TL;DR:
- Japan is lagging behind in generative AI development compared to the US, China, and Europe.
- Shortcomings in deep learning and software development contribute to Japan’s slower progress.
- The country faces a deficit of software engineers and ranks 28th in technological knowledge.
- Hardware challenges arise as Japan lacks world-class AI supercomputers for LLM training.
- Government-controlled supercomputers like Fugaku are crucial to Japan’s pursuit of LLMs.
- Japanese institutions are partnering with Fugaku’s developers to advance LLM technology.
- SoftBank, NTT, and other Japanese companies are investing in generative AI.
- Open-sourced LLMs and collaboration with previous pioneers can aid Japan’s progress.
- The competition in generative AI will be intense and require substantial investments.
- The Japanese government is positive about AI adoption and considering regulations.
- Clear guidelines and regulations are necessary to ensure the responsible use of generative AI.
Main AI News:
Generative artificial intelligence (AI) has taken the tech world by storm, and countries are actively vying to develop their own algorithms in this field. However, when it comes to generative AI, Japan, known for its high-tech prowess, is currently lagging behind. This article explores the reasons behind Japan’s slower progress in generative AI and discusses the potential strategies the country can employ to catch up and establish its own large language models (LLMs).
Generative AI has emerged as one of the most prominent trends in technology since OpenAI’s revolutionary chatbot, ChatGPT, captured global attention. According to Goldman Sachs research, breakthroughs in generative AI have the potential to drive a staggering 7% increase in global GDP, equating to nearly $7 trillion, over the next decade.
Key to the development of generative AI is large language models, which serve as the foundation for advanced AI systems like ChatGPT and Baidu’s Ernie Bot. These models possess the ability to process vast amounts of data, enabling them to generate text and other types of content. Unfortunately, Japan is currently trailing behind the United States, China, and the European Union in the race to develop these crucial algorithms, as highlighted by Noriyuki Kojima, the co-founder of Japanese LLM startup Kotoba Technology.
Research conducted by a consortium of state-run institutes revealed that Chinese organizations, including tech giants Alibaba and Tencent, have domestically launched a staggering 79 LLMs over the past three years. Meanwhile, US corporate powerhouses such as OpenAI, Microsoft, Google, and Meta are playing a pivotal role in propelling the country’s advancements in LLM technology, as pointed out by Kojima.
So, what factors contribute to Japan’s lag in generative AI development? According to Kojima, the nation’s trailing position in this field can be primarily attributed to its comparative shortcomings in deep learning and extensive software development. Deep learning, a crucial component of generative AI, requires a robust community of software engineers to develop the necessary infrastructure and applications. Unfortunately, Japan is facing a deficit of 789,000 software engineers projected by 2030, as reported by the Ministry of Economy Trade and Industry. In addition, Japan currently ranks 28th out of 63 countries in terms of technological knowledge, according to the IMD World Digital Competitiveness Ranking.
Apart from the software challenges, Japan also faces hardware limitations in the development of LLMs. These models necessitate training using AI supercomputers, such as IBM’s Vela and Microsoft’s Azure-hosted system. However, no private company in Japan possesses its own world-class machine with these capabilities, as reported by Nikkei Asia.
The solution to Japan’s generative AI lag lies in government-controlled supercomputers, such as Fugaku. These supercomputers hold the key to Japan’s pursuit of LLMs, as accessing large-scale computing power has traditionally been a significant bottleneck in the development process, as explained by Kojima.
Recognizing this potential, the Tokyo Institute of Technology and Tohoku University have partnered with Fugaku’s developers, Fujitsu and Riken, to utilize the supercomputer in the development of LLMs primarily based on Japanese data. The organizations aim to publish their research results in 2024, contributing to the advancement of LLM technology among Japanese researchers and engineers. Furthermore, the Japanese government plans to invest 6.8 billion yen ($48.2 million), half of the total cost, to construct a new supercomputer in Hokkaido, specialized in LLM training, which is expected to commence service as early as next year.
In a show of support for generative AI technology, Japanese Prime Minister Fumio Kishida expressed his backing for industrial applications of AI after meeting with OpenAI CEO Sam Altman. OpenAI itself is considering establishing an office in Japan, signaling the country’s increasing focus on generative AI.
Not only is the government taking steps to bolster Japan’s standing in generative AI, but big tech players are also joining the race. SoftBank’s mobile arm recently announced its plans to develop its own generative AI platform, with CEO Masayoshi Son stating the company’s ambition to lead the AI revolution. SoftBank’s shift in focus toward AI, highlighted by its intention to list its chip design company Arm in the US IPO market, is aimed at accumulating funds and resources for AI endeavors.
Additionally, NTT, the Japanese telecommunications company, has revealed its plans to develop its own LLM this fiscal year, focusing on creating a lightweight and efficient service for corporations. NTT intends to invest 8 trillion yen over the next five years, a 50% increase from its previous level of investment, in growth areas like data centers and AI.
While Japan may be behind in generative AI at present, it is making its initial strides through these private sector efforts. Establishing a robust infrastructure is crucial, and once accomplished, the remaining technical challenges can be significantly mitigated by utilizing open-sourced software and data from previous pioneers, according to Kojima. Notable open-sourced LLMs, such as Bloom, Falcon, and RedPajama, offer vast datasets that can be downloaded and studied to aid further progress.
However, companies venturing into the generative AI space must be prepared for competition that spans a relatively long timeframe. Developing LLMs requires substantial capital investment and a highly skilled workforce proficient in natural language processing and high-performance computing. Therefore, it will take time for SoftBank, NTT, and other players to make a substantial impact on the AI landscape, cautions Kojima.
As Japanese tech companies actively engage in generative AI development, the country is also adopting a positive stance on AI adoption in various sectors. A survey conducted by Teikoku Databank revealed that over 60% of companies in Japan have a positive attitude toward using generative AI in their operations, with 9.1% already implementing it.
Recognizing the need to address risks associated with generative AI, companies like Hitachi have established specialized centers to promote safe and effective usage of the technology. Hitachi’s generative AI center, staffed with data scientists, AI researchers, and relevant specialists, aims to formulate guidelines for mitigating risks associated with generative AI applications.
The Japanese government is also considering the adoption of AI technologies like ChatGPT, provided that cybersecurity and privacy concerns are adequately addressed. Chief Cabinet Secretary Hirokazu Matsuno emphasized the importance of establishing guidelines and regulations to ensure the responsible and secure use of generative AI. Hiroki Habuka, a research professor at Kyoto University’s Graduate School of Law, suggests that the government should facilitate soft guidelines and evaluate the need for stricter regulations based on specific risks, thus ensuring proper governance and alignment with societal values.
Conclusion:
Japan’s current lag in generative AI development presents both challenges and opportunities for the market. While the country faces shortcomings in deep learning, software development, and hardware infrastructure, efforts are being made to leverage government-controlled supercomputers and foster collaborations. Japanese companies, including tech giants like SoftBank and NTT, are investing in generative AI, aiming to improve Japan’s position. However, catching up and establishing a strong presence in the generative AI market will require significant investments, a skilled workforce, and the development of clear guidelines and regulations to ensure responsible and secure usage. The market will witness intense competition, and companies must strategically navigate the evolving landscape to succeed.
