The Opportunity That Emerges Once Japan Accepts It Cannot Beat the U.S. and China in AI

“Japan is a year and a half behind the U.S. and China in AI development.”

An article making that claim had attracted attention. It said the reasons were slower decision-making and a gap in financial resources.

That makes a great deal of sense to me.

AI development today is not a field where one brilliant researcher can turn everything around by writing a single groundbreaking algorithm. It requires building enormous data centers, securing semiconductors, supplying them with power, and recruiting excellent researchers from around the world.

In short, it is a war of sheer resources.

In the United States, a single company can mobilize enormous sums of money, while China is also investing at tremendous speed through both the state and private companies.

Against that backdrop, it is difficult to argue that Japan can catch up simply by investing more in homegrown AI. The other countries are not going to stop and wait.

Even if we assume that sufficient funding is made available, Japan would then repeat the usual rounds of internal coordination, approvals from the relevant ministries, and reviews by third-party committees. It would take a year just to get started, during which time U.S. and Chinese models would probably advance by another two generations.

Financially and culturally, AI development is impossible for Japan.

If that is the case, we might as well admit it first.

It will be difficult for Japan to beat the U.S. and China in the race to build AI itself.

Instead, Wait for AI to Become a Commodity

Does that mean Japan has no chance of winning?

I think the opposite is true.

Japan should aim not at AI itself, but at the stage that comes next.

AI will eventually become a commodity. AI that would have been available only at the laboratory level a few years ago can now be used for a few thousand yen a month.

As this trend progresses, AI will become a technology like the cloud—something whose existence is simply taken for granted.

That is where Japan should compete.

Japan did not invent the automobile either. It was a complete latecomer.

But once the technology matured and the competition shifted from “Who can invent the automobile?” to “How can we mass-produce cars that break down less often, cost less, and use less fuel?” Japanese companies were strong.

The inventor does not necessarily remain the winner in the industries built around an invention.

The same thing could happen with AI.

If Japan Is Going to Compete, It Should Be on the Side That Uses AI Relentlessly

Manufacturing, healthcare, logistics, construction, and retail—the front lines of these fields contain enormous amounts of know-how that only a particular industry or company understands.

“When this machine starts making this sound, it is about to break down.”

“The manual says to do it this way, but this is how we actually do it on site.”

Simply building an enormous AI model does not give you this kind of knowledge.

I have worked in IT for many years, and even in system development, the difficult part is often not the technology itself but translating the realities of the workplace into a system.

The same is true of AI.

Even if Japan cannot build the world’s highest-performing model, it can still incorporate that model into business operations, connect it to data from the field, and turn it into a tool that people can actually use.

This, I believe, is where Japan should focus.

Physical AI—the combination of robots and AI, for example—might also provide Japan with a tailwind.

I believe Japan remains among the best in the world when it comes to artisans’ skills, tacit knowledge in manufacturing, and knowledge grounded in physical experience. Combining these strengths with AI could be powerful.

But Japan Must Not Fall Completely Behind

Of course, this does not mean that Japan should stop researching AI itself and simply buy the latest models from the United States.

If you do not understand the technology, you cannot use it effectively either.

Even if Japan cannot take the lead, it needs to remain close enough to understand what is happening.

It should keep running one lap behind, or at most about one and a half laps behind.

Then, the moment the technology matures and prices fall, it should move rapidly to put it into practical use.

Japan may not be able to win the resource war of building foundation models, but the competition to bring AI into manufacturing, healthcare, logistics, small and medium-sized businesses, and homes is a different matter.

Japan Does Not Need to Build the World’s Best AI

Hearing that “Japan cannot beat the U.S. and China in AI” can sound like a declaration of defeat.

But:

Leaving a competition that Japan cannot win is completely different from leaving the AI era behind.

Japan does not need to build the world’s best AI.

It should become the country that most relentlessly forces AI into real-world workplaces.

Japan has long excelled at improving things introduced from abroad and using them so thoroughly that their original form becomes unrecognizable.

That approach is just about right for AI too.

The United States can build the world’s most powerful AI.

We will have that AI work overtime for us.