Weekly Scroll: Ghibli-fy Everything
Plus: Elon's fake numbers and a Good Shrek Prediction
Welcome to the Weekly Scroll! As a reminder, for the next two days you can get 30% off a paid subscription to Infinite Scroll. Paying subscribers are what powers this blog for everyone, so join up today!
Are the Studio Ghibli AI Pictures Evil?
Last week, X user Grant Slatton posted the first truly viral AI meme of 2025:
This seems to have started as a result of ChatGPT’s latest update, which greatly improved the 4o model’s ability to edit and manipulate user-uploaded images. Within hours hundreds/thousands of X users were posting pictures of their families and friends made into very cute Studio Ghibli style characters.
Three additional factors made the trend go mega-viral. First, they’re just very cute and wholesome. Second, people realized you could do this to any image, meme, or historical photo:
And third, some people get really, really angry about AI art, which leads to The Discourse. Social media seemed split into two camps - those who loved the images and wanted to make more of them, and those that thought the AI-generated images were a moral travesty. Or as Slatton said, “If anyone is curious my DMs look like this x100”:
This is the first really big AI meme since we had the cartoon characters doing 9/11 memes back in 2023. How big? Sam Altman had to announce rate limits for creating new images, claiming that OpenAI’s servers were melting under the strain. Since the last time we did this back in 2023, the discourse around the ethics of creating and using AI-generated images has exploded. Some of the largest pushback I’ve ever gotten here at Infinite Scroll came not from anything I wrote, but from my occasional use of AI images.1 This is a subject that many people are very, very passionate about - I respect their opinions and I want to tread carefully here. But while treading carefully, I think the people who are really angry about this are misguided.
Before we get to the ethical issues with AI images, it’s important to state the obvious benefit - people like it! The upside is that people get real enjoyment and happiness out of this tool. This trend really broke containment - my 60 year old mother texted me with a request to Ghibli-fy a family picture. Lots and lots of people had cute and nostalgic moments thanks to ChatGPT.
What are the downsides? Depends on who you ask. Some of the downsides are basically fake, like the overblown concerns about water usage.2 Some people are concerned this steals jobs from artists, but I think that misses the mark as well. The counterfactual here isn’t “All of these images would have been drawn by real people, and we’d live in a golden age of artists each getting hundreds of commissions online”. The counterfactual is that none of the images would have existed at all. Would it really be better if images like this never existed? Who would be helped?
Others have concerns about intellectual property. I think this is a reasonable concern! But I can’t help but notice it’s selectively applied to AI-generated images and no other kind of online content. I’ll see a post that claims AI users deserve to be tarred and feathered for stealing intellectual property, and then see the same user thirty minutes later posting Sopranos memes. For some reason it never occurs to them that the Sopranos is also a work of art made by hard-working artists and protected by copyright! If we applied IP protections strictly, half the internet would disappear overnight.
It doesn’t seem likely to me that Studio Ghibli is going to be hurt in any way by this trend. If anything, I’d bet that we’ll see a resurgence in interest in their films - especially among the younger generations who may not yet know about how great movies like Spirited Away or My Neighbor Totoro are. This reminds me of how sports leagues like the NFL and the Premier League were initially furious that fans would post highlight clips on social media. They spent a massive amount of effort copyright striking people’s posts and trying to ensure there were no IP violations, which was stupid and counterproductive. They eventually realized online clips were helpful and not harmful in growing their fanbases, and embraced fans spreading short videos of their athletes.
Perhaps the most serious objection has come from Studio Ghibli co-founder Hayao Miyazaki himself. During a documentary he filmed in 2016, he said of computer-generated animations “I am utterly disgusted… I strongly feel this is an insult to life itself”. This is a pretty serious objection! If someone explicitly asks you not to ape their visual style, perhaps we should listen to them.
