Since OpenAI created a new wave with its ChatGPT chatbot, generative AI has become the most talked-about topic in the tech world. According to Goldman Sachs' predictions, new AI breakthroughs could drive global GDP growth by 7%, equivalent to nearly $7 trillion, over the next decade.
The key to developing generative AI lies in large-scale language models (LLMs), capable of processing massive datasets to generate text and images. Noriyuki Kojima, co-founder of the LLM startup Kotoba Technology, says Japan currently lags behind the US, China, and Europe in developing these algorithms.
According to CNBC, companies in China, including Alibaba and Tencent, have launched at least 79 LLMs in the past three years. Major US tech corporations such as Microsoft, Google, and Meta also play a significant role in driving LLM development.
However, Japan is lagging behind the US, China, and Europe in terms of the scale and speed of LLM development. Kojima stated that the reason for Japan's delay in generating AI technology largely stems from shortcomings in the fields of deep learning and software development.
Deep learning requires a highly skilled community of software engineers to develop the necessary infrastructure and applications. According to Japan's Ministry of Economy , Trade and Industry, the country will face a shortage of 789,000 software engineers by 2030. Based on the IMD Digital Competitiveness Ranking, Japan is currently ranked 28th out of 63 countries in the digital knowledge category.
According to Nikkei Asia, Japan also faces hardware challenges because LLM needs to be trained using AI supercomputers like IBM's Vela or Microsoft's Azure storage system. However, no private company in Japan possesses a world-class supercomputer with similar capabilities.
Kojima believes that government- controlled supercomputers like Fugaku will "hold the key" to Japan's pursuit of LLM. He stated that access to these supercomputers will become the backbone of LLM development.
The Tokyo Institute of Technology and Tohoku University are planning to collaborate with supercomputer developers Fujitsu and Riken to use Fugaku to develop a Low-Level Programming (LLM) system based on Japanese data. They expect to publish research results in 2024 to help other Japanese researchers and engineers develop LLMs.
The Japanese government will also invest $48.2 million to build a new supercomputer in Hokkaido, expected to begin operation in early 2024. This machine will specialize in training LLMs to boost the development of generative AI in Japan. In April, Japanese Prime Minister Fumio Kishida expressed support for the use of generative AI technology in industry.
Japanese companies are chasing AI for proliferation.
Tech giants have also joined the competition to enhance Japan's position in the AI field. In June, SoftBank's mobile division announced plans to develop its own AI generative platform. SoftBank CEO Masayoshi Son emphasized the investment firm's plan to shift from a "defensive" to an "offensive" approach and increase its focus on AI. Masayoshi stated that SoftBank aims to be a leader in the AI revolution.
SoftBank Group has sold 85% of its stake in SB Energy and agreed to sell 90% of its stake in Fortress Investment Group. Cutting other investments will help SoftBank free up cash and inject it into AI through its venture capital firm, Vision Fund.
Despite investing $140 billion in AI, Softbank is still lagging behind in this race.
SoftBank-owned chip design company Arm is also preparing for an IPO in the US later this year. Amir Anvarzadeh, a Japanese equity market strategist at Asymmetric Advisors, said it would be the largest IPO in the world.
Arm initially sought to raise only $8-10 billion. However, with the soaring demand for semiconductor chips, Anvarzadeh believes Arm could raise $50-60 billion, equivalent to 85% of SoftBank's market capitalization.
While SoftBank's stock price may rise, this doesn't guarantee the success of its AI investment efforts. Anvarzadeh argues that SoftBank is not the savior in Japan's AI development race.
Japanese telecommunications company NTT has also announced plans to develop its own LLM to create a “lean and efficient” service for corporations. In May, digital advertising company CyberAgent released an LLM that allows companies to create AI chatbot tools, one of the few models specializing in Japanese language and culture.
Although Japan is yet to catch up in the AI race, it is making initial strides thanks to contributions from businesses. Developing LLMs requires a highly skilled workforce and significant investment. Kojima says that once a robust infrastructure is established, the remaining technical challenges can be significantly mitigated by utilizing open-source software and data from pioneers.
However, companies entering this field should anticipate long-term competition. SoftBank and NTT's involvement in AI development will not change the current situation in the short term, Kojima warned.
AI regulations in Japan
A survey by Teikoku Databank shows that over 60% of companies in Japan have a positive attitude towards using AI for generation, while 9.1% are currently applying AI to their operations.
Hitachi Corporation has established a generative AI center in an effort to promote the safe and effective use of the technology by its employees. With the participation of data scientists, researchers, and AI experts, the center will develop guidelines to mitigate the risks that generative AI may pose.
Japanese Chief Cabinet Secretary Hirokazu Matsuno said the government would consider applying AI technology like ChatGPT if cybersecurity and privacy concerns were resolved.
Professor Hiroki Habuka of Kyoto University Law School argues that the Japanese government should develop and facilitate soft rules for the use of generative AI, while also considering and implementing strict regulations to protect against potential dangers from AI.
Source link
Comment (0)