Like many other AI scientists, Sakana AI is pursuing technologies that can validate, trust and help scientific research. But where Tokyo-based work stands outside the country’s model developers is the focus on specific products of Japanese society, language and industry.
Investors are excited about the idea. Sakana announced today (Nov. 17) that it raised nearly $135 million in Series B funding, valuing the company at $2.6 billion, making it one of the most important AI startups in Japan. Backers include the MITITSUBISHI UFJ group, Khosla Ventures, Nea and Lux Capital, among others.
Sakana was founded in 2023 by a group of former Google and AI researchers, including Deo David Ha. The company’s young products so far include consulting models, a self-improvement agent and “AI Scientist,” a program designed for scientific discovery.
The startup also works with major Japanese institutions to develop customized products for the financial industry. Its clients include MUFG, Japan’s largest bank, and the Daiwa Securities group. Sakana is now looking for increased defense, intelligence and productivity.
“As a business leader myself, I feel a responsibility to lead the implementation of AI in all Japanese companies,” said Hinori Kamezawa, CEO of MUFG, in a statement. “Through this investment method, I hope that the benefits of AI will extend beyond the use of banking services to Japan’s diversified industry.”
Building a dynasty ai
Expansion strategies and possible investments, collaborations and possible acquisitions, Sakana will use its new capital to further research on large-scale linguistic models (LLMS) that reflect Japanese national culture. “This allows us to develop and optimize the Frontier models for the specific needs of Japan, providing a unique approach and competitive advantage for the major competition,” the company said in its financial announcement.
More than half of all content on the Internet today is written in English, which poses a challenge for training LLMS serving non-English speaking markets. Less than 5 percent of the world’s 5 most widely spoken languages have meaningful online presences, according to a paper from Stanford in a STANFFORD LLMS issue called “the digital divide.”
Sakana is not alone in trying to close this gap. Krutrim, an AI startup launched by Ola Co-Founder Bhavish Aggarwal, is building LLMS in Indian languages, while France’s AI is pitching its le le chaxoya assistant. Earlier this year, Microsoft announced that it has launched efforts, including a call for proposals to increase training data in European languages, to further develop multilingual AI.
Despite the limited number of AI systems being developed in Japan, Sakana argues the needs and benefits are clear. “For Japan, with a shrinking workforce and an aging population, the benefits of AI technology are clear,” the startup said. “We believe it is the right way for Japan, a country with limited resources, to develop our AI Ecosystem sustainably through innovation and strong collaboration in the social application of technology.”