In an attempt to accelerate the catch-up with China and the United States, Japanese companies have stepped up their deployment of AI, focusing on applications

Following Nvidia's announcement on the 4th that it would cooperate with SoftBank and other Japanese companies to develop generative artificial intelligence (AI), there is new news about Japan's development in the field of AI. According to a report on the CNBC website on the 11th, Hiroshi Mikitani, CEO of Rakuten Group, said on the same day that the company plans to launch its own artificial intelligence language model in the next two months. In addition to this, Japanese telecommunications giant NTT has also announced that its proprietary large language model will be launched in March next year. According to the report, Japanese companies are lagging behind their U.S. and Chinese counterparts in the development of large language models, but they are trying to catch up quickly.

Nvidia plans to collaborate with Japanese companies, including SoftBank, to develop generative AI. Pictured here is Nvidia CEO Jensen Huang on December 5 in talks with Japan's Minister of Economy, Trade and Industry. (Visual China)

"The first step is being taken through the efforts of the private sector"

According to reports, Rakuten Group is involved in a variety of business areas such as banking, e-commerce and telecommunications, so it has a large amount of "very unique" data. Hiroshi Mikiya said that the training of large language models requires a huge database to support, and "no one has a database like ours." At the same time, he said that the push for AI would bring "huge profitable growth" to businesses. The company plans to use AI models internally to improve operational efficiency by 20%.

Large language models are key to generative AI development. Since the release of ChatGPT by American start-up OpenAI last year, generative AI has become a hot topic in the tech world. According to research by Goldman Sachs, breakthroughs in generative AI research could drive global GDP growth by 7% over the next 10 years. In this context, global technology companies are also working on large language models. CNBC reported that to date, major tech giants in the United States and China have launched their own large language models. Google recently released the Gemini model, and xAI, which has only been established for more than half a year, has also released its first AI model product. In the Chinese market, Baidu, 360 Group, Alibaba, Huawei, etc. have successively launched their own large language models.

According to the report, although Japan is famous for its high technology, it has fallen behind in this wave of generative AI competition. At present, many Japanese companies are accelerating their efforts in the field of AI, trying to catch up with their counterparts in China and the United States. "We want to be a leader in the AI revolution. Masayoshi Son, CEO of Japan's SoftBank Group, said at this year's annual general meeting of shareholders. According to the Nikkei Asian Review, SoftBank has been selling some company shares since April to free up cash and focus on AI research.

Kenyuki Kojima, co-founder of Kotoba Technology, a Japanese large language model startup, said that while Japan has not yet caught up in the field of generative AI, it is taking its first steps through the efforts of the private sector. He added that once a "robust infrastructure" is in place, the remaining technical challenges could be "significantly mitigated" by leveraging open source large models and associated data.

The lack of software engineers is a big problem

Kojima's current lag behind the United States, China, and the European Union in large-scale model development is largely due to its shortcomings in in-depth Xi and broader software development.

CNBC reported that deep learning Xi needed "strong software engineers" to develop the necessary infrastructure and applications. However, according to Japan's Ministry of Economy, Trade and Industry, Japan will face a shortage of 789,000 software engineers by 2030. According to the International Institute for Management Development in Lausanne, Switzerland, Japan is currently ranked 28th in terms of technical knowledge out of 63 countries.

In addition, the development of large models in Japan also faces hardware challenges. Training large language models requires the use of supercomputers such as IBM's Vela and Microsoft's Azure, and according to the Nikkei Asian Review, no private company in Japan has such a "world-class machine".

The Nihon Keizai Shimbun also mentioned that the salary increase of high-end information technology talents in Japan is weak, and it is also at a disadvantage in Asia. The salary gap between data specialist positions in Japan and China has widened from around 30% in 2019 to 70%. Among them, the highest annual salary of "data scientists" engaged in big data analysis and other work is 23.5 million yen (about 1.16 million yuan) in China and 14 million yen in Japan. The highest annual salary for a "data architect" who directs the use and construction of data in a company is 23.5 million yen in China and 18 million yen in Japan, a difference of 5.5 million yen. The highest annual salary for an "AI Global Solution Architect" who uses AI to help solve customer problems is 23.5 million yen in China, which is 5.5 million yen higher than in Japan. The salary gap between Japan and China widened in 2022, and although the yen continued to depreciate and other factors, the overall salary increase for AI-related talents in Japan was generally weak.

Most Japanese companies value generative AI

The Nikkei Sangyo Shimbun quoted Germany's Statista as predicting that the global generative AI market will reach $44.8 billion in 2023 and is expected to exceed $200 billion in 2030. China and the U.S. are the main drivers of AI development, but in Japan, companies are focusing more on AI applications for purposes such as improving productivity.

In May of this year, the Japanese government held an "AI Strategy Conference" to convey the message that Japan envisages that the development stage of generative AI will be mainly led by major companies such as OpenAI and Google in the United States, and that they should disclose the structure and functions of their development results to the world. There are reports that Japanese companies are more willing to focus on the application of generative AI unless they have special financial strength.

According to a survey previously released by the Nihon Keizai Shimbun, only one of the 94 major companies surveyed does not plan to use generative AI. According to the survey, 83% of companies are aiming to reduce working hours, 67% are aiming to improve productivity and increase sales, and 63% are aiming to reduce sales and management costs and labor costs.

It is also reported that the trend of using generative AI for cyber defense is also becoming more pronounced in Japan. NEC is using fake emails written by ChatGPT to train employees. ITOCHU Technology Solutions has introduced generative AI from Microsoft Japan in its Security Operations Center to help analyze the types of cyberattacks on companies and provide feedback on countermeasures to those responsible.