Nomura Investment Forum: Using Generative AI in Japan’s Financial Sector

The emergence of generative AI (artificial intelligence) represents a springboard for economic growth and digital transformation for Japan in particular, with the country boasting one of the most receptive environments for AI development, said Kangsoo Kim, Special Research Fellow at Financial Services Agency, and Board Director of Special Situations Group at Matsuo Institute.

Thanks to favorable factors from improvements in machine processing power, a growing body of data, advances in deep learning, to the arrival and practical application of generative AI, AI technology is undergoing another boom, said Kim during his presentation at Nomura Investment Forum on December 5 in Tokyo.

Among globally adopted examples of such technology is Open AI’s ChatGPT. Launched in late 2022, ChatGPT’s user base reached 1 million within one week of release, and 100 million within two months. It is also capable of more advanced semantic understanding and conversations than previous large language models (LLMs), Kim added, with the potential of being adopted in a wide range of fields, including text processing and generation, and programming, by giving appropriate prompts.

While OpenAI’s latest GPT-4o is by far the most accurate large language model, capable of faster and more precise generation, other technology giants are ramping up investments in generative AI: Google announced in May 2024 that it would incorporate its LLM Gemini into its services, with Meta releasing open-source software Llama3, allowing the model to be continuously trained on its own data and language.

As tech giants continue to drive further innovation in generative AI and integrate such technology into their services, the Japanese government’s receptive stance has drawn interest from these companies in expanding in the country. OpenAI, for example, set up its first Asia Pacific office in Japan, while Amazon Web Services plans to invest more than 2.2 trillion yen in constructing data centers by 2027.

“In 2023, the Japanese government’s response [to new developments in the space of generative AI] has been as fast as any other country,” said Kim.

“In addition to responding to risks, promoting the use of AI and strengthening AI development capabilities are as important as AI measures in economic policies,” he added.

One of the major steps taken by the government is the launch of its AI Strategy Council, a forum with a remit focused on mitigating the risks of adopting generative AI, the technology’s potential to accelerate digitalization and improve productivity in Japan as a whole, as well as strengthening AI’s capabilities.

Japan’s regulatory support, its relatively low labor costs and ample room for digitalization across corporates, particularly larger ones with exposure to global markets, burnish its appeal to tech giants as a prime location to promote the adoption of AI, said Kim.

Should generative AI become an integral part of Japanese financial corporates, he added that the application of such technology would increase efficiency and automation, bringing down costs, as it could perform analysis of companies, sectors, asset classes, capital markets, and across various geographies. Another potential application of generative AI for financial institutions , Kim said, is combating and strengthening their defense against financial crimes.

“As financial transactions become increasingly digitized, the possibility of countermeasures with AI are enormous. Among them, the efficiency of transaction monitoring could be achieved in an unprecedented way,” Kim said. Given that current cybersecurity measures involve manual analysis and checks, there is a strong need for systems that use LLM as a tool to support security experts - multilingual LLM has been shown to detect phishing attacks using different languages and techniques with high accuracy, facilitating a variety of cybersecurity measures and enhancing corporates’ cyber defense systems.

“On the other hand, financial crimes also use AI, meaning that it’s highly likely that countermeasures and crimes will increasingly be a cat-and-mouse game as technology develops further in the future,” he added.

As the technology of generative AI continues to make rapid progress, senior corporate leaders need hands-on experience using large language models, to accelerate organizational penetration of digitalization, Kim said.

For more on Japan's financial sector, view the session replay here (requires client login).