Japan, renowned for its prowess in automobiles and consumer electronics, is now at a crossroads in the technological landscape, particularly in the realm of artificial intelligence (AI). As the nation grapples with a declining population, Japanese scientists recognize the urgent need to make significant strides in AI and robotics to ensure sustained productivity. While Japan has employed ChatGPT for various governmental tasks, there is growing consensus among experts that the country should embark on creating its own unique version of ChatGPT.
In June, Japan emerged as the third-largest source of traffic to OpenAI’s website, as reported by analytics firm Similarweb. One notable example of its application is in Yokosuka City, where ChatGPT has been instrumental in streamlining administrative processes and fostering better workflow within government operations.
Keisuke Sakaguchi, a natural language processing specialist at Tohoku University in Japan, underscores a critical challenge: “Current public Large Language Models (LLMs), such as GPT, excel in English but often fall short in Japanese due to differences in the alphabet system, limited data, and other factors.”
Why Japan is in dire need of its own ChatGPT becomes evident when considering the linguistic intricacies unique to Japanese culture. Unlike English, which has 26 letters, written Japanese comprises two sets of 48 basic characters along with 2,136 commonly used Chinese characters or kanji. With most kanji having multiple pronunciations and an additional 50,000 rarely used kanji, the complexity of the Japanese language poses a formidable challenge for generic LLMs like GPT-4.
Major technology giants in Japan, such as NEC, Fujitsu, and SoftBank, have committed substantial financial resources to the development of artificial intelligence tools. However, the nation faces a knowledge deficit, trailing behind the United States, China, and the European Union in the race to advance AI algorithms. The situation is exacerbated by projections indicating a shortfall of 789,000 software engineers in Japan by 2030, attributed to an aging population.
According to the IMD World Digital Competitiveness Ranking, Japan currently ranks 28th out of 63 countries in terms of technological knowledge. Furthermore, Japan lacks the formidable AI supercomputers required to train AI tools with vast datasets, which are crucial for AI development. No Japanese private company possesses a “world-class machine” with such capabilities.
Yet, Japan’s lag in AI technology and manpower can be remedied with the allocation of sufficient resources. “Japanese LLMs are certainly improving, but they still lag behind GPT-4,” notes Sam Passaglia, a physicist at the University of Tokyo specializing in Japanese language models. Passaglia and his colleagues have launched Rakuda, a platform used for ranking and benchmarking Japanese AI assistants.
Notably, government supercomputers like Fugaku hold the key to Japan’s pursuit of LLMs, according to Noriyuki Kojima, co-founder of Japanese LLM startup Kotoba Technology. He emphasizes that access to large-scale supercomputers is integral to LLM development, historically serving as a critical bottleneck in the process.
As Japan stands at the precipice of an AI revolution, it must leverage its technological potential and foster innovation to create a ChatGPT tailored to its unique linguistic and cultural landscape. With strategic investments, partnerships, and access to advanced computing resources, Japan can navigate the path towards AI excellence, ensuring its continued relevance on the global technological stage.