The latest version of ChatGPT lets users make Studio Ghibli-style anime memes. It again raises questions about AI and copyright law.
On Tuesday, tech firm OpenAI launched the new build of its AI image-generator software, GPT-4o. Billed as being “capable of precise, accurate, photorealistic outputs” it has instead been used for something more whimsical: to make memes in the style of Studio Ghibli.
Over the past couple of days, social media – specifically X/Twitter/The Upside Down – has been awash with images that are essentially riffs on existing memes, but rendered in an art style similar to that of master animator Hayao Miyazaki.
There’s a Studio Ghibli-style version of the kid smiling into the camera lens as a house blazes in the distance. There’s an anime JD Vance, and a cartoon rendition of a couple walking hand in hand, the male looking back distractedly at a lady in a red dress. If you can think of a recent viral image, there’s probably a Studio Ghibli version of it on the internet by now. We won’t bother posting them here, but Daily Beast has a round-up of tweets if you’re interested.
For web dwellers looking for a quick source of clicks with minimum effort, GPT-4o is a dream come true; already, Twitter users have racked up millions of views with their AI-generated images. One resident on the platform recently decided to use their viral post, which has received over nine million views, to promote their crypto newsletter.
Noting all the Ghibli memes, OpenAI boss Sam Altman got in on the act himself, even changing his profile pic to an anime-fied version of his faux-casual image (thanks, Variety).
In response, some (including the Daily Beast) have pointed to Miyazaki’s known distaste for generative AI. On seeing a demonstration of an early bit of AI tech a few years ago, he famously condemned it as “an insult to life itself.”
Between the opposing reactions of glee and distaste, however, an important question emerges. Take a close look at some of the memes floating around, and you can see that the characters have the distinctive pen marks and shading of a hand-drawn anime cel. In order to copy the specific style of an anime cel so closely, GPT-4o must be trained on hundreds, thousands or perhaps millions of still or moving images. If they’re taken from the work of Studio Ghibli or other animation houses, then that means that OpenAI is once again making a business out of taking copyrighted work, laboured over for countless hours by flesh-and-blood artists.
Read more: AI | 16 firms that have used artificial intelligence, and how
Is this even legal? We’ll wait and see whether Studio Ghibli has a response to the memes – particularly ones where its style is used to riff on such things as, say, the assassination of John F Kennedy. Some publishers and authors have attempted to push back on generative AI, with court cases currently deciding on whether ‘training’ software like ChatGPT counts as fair use, as Sam Altman insists. Meanwhile, tech companies are also trying to lean on governments to change copyright laws so that anything can be fair game for their training models.
What we’re seeing, in essence, is cultural vampirism. Generative AI feeds on the work of artists to create its output. Creative communities are finding it ever more difficult to find resources, even as tech companies rake in trillions from investors. The way things are going, we won’t see hand-drawn masterpieces like Studio Ghibli’s The Boy And The Heron produced for much longer.
At some point, we’ll all have to collectively decide which future we want: one filled with art that has intent and meaning, or endless torrents of self-replicating bilge. I know which I’d prefer, and it isn’t anime-style images of JD Vance.
