I thought that this week would be interesting when I read my colleague Kylie Robison’s article about OpenAI’s new image creator on Tuesday. I find generative AI images incredibly entertaining. I spent a large part of the day watching other Verge staff test ChatGPT, in ways that ranged from cute to cursed.
On Thursday afternoon, however, the White House decided that it would ruin it. Its X page posted a photo of a crying detainee, claiming that she was an undocumented alien and a fentanyl dealer. It then added a cartoon, almost certainly AI-generated, of an officer handcuffing a sobbing woman – not attributed to a particular tool, but in recognizable style of ChatGPT’s super-popular Studio Ghibli impersonations, which have been sweeping the internet for the past week.
A software tool’s ugly use shouldn’t be a reason to condemn it. As the picture appeared on my social media feeds, I began to feel that the adorable Ghibli filters and the White House’s social media blitz were made for each other. Both are, as contrary as it may sound, the result of a mentality that views basic decency and callousness in the same way.
While we’ve reached out for more information to OpenAI and to the White House, the move was a bizarre advertisement from a company that President Donald Trump has close ties with. We don’t know whether ChatGPT, or another AI generated the image. If the White House did commission an artist, and they read this, I would love to hear from the artist. OpenAI CEO Sam Altman is promoting Ghibli-style images as a cool, exclusive feature for ChatGPT’s premium tiers. Trump has been a vocal supporter of OpenAI’s Stargate Project, announcing the project at a joint press conference with Altman.
AI Ghibli and Trump seem to fit together in a bizarre way. The White House clearly wanted to engage in a form of online performative sadness. This is the same account which posted a video of prisoners clinking their chains called “ASMR: illegal Alien Deportation flight”. Even if all the information is accurate, it’s still gross and juvenile. No one can defend the public humiliation, in a joke, of what appears to a low-level immigration suspect as good governance, effective messaging, or moral good.
Ghibli’s aesthetic is so wholesome, it undermines this. Even a prominent Silicon Valley conservative pointed out how depicting an anime woman crying while being arrested by a stone faced agent did not make most people sympathize with the agent.
AI in general is the MAGA movement’s primary aesthetic producing plenty of strange, tasteless works. It’s the natural result of their love of photoshopped images and political cartoons portraying Trump as a brash muscleman. It’s also a product of links between Trump, the AI industry, and things like “Stargate” and Elon Musk’s “First Buddy.” Eight years ago, a company might have distanced themselves from someone using its memes to promote mass immigration.
i don’t know what OpenAI and Altman think about the White House promoting a ChatGPT joint advertisement and a brutal, and likely partially illegal,effort to expel immigrants from America. Altman was a known supporter of progressive causes before this administration. OpenAI’s team claimed that ChatGPT was supposed to have highly flexible guardrails. They may argue this is no different than using Photoshop offensively. It may go without saying, however, I don’t think OpenAI should or be able to block the production of an image like this — if the White House hadn’t posted it, you could read it as a form of protest against these arrests.
Eight years ago, when Silicon Valley was at odds with Trump, a major technology company may have disassociated itself. It’s not difficult to make a statement like “OpenAI believes that maximum artistic freedom is important, and that we are responsive to user requests. However, this administration’s position does not reflect the values of our company.”
There is a lot of social and political pressure on OpenAI to not do that. OpenAI’s staff may have their own opinions, but it’s not good business to be feted by an avenging president, then criticize his policies. This is especially true in the context of a Silicon Valley rightward shift.
There’s also a deeper issue at play because the Ghibli filters themselves have a sour taste — at its core it’s an echo of the Trump era utter disregard for human beings.
Ghiblifying images is not something I can resist. Seriously. Some of them are really cute. I don’t believe that these images were created maliciously. People have been using anime filters for many years. Hayao Miyazaki is a famous anti-AI artist, and his name is synonymous with that of the animation studio. He is quoted as saying that an earlier version AI animation was “an insult to the life itself.” There’s no indication that he approves ChatGPT being used by OpenAI to imitate his signature art style, probably due to training in his art. Brian Merchant says that using Ghibli’s artwork for publicity is a powerful move. It says loudly to the artists who create ChatGPT that we will take what we want and tell everyone about it. Do you consent? We don’t care.
OpenAI should have approached artists as collaborators, not as obsolete producers of raw data for training.
Modern tech and politics share an ideology of dominance: the idea that power, money, and authority can be best used by forcing others to do whatever you want. This is self-explanatory for Trump. It manifests itself in every useless AI feature that replaces anything useful. Criticism is the senseless destructionof great men. Empathy, self examination, and compromise are weak .
It is ironic that the Ghibli filter has become wildly popular in a sea full of pointless and dysfunctional AI applications. OpenAI has captured its appeal in a world that doesn’t show disrespect to the people who’s work it is building. AI companies could have easily (if not at the same cost) built their products by approaching artists as partners, instead of obsolete producers who provide raw training data. OpenAI could have found a cartoonist or animator who would be willing to adapt ChatGPT for their style, promoting a lesser known artist. It would take a belief that people who don’t fit the Great Men category are worth learning from and working with, not just overpowering.
Do I still think that paying for ChatGPT is a sign of a bad character? Paying for anything can fund something inhumane, harmful, and often in a more destructive way. We all draw our own lines, and I am not sure where my line falls.
However, the Tesla Takedown protests do show how tying a business to toxic politics backfires. ChatGPT is used by many people to create cute pictures of loved ones. It’s sad that OpenAI quietly let the White House use the meme to bully those who are powerless. Do OpenAI researchers believe this is a good thing for “AI for Good”? Where will Silicon Valley companies draw their and boundaries as they compete to sell their AI systems?
