Japan, the "most AI friendly country in the world"

The country's leaders are gung ho for AI. But will it kill the golden goose of its content industry?

(Astro Boy, ©Mushi Pro)

Just a reminder that I’m continuing to record “audiobook” versions of my essays. Love ‘em? Hate ‘em? Let me know what you think! And stay tuned as I’m planning to start releasing video excerpts on my long-dormant YouTube channel in the near future.

In 1963, a medical student turned manga artist named Osamu Tezuka coined the term “anime” to differentiate his latest creation from the “animeshon” imported from abroad. The television series Mighty Atom, better known in translation as Astro Boy, starred a robot kid with an atomic heart and an AI brain. The show’s popularity made Astro Boy the defacto face of technology in Japan – gentle but powerful, a friend to children and defender of justice. It proved equally popular abroad. One of its fans was Stanley Kubrick, who sent Tezuka a letter pitching a collaboration. It didn’t pan out, but six years later, Kubrick would unleash a far different sort of AI on the global consciousness with the murderous HAL of 2001: A Space Odyssey.

When Japan’s digital information minister Taro Kono spoke to Bloomberg last week, the dichotomy between Astro Boy and HAL was certainly on his mind. “In Japan, we grew up with manga, anime, where AI controlled robots are helping you to do homework, or fighting against villains coming from outer space. So AI technology is always friendly to the Japanese... Unlike European countries, we are not really worried about AI stealing jobs.”

Kono’s remarks followed a June Nikkei Asia report that Japanese lawmakers were looking to make their nation the “most AI-friendly country in the world.” This in turn built on a 2023 LDP whitepaper that touted, almost gleefully, that there were currently “no consideration of ‘hard law’ regulations focusing on AI in general.” The hospitality seems to be attracting foreign talent: Sakana AI, launched by several ex-Google employees, opened last year, and OpenAI set up shop in Tokyo in April. “Japan is a super important country to us,” OpenAI CEO Sam Altman said during a visit last year. “I think there is a long history of humans and machines working together here.” Score another for team Astro Boy.

But a recent news article belies both Kono’s and Altman’s rosy assessment of Japanese citizens’ feelings about AI. Last week, Nikkei reported that a newspaper association is calling for legal reforms to curb AI search, claiming that the systems summarize their reporting without directing users back to the source. Journalists aren’t the only ones worried about AI’s impact on their careers. In a 2023 survey, 94% of content creators professed deep ambivalence about generative AI. (Hayao Miyazaki famously proclaimed that he was “utterly disgusted” by an earlier iteration of the technology.)

This disconnect is a bigger deal than it might seem, because post-industrial Japan increasingly depends on the production and export of content, both for economic reasons and to stay relevant globally. The domestic market is worth roughly the equivalent of $30 billion USD currently, and the government is keenly aware of its growth potential. Officials say they hope to expand that more than fourfold to $129 billion USD over the next decade. The idea is to position content creation as one of Japan’s key industries.

Japan made itself rich in the postwar era by selling the world cars and computer chips. But it made itself loved by selling cultural products – games, manga, anime, literature, music, magical-realist literature, life-changing cleaning magic, you name it. And all of these products were created by hand. Painstakingly. Really, an absurd amount of handiwork goes into making them. Having worked in the game industry myself, and having many manga artist and animator friends, I can tell you that the level of work it takes to maintain illustrated entertainment empires like ONE PIECE or Demon Slayer, or virtual worlds like Elden Ring, or even seemingly lighter fare like the latest Mario title, is mind-boggling. Yet that effort is precisely what made Japan the planet’s factory of dreams.

Japan’s leaders want their nation to be an AI superpower. They also want it be an content-creation superpower. But these twin drives are in sharp conflict. Proponents tout the potential of generative AI to boost the efficiency of creative industries (even as they downplay the fact that they want to eliminate many of the humans involved with its production.) And AI models need to be trained on copious amounts of human-created content, the creators of which aren’t compensated or even acknowledged. (I know this from experience, as several books Hiroko and I wrote turned up in leaked training data for Meta’s “Llama” system.) Once trained, AI models output imitations of human-made content at a fraction of the cost. This is why so many creatives view it less as a tool than a piratical destroyer of their livelihoods.