Unfortunately, while dozens of internet trend reporters are breathlessly repeating the ‘insult to life itself’ line… it’s not really true. Miyazaki did say that, but the wider context is that he was watching this monstrosity:
Almost ten years ago, one of Miyazaki’s employees showed him a horrifying eldritch abomination crawling along like the girl from The Ring. Of course he hated it! There’s no indication he would feel the same way about current gen AI - maybe he would, maybe he wouldn’t.3
I do get why some people feel uneasy about this stuff. If you prefer not to engage with it, I understand. But at the same time, the level of vitriol online can be pretty excessive here. People are posting guillotine memes and telling AI posters they have ‘rotted souls’ and don’t deserve to be treated like humans. For making cute version of family photos to show their wife or husband. Is this really that offensive?
Either way, the meme has now broken containment. We have people posting offshoots in Rick and Morty style, as 18th century portraits, in Van Gogh style, and more. But what will ultimately end enthusiasm for the trend is the White House jumping in to Ghibli-fy an image of an illegal immigrant getting deported. There’s nothing to kill a meme like politicians and brands trying to jump on board.
Donald Trump has no idea what Signal is
When asked for his thoughts on the Signal scandal:
“It’s a bad signal” says confused elderly man about new fangled technology.
Meanwhile, the Signal scandal continues to build steam. It was revealed by Der Spiegel that the account and password information for Tulsi Gabbard, Pete Hegseth, and other administration officials was publicly available in hacker data dumps. One additional side effect of the scandal - Signal has had their biggest week ever in new downloads.
EU and Meta
Meta is once again attempting to roll out a $14 per month plan to get ad-free Instagram in Europe. This move comes in response to European regulations concerning targeted ads - Meta’s workaround is to give users the option to either approve targeted ads on their account or pay a monthly subscription. This isn’t a new plan from Meta:
In July 2024 I wrote:
Last week, the EU charged Meta with violating the DMA’s regulations. Here, the EU had ruled that targeted advertising that tracks consumer preferences with cookies was against the DMA, and ordered Meta to come up with an alternative business model. Meta complied, and allowed users to either choose between receiving targeted ads or to pay a monthly fee for a no-ads experience. The EU then ruled both of those options were out of compliance, and has sued Meta and ordered them to find yet another business model.
They don’t like targeted ads. They don’t like monthly subscriptions. Essentially, they want Meta to provide their product to every European for free, while making the business model that pays for that product illegal.
It’s a continuous theme here that the EU has absolutely horrendous tech regulations, and I continue to think that Europe is being ridiculous in this case. Advertising is what keeps Facebook and Instagram operating, and non-targeted ads are worth a tiny fraction of the targeted variety. Forcing Facebook to operate on a non-target business model is the worst kind of regulatory meddling, and in the long run could even potentially lead to the company abandoning the EU market. We’re a long, long way from that, but companies like Apple are now routinely delaying or de-featuring their products in the EU due to the insane regulatory burdens there.
Some Very Good Shrek Predictions
Tiktok failed to load.Enable 3rd party cookies or use another browser
Fake Number purchases other Fake Number
Earlier this week Elon Musk announced that his company xAI has purchased X née Twitter for the sum of ‘I guess some money, probably what we spent on it, who knows’.
The precise numbers look like this: xAI is acquiring X in an all stock deal that values xAI at $80 billion and X at $33 billion, roughly Elon’s initial $45 billion purchase price minus the $12 billion of debt he loaded into the company at time of purchase.
Those sure are some numbers, I guess. But let’s be honest. The precise numbers don’t matter. This is one Elon-controlled company buying another Elon-controlled company, the numbers could have been anything he wanted them to be. That’s how selling things to yourself works.
Allow me to get moderately side-tracked talking about theories of financial value. When we say that McDonald’s is worth about $220 billion, what we roughly mean is that if you multiply out the stock price by the number of shares, it comes to that total. That number isn’t provably true in the same sense that 2+2=4 is true, but it’s a very strong estimate. McDonald’s is a publicly traded company. The stock price is the result of hundreds of thousands of people making data-informed estimates of what the company’s future cash flows are worth, and large markets like this one tend to be pretty efficient. So in that sense, we can trust that McDonald’s is “worth” $220 billion.
