My Discomfort with Japan's AI Business: What Does "Born in Japan" Really Mean?

Recently, I keep seeing the name Sakana AI.

“An AI startup from Japan.” “Japan’s fastest unicorn.” “An AI company taking on the world.”

When you hear phrases like that, it is natural to want to feel hopeful. Has Japan finally produced a company that can compete with OpenAI or Anthropic?

But to be honest, something about it bothers me.

First, let’s check the facts

Sakana AI is an AI R&D company founded in Tokyo in 2023. Its founders are David Ha, Llion Jones, and Ren Ito. On the company’s official site, David Ha is listed as CEO, Llion Jones as CTO, and Ren Ito as COO.

Llion Jones is one of the co-authors of the famous paper “Attention Is All You Need,” which laid the foundation for the Transformer. David Ha also has research experience at places such as Google Brain Japan. In other words, the research credentials are very strong.

Sakana AI has also been selected for Japanese government support related to the development of generative AI foundation models, and it has received computing resource support from NEDO.

Looking at all this, there is certainly a basis for calling it a “Japanese AI company.” It is headquartered in Tokyo, and it talks about AI for Japan, Japanese-language AI, and AI for Japanese companies and the public sector.

Still, a small discomfort remains in me.

The strange feeling behind the phrase “born in Japan”

Sakana AI is legally and geographically a Japanese company. So calling it “born in Japan” is not wrong.

But I feel there is a slight gap between what ordinary people imagine when they hear “AI from Japan” and what the reality actually is.

At the center are world-class AI researchers. And a major part of that research brand depends heavily on former Google talent from overseas.

Of course, that is not a bad thing. If Tokyo was able to attract world-class researchers, that is an achievement in itself.

But converting that directly into a story that “Japan’s technological power has finally caught up with the world” feels a little too rough.

It is located in Japan. It is built for Japan. Japanese companies have invested in it. Therefore, it is Japanese-born.

I understand the logic. But there is also something slightly dangerous about it, as if attaching a Japanese flag somehow makes something domestically made.

Are they trying to become OpenAI?

Another thing that concerns me is what Sakana AI is actually doing.

Sakana AI talks about AI inspired by natural evolution and collective intelligence, model merging, Japanese-language models, and AI Scientist. Reuters has also reported that the company created a Japanese-language model by evolutionarily combining existing models.

This is interesting. Technically, it is not just a thin wrapper company.

However, it seems to occupy a slightly different position from companies like OpenAI or Anthropic, which pour massive compute resources into building frontier foundation models head-on.

Rather, it looks closer to a company that skillfully combines existing models and research assets, adapts them to the Japanese market, and implements them for companies and the public sector.

That is important in itself. In fact, that may be exactly the kind of company Japan needs.

But if so, they should say that.

Not a dream story about “winning from Japan with a world-leading foundation model,” but a company that “makes global AI technology usable in Japanese workplaces.”

That would feel much more honest.

If Japan wins in AI, will Japanese people become richer?

To begin with, becoming a global leader in AI and making Japanese people richer are not the same thing.

You can see this by looking at the United States. AI companies’ market capitalizations have exploded, and investors and a small number of researchers are earning enormous compensation. But whether society as a whole has become broadly richer because of that is highly questionable.

The AI industry is different from the automobile industry.

With automobiles, there are factories, parts suppliers, logistics companies, dealerships, and mechanics. The industrial base is broad, and many people can enter the industry.

AI, by contrast, is decided, in the extreme, by a small number of brilliant researchers, GPUs, capital, cloud infrastructure, and data centers. It tends toward winner-takes-all dynamics.

Does Japan really need to force itself into that game and chase after the United States? That is something we should think about calmly.

What Japan should do

I am not saying Japan should abandon AI.

Quite the opposite. I think Japan should use it.

But instead of getting drunk on the label of “amazing AI from Japan,” we should think more seriously about how to embed AI into existing industries and real workplaces.

Manufacturing, healthcare, nursing care, government administration, logistics, education, back-office work for small and medium-sized companies. Japan has countless unglamorous workplaces that AI could improve.

What those places need is not only papers that surprise the world. They need the ability to understand operations, understand workplace constraints, connect with existing systems, and run things responsibly.

In that sense, the AI business Japan should aim for is not an inferior copy of OpenAI.

Rather than “building the world’s strongest model,” it should be: “Using the world’s AI to make Japanese workplaces a little better, step by step.”

That seems far more realistic to me.

In the end, what is the source of my discomfort?

Sakana AI is not a suspicious company. Rather, it has accomplished researchers, it has raised money, and it is doing technically interesting work.

But when the phrase “born in Japan” starts walking around on its own, the conversation becomes a little strange.

Being based in Tokyo. Building for Japan. Receiving investment from Japanese companies. These things certainly matter.

But the moment people start saying, “This means Japan’s AI technology will take over the world,” it suddenly begins to smell like advertising agency copy.

That is what bothers me.

AI is not a dream. It is a tool. It is not a poster for national pride.

Instead of dancing around the label “born in Japan,” we should ask: Whose work becomes easier because of it? Whose life becomes a little better?

At least for me, that is the kind of AI business I find more trustworthy.