’t currently available in Hong Kong, many users have shared what other platforms are available to transform pictures. This has gone viral on social media during late March to early April, most people hopping on to reimagine themselves in this art style. Aside from humans, people have also used this to reimagine locations such as tourist spots in HK. Some use this to convey a message, such as the white house posting an image of an arrest that garnered major backlash. Others use this to remember important people and moments in their life.
While this may seem like a lighthearted and fun trend on social media, the involvement of generative AI and a specific art style has raised questions about the ethics of this phenomenon. How does AI make “art” and what are the ethics concerned behind it?
How does AI make “art”?
Generative AI is a neutral network that draws on patterns to generate something new. The user inputs a prompt to the GenAI, which is the description of what they want the GenAI to generate. In this case, the input also includes the image the user wants to transform. Once you input your prompt, the GenAI will pull information from their trained data base for what best fits the description. Then the GenAI will use the information gathered to generate image options it thinks best suits the prompt. For example, type “Three eiffel towers in desert with river; photorealism” and you’ll get this uncanny image that matches the prompt.
There are many GenAI models that can generate images now, including Adobe Firefly. But, where do these models get their training data?
Copyright concerns
GenAI models use large databases of existing information to train their machine to give users the best output. OpenAI states that their products are developed using publicly available information on the internet, information with third parties access, and information that users or human trainers and researchers provide or generate. The trend of generating Ghibli-styled images has raised concerns on whether this is considered copyright infringement or not.
Masato Imai, member of the House of Representatives in Japan questioned the Ministry of Education, Culture, Sports, Science and Technology if generating Ghibli styled AI images was copyright infringement and what did and did not constitute fair usage under current law. Their response states that according to law, creative expressions of thought and feelings in the form of art, music and literature are protected under copyright law. The more style or creativity alone does not fall under the law’s regulatory area. Each case has to be individually evaluated. However, if the art shows “high similarity” and clearly traces from the original, it may infringe on copyright.
Ethics of AI art
Many artists have been outspoken against AI art, including Hayao Miyazaki himself, the animator behind Ghibli films. While he hasn’t said anything about the current Ghibli AI trend, he had a very strong reaction to AI animations back in 2016, calling it “gross”. His son, Goro Miyazaki says that AI may one day replace humans in animation, but the heart and soul put in by his father can never be replaced.
Artist’s main gripe against AI art is how the training material is taken without consent from the artist to create an AI that replaces the artist themselves. Japanese artists are especially vocal against this. Japan considers anime and the art associated with it an export and culturally significant. AI bots that replicate this art style is not only a threat to artist’s livelihood but also a threat to something culturally significant. Many artists, including vtubers have explicitly stated that they do not want their artwork used to train AI models. The phrase “AI学習禁止” (“No AI Learning”) trended on Japanese Twitter in response.
The public’s distain towards AI art has also caused issues for artists whose style is “similar” to what AI can generate. A Japanese anime artist was obligated to post a screenshot of their work in progress to combat allegations of them using AI to create their art. Since AI related ethics and copyright are still hard to regulate, artists have take it upon themselves to protect their works. Some famous vtubers have stated online that they do not allow use of AI in fan art or it may infringe on copyright issues.