Opinion | Why Does Your AI 'Ghibli' Look So Terrible? Because It Is

Earlier this week, OpenAI CEO Sam Altman joked on X (formerly known as Twitter), “Can y'all please chill on generating images, this is insane our team needs sleep”. This was a reference to the massively popular online trend of Studio Ghibli-style images being churned out by OpenAI's new ChatGPT Image Generator. Altman's words struck me as deeply ironic — the whole point of AI art, as its proponents keep reminding us, is that sleepless nights are not required to create these images. Users worldwide currently ‘Ghiblifying' their personal photographs aren't losing any sleep. The only people involved in this situation who've spent sleepless nights creating art are Studio Ghibli's Hayao Miyazaki and his team of artists, animators, et al. In other words, the people currently being ripped off by half the planet because in 2023, the Japanese government declared that copyrighted works could be used, sans permission, for AI-training endeavors (a position that the US, the UK and the EU are still distant from, legally speaking).

When Theft Is Art

The ongoing Studio Ghibli situation really is the perfect distillation of AI art discourse—an act of material theft and creative bankruptcy is being paraded as pathbreaking innovation. And you're being asked to shift your gaze from the scene of the crime and focus instead on the genteel faces and kind eyes of Studio Ghibli grandmas.

To my mind, there are three principal aspects of the AI art debate: material (pertaining to the labour-capital relationship between artists and corporations), legal (pertaining to the ownership/distribution of art) and aesthetic (pertaining to inherent artistic value). And on all three planks, generative AI—at least in its current shape and form, exemplified by OpenAI's ChatGPT—seems to me to be a net negative for humanity in general, and a crushing blow in particular to artists and creators everywhere.

Manufacturing Consent

First, the material aspect. Let's say you are a freelance artist and designer who was commissioned to create a few minor logos and webpages for a mega-corporation, a couple of years ago. At the completion of said project, you were paid what you assumed was a fair wage. But now, a couple of years later, you realise that the mega-corporation is also deeply invested in AI, and has been feeding all of your work (not just the finished product, but stage-updates and mockups as well) into the data-guzzling, AI-training machine. You never consented to this process, but your consent may have been manufactured by vaguely-phrased clauses buried deep within your standard-issue freelancer's contract.

In this way, the mega-corporation gets to use your work twice: once in the usual way, and then once more for AI-training. The mega-corporation gets two bites at the apple but has to pay only for the first one. This is theft in the most ab initio, first principles sort of way. Every day is a cost-cutting day if you can simply get free work off people.

The Copyright Loophole

The second aspect speaks to legalities, which in this case means the concept of copyright. For successive decades, large corporations told us that copyright law was sacrosanct, that without the legal protections that controlled access, writers and artists and creators everywhere would be impoverished. You might remember the tragic suicide of Aaron Swartz (1986-2013), the genius programmer behind the RSS web feed format and the development of Reddit. Swartz was threatened with 50-plus years in prison for downloading a stash of JSTOR articles (JSTOR is a repository of paywalled academic articles) off MIT servers, which prompted him to take his own life.

But now, we are suddenly being told by Altman and co. that unless we suspend the idea of copyright completely and allow copyrighted works to be chewed up by AI-training ventures, we are being anti-progress and undemocratic. It's funny how ideas of progress, fairness and the “greater good” seem to gravitate around the interests of billionaires, shifting one way and then another in sync with share prices and investor valuations. Why should independent artists and creators embrace a technology and an extra-legal framework explicitly designed to undercut and eventually replace them? In 2024, Mira Murati, then-CEO of OpenAI, said that AI will definitely replace some creative jobs but that it was alright because “maybe those creative jobs shouldn't have existed in the first place”. In the ‘growth-hack' model of the universe favoured by Silicon Valley executives like Murati, creativity is an active threat that can and should be snuffed out by machine logic, the imperceptible hum of digital ones and zeroes. No surprise then, that influential groups of creators worldwide are suing OpenAI and co. for copyright infringement—the comedian Sarah Silverman is leading one such lawsuit against Meta, for example, while The New York Times has its own lawsuit against OpenAI.

Art Minus The Artist

The third aspect of the situation speaks to the purely aesthetic, the raw artistic value (if any) of AI-generated imagery. Every single professional artist I have spoken to or heard online testifies to how terrible, soulless and emotionally hollow AI-generated art is. Quite simply, the tech simply isn't very good. How could it be any good, when the vast majority of culturally significant art is, in fact, still protected by copyright law? Existing image generators are being trained on leftovers, discards and amateur/learner fare, for the most part, with Japan being a rogue exception, of course, thanks to its whimsical interpretation of copyright. For professional artists, the most common ongoing use case is to create mockups or demos for prospective clients—demonstrating what a particular style or color palette might plausibly look like. And even there, as professionals would know, there are significant limitations to the output.

The reason why AI-generated art is so terrible brings us to an intangible fourth aspect of the situation: the philosophical. Think about the very concept of image generators for a minute—they seek, essentially, to conjure art out of thin air while completely isolating the (ChatGPT) user from the process. An artefact minus the artistry. A hunk of butter that's all hunk and no butter is futile, both nutritionally and ontologically.

An artist's work, especially when it's beyond a certain minimum quality, speaks to their very soul, the story of their life. And the story is ingrained upon their callused hands, like a grand old tree marking the passage of every solar cycle with a concentric ring. To remove the process of learning your craft, to remove the inevitable sequence of mistake-correction-growth from an artist's life, is downright dystopian and a fundamentally antagonistic attitude towards art and artists.

An Insult, Indeed

Back in 2016, Hayao Miyazaki himself was given a brief demonstration of AI-based animation techniques, and it left the veteran animator and filmmaker unimpressed, even shaken. The phrase he used to describe the experience was “an insult to life itself”. I think he was absolutely right. If you're currently getting a chuckle out of ‘Ghiblifying' your last year's Himachal road-trip, I'd urge you to listen to the man responsible for those genteel faces and kind eyes.

(Aditya Mani Jha is an independent writer living in New Delhi. His first work of nonfiction will be published by Oxford University Press in 2025)

Disclaimer: These are the personal opinions of the author