When we say that X is worth 33 billion, that’s entirely the result of Elon Musk’s willingness to overpay for it at that price. It’s not a real value tethered to business fundamentals or market forces. xAI’s valuation is similar. Sure, some venture capitalists might invest in it at a certain number. But xAI is a company doing virtually no revenue compared to its size. It may have value years from now, but its current value is basically the fever dream of a very small number of people not really subject to market forces that would keep them honest. Is the combined new company worth more than a hundred billion dollars? In some sense, sure, I guess? But it could be a trillion. Or zero. Or anything in between.
I think the big actual thing Musk gets out of the xAI/X merger is ego salvation from his Twitter overpay. It’s been a huge embarrassment that he overpaid for the bird app and then banks immediately started writing it down, anticipating that it might lose as much as 90% of its value. Because Elon Company A bought Elon Company B, he could instruct them to pay a price which avoids that particular shame.
In other Musk news: he’s planning to step down as head of DOGE in May. This may just be a legal maneuver to keep his status as a limited-time special employee and thus avoid the oversight and regulations that would come with being a full-time employee of the White House. Or it could be that he underestimated how much everyone would hate what he’s doing and how it would end up cratering Tesla’s stock, so he’s trying to declare victory and retreat. Or maybe he just needs more time in his schedule to email Reddit’s CEO about mean posts.
Links
The State Department is moving to use ‘enhanced’ social media vetting of student visa applicants. This is plainly anti-free speech and a pretty gross violation of America’s values.
A survey from Deloitte shows that social media and ‘creator content’ are taking an increasing share of entertainment, to the detriment of traditional television and movies.
The rise of IRL, on location live streaming has led to an increase in ‘stream sniping’ - when fans track down streamers in public to approach, accost or even threaten them.
How is Spotify’s move into video podcasting going? Well, porn keeps popping up in their ‘Top Video Podcasts’ charts, so…
Netflix’s show about a social-media driven pre-teen murder, Adolescence, is breaking records.
Facebook is introducing a Friend and Family tab to try to recreate the feel of the original Facebook, where you got life updates from people you knew rather than content from content mills.
Posts
Horrifying Dora The Explorer Foot Mukbang
The NY state government is Minion Posting
It’s very occasional - by my count, I’ve used an AI-generated image as the header image for a post only twice in the last four months.
A typical ChatGPT query has the same environmental impact as driving a car 15 feet.
Of note - in Japan, copyright specifically does not apply to training AI models.
I’m a pretty new subscriber, but I’m hopping off here given what I feel is a very intellectually dishonest stance on AI art.
Illegally (at least in most territories, and at the least without permission) training models on intellectual property so you can create a program that allows people to make content in the explicit branded style of a studio you are in no way affiliated with is not the same thing as screenshotting a TV show and putting some white text over it, and I think you know that.
And while you can write various justifications to minimize the environmental cost of AI usage, the fact that the “average query is equivalent to driving a car 15 feet”…adds up fast when there are millions of queries. When we’re staring down cataclysmic climate change, why are we ADDING a completely superfluous way to contribute to that? Not to mention that image generation, as I’m sure you know, takes more energy than a ChatGPT query, though you’ve misleadingly only mentioned the latter’s cost here.
I also think you know younger people today, who have attention spans that run the length of TikToks, are not going to see this meme (which only seems to appeal to older people who are already familiar either Ghibli anyway) and run out and watch Mononoke legally. All this stands to do is cheapen the image of a studio and artist long associated with auteur creation.
I’ve really enjoyed your write-ups and round-ups even though I have side-eyed the usage of AI header images. I don’t believe every use of AI = one artist’s gig lost—that’s naive—but the continued use does serve to devalue the creative arts, enrich some of the scummiest Silicon Valley vultures to ever exist, and negatively affect the environment for what? Produce the equivalent of a photo filter? What price we pay for progress.
I’m sure this will fall on deaf ears, but I was surprised to see you make such disingenuous arguments, particularly the comparison between AI image generation and friggin’ screenshot memes, and wanted to let you know why you lost a subscriber today.
I dunno man, the EU might be on to something with banning advertising. The ad model is a huge part of the reason our media ecosystem is so screwed. In a world of CPM and CPC business models, traffic is everything. The easiest way to drive traffic is outrage and lies.