You’d think the same politicians who are intent on ensuring Japan’s existence as a content superpower might want to pump the brakes a bit before inviting the digital equivalent of a vampire inside their home. I mean, take a look at how OpenAI made ostentatious use of ersatz Japanese imagery in the announcement of its Sora text-to-video model. Why buy Japanese stuff from Japan when you can generate mediocre simulacra for free? Yet OpenAI’s CEO seems to have made quite the impression on Prime Minister Kishida in a visit last year. Kishida seems to have been swayed by the promises of Altman and other AI proponents that their tools will turbocharge economic growth rates by increasing labor productivity. This is absolute catnip to the leaders of a nation whose population is ageing so rapidly that more grown-ups are in diapers than children.

Not everyone in the Japanese government seems oblivious to the potential downsides of generative AI. In February, the Ministry of Culture weighed in on a situation involving AI models being used to create parodies in the styles of well-known manga artists. “The ability of AI to mass-produce approximations of art will significantly reduce scarcity and customer demand…. we believe that creators' artistic styles should be protected in some way, and the use of generated AI should be regulated.” So much for that whole non-regulation thing. To be fair, the Ministry of Culture is part of the bureaucracy, a different sort of animal from the LDP, the ruling political party that is pushing for unfettered AI. It’s a hint that we can expect more debate about this supposedly settled topic in the future.

This seemingly innocuous document also articulates the precise problem that Japan will face in the integration of AI tools into its creative economy. Training AI models on Japanese content makes a lot of sense. For a variety of reasons foreign-made ones don’t do a very good job of reproducing Japanese imagery. Fans are savvy enough to want authentic stuff from Japan, and they want stuff from specific Japanese creators. If the handiwork of those creators is excluded from AI training data to protect their livelihoods (and for the record, I think giving creators the option to opt out is a good thing), the resulting models won’t be able to produce content that evokes them. What might these sanitized AI make instead?

There’s a hint in a METI report, reported by the Animenomics newsletter last week, called the Guidebook for the Utilization of Generative AI in Content Creation. A truly low-effort AI-generated cover sets the tone for the contents, which identify risks and highlight successful “use cases” for generative AI. The problem is, none of them are particularly successful: forgettable character designs, rote text-generation systems, poorly reviewed video games, and bland advertisements. AI may well have saved time in producing all of these things, but, to paraphrase KonMari, life-changing magic they aren’t. And therein lies the biggest problem of all.

Generative AI has its uses, but creating compelling content isn’t one of them. “AI thrives when our need for originality is low and our demand for mediocrity is high,” as sci-fi author Ray Nayler put it in an essay for Time last year. When it comes to Japan’s cultural products, consumers crave the opposite. They want novelty and excellence — which translates into authenticity. Generative AI tools may well make content creators more efficient. But efficiency has never, ever been the name of the game when it comes to making popular art in Japan. The meticulous process of making the ukiyo-e woodblock prints and handicrafts that charmed global audiences in the 19th century was anything but efficient. Little has changed in the process of making the games, manga, and anime that obsess modern-day fans.

Can Japan become an AI superpower without compromising everything that makes its content so beloved around the world? Can its creators continue to make a living when their own content is being used to train their replacements? If their handiwork is excluded, will Japanese AI offer anything the world really wants? These are big questions, but Japan’s leaders don’t seem to be interested in asking them.

In some ways, this is just business as usual in a nation that long underestimated the contributions hardworking creatives made to society. Using a pop hero like Astro Boy to claim that Japanese aren’t worried about AI isn’t simply disingenuous — it’s a slap in the face to the content creators who are most invested in how this new technology will affect their lives. To borrow Kono’s words, it sounds suspiciously like something some “villain from outer space” might say.