SoftBank Unveils SB Intuitions: Japan's OpenAI for Developing LLMs and Generative AI in Japanese Language

AI may have found a significant hub in the Bay Area, but there's a rising wave of innovation emerging from other regions as well, reflecting a growing belief that market opportunities exist beyond a single geographic location.

Today, SoftBank has made a strategic move in Japan by launching SB Intuitions, a new venture aimed at "researching and developing homegrown Large Language Models (LLMs) specifically tailored for the Japanese language." In addition to this, SB Intuitions plans to create and distribute generative AI solutions focusing on Japanese contexts.

While details remain limited, here's what we know so far: SB Intuitions will be fully owned by SoftBank, with no current plans for outside investment. The new company will rely exclusively on data from Japan-based data centers, though specifics on data sourcing and related agreements have yet to be disclosed. It's expected that SB Intuitions will leverage SoftBank’s extensive consumer and enterprise networks within Japan.

Initial plans for this initiative were set forth in March of this year, and today's announcement officially reveals the name and vision for SB Intuitions. To support its operations, SoftBank will utilize a robust AI computing platform powered by Nvidia GPUs, designed for developing generative AI and various applications, including academic research and enterprise solutions. This platform is expected to be operational by the end of March 2024.

Hironobu Tamba, a veteran of SoftBank who has recently overseen the company’s AI and IoT division, will lead this new venture. It starts with a paid-in capital of 150 million yen, equivalent to just over $1 million at current exchange rates. However, the total investment commitment for the endeavor remains undisclosed.

For perspective, consider that OpenAI has garnered over $11 billion in funding and holds a valuation nearing $30 billion based on its latest financial updates this year. This underscores the substantial investments required for ambitious companies developing LLMs and generative AI services.

Furthermore, SoftBank has announced plans to issue a new bond worth 120 billion yen (approximately $840 million), with a portion of the proceeds earmarked for AI investments. This could indicate potential financial resources for SB Intuitions in the future.

It’s intriguing, and slightly ironic, for SoftBank to unveil a new LLM venture given its mixed history with AI. The tech giant recognized AI as a pivotal development area long before the current surge in interest. However, evidence of successful in-house services or effective partnerships to bring its AI strategy to life has been scarce.

SoftBank's investment history in AI is equally noteworthy; the company has invested around $140 billion into approximately 400 AI-related projects over the years. Its most notable endeavor has been the Vision Fund, which has surpassed $150 billion in funding aimed at startups, with a significant focus on AI. Yet, despite these extensive investments, the consensus is that SoftBank has been largely excluded from recent AI advancements.

SoftBank's new initiatives underscore both a missed opportunity and a larger existential strategy.

As SoftBank CEO Junichi Miyakawa stated: "Just as search engines, operating systems, and cloud computing have transformed the internet landscape, generative AI will be a groundbreaking technology." He emphasized that while major foreign entities are currently leading the development of LLMs—based predominantly on English and Chinese data—there's a critical need for domestically-produced generative AI tailored to Japanese cultural and business contexts. He noted that Japanese data can better accommodate unique expressions used in sectors like public services and healthcare.

This initiative will not exist in isolation. During the announcement, Tamba highlighted SoftBank’s strategic partnership with Microsoft, which includes creating a secure data environment for enterprises in Japan interested in AI collaborations, including generative AI projects. SoftBank plans to utilize Microsoft’s Copilot service internally as well.

Miyakawa concluded by stating, "We will not only develop our own LLMs but also create a multi-generative AI framework that allows us to select the best model for our clients from options developed by OpenAI, Microsoft, Google, and others."