Fujitsu partners with Cohere to build LLMs for Japanese enterprises

Japanese cloud service and data intelligence firm, Fujjitsu, has formed a strategic alliance with Cohere, a Toronto and San Francisco-based enterprise AI company known for its focus on security and data privacy, to develop and provide secure, cutting-edge generative AI solutions for Japanese enterprises.

This collaboration focuses on building large language models (LLMs) specifically tailored for the Japanese language, empowering businesses to enhance customer and employee experiences, the companies said in a statement.

The model, tentatively named “Takane” is expected to be released in September 2024.

The model, the companies said in the statement, will be integrated into Fujitsu’s Kozuchi AI services, specifically designed for private cloud environments that demand high security, such as financial institutions, government agencies, and research and development units.

“We decided to enter into an agreement with Cohere because we felt that the collaboration with Cohere would accelerate our efforts to create a specialized LLM that meets corporate needs,” said a Fujitsu spokesperson. “Cohere sees Japan as a promising market for enterprise LLMs, and was looking for a partner, who has a strong presence in the Japanese market, to jointly develop a Japanese-enhanced LLM.”

As part of the partnership, Fujitsu will become the exclusive provider of jointly developed services on the global market. The Japanese firm plans to provide the jointly developed AI technology to customers through Fujitsu Data Intelligence PaaS, a cloud-based all-in-one operation platform, and Fujitsu Uvance, a cross-industry business model to solve social issues.

“Combining our expertise with Cohere’s secure LLMs, we aim to provide businesses with powerful and adaptable AI solutions that address specific needs and accelerate the adoption of generative AI globally,” Vivek Mahajan, corporate vice president, CTO, and CPO at Fujitsu said in a statement.

“We expect business-specific LLMs based on Takane will be a focal point for aggregating the collective wisdom of Japan. By integrating this LLM into each of Uvance’s offerings, we anticipate it will address resource shortages and knowledge gaps, ultimately contributing to solving societal issues,” the Fujitsu spokesperson said.

The strategic enterprise push

The partnership between Fujitsu and Cohere is not just a technological collaboration; it is a strategic move that underscores the shifting dynamics in the AI market. By focusing on creating LLMs that cater to specific linguistic and security needs, Fujitsu aims to consolidate its position in the Japanese market.

“This partnership combines Fujitsu’s Japanese language expertise with Cohere’s advanced LLM technology,” said Charlie Dai, VP and lead analyst at Forrester. “It will speed up the development of more innovative solutions for enterprises in target market segments, augmenting the competitiveness of both vendors in the meantime.”

Enterprises are the core focus of this partnership’s offerings.

“We plan to launch Takane as one of the options within the Kozuchi Generative AI lineup in September. The primary target customers are those in sectors such as finance, government agencies, and R&D, where high security is paramount. We intend to cater to these customers, including offering for private environments, such as private clouds,” the spokesperson added.

This move is particularly significant given the increasing demand for AI applications that can handle industry-specific requirements while ensuring data privacy and security.

The partnership, Dai feels, needs to address a range of technical and regulatory challenges, such as “security and compliance in hybrid cloud environments, resource constraints towards scalability and performance, model adaptability for various business needs, and the AI readiness of data.”

“Combining our knowledge graph extended RAG technology and generative AI amalgamation technology with Cohere’s secure enterprise LLMs, we aim to provide businesses with powerful AI solutions that address specific needs and accelerate the adoption of generative AI globally,” stated Mahajan in the statement.

Cohere’s LLM, Command R+, forms the backbone of the Takane model. Known for its retrieval-augmented generation (RAG) capabilities, Command R+ is designed to mitigate the issue of hallucinations in AI models, ensuring accuracy and reliability. This is particularly crucial for enterprises that require precise and dependable AI outputs.

RAG has become a critical area for enterprises to reduce model hallucination and drive business outcome, Dai pointed out. “This partnership will hopefully help enterprises to kill the complexity and accelerate the adoption.”

Market Implications

The enterprise AI market is witnessing a surge in demand for customized AI solutions that can operate within secure and private environments. By developing the Takane model, Fujitsu and Cohere are trying to address this demand by providing a solution that combines high performance with stringent security measures.

This development is expected to drive digital transformation across various industries, promoting the adoption of AI technologies that are tailored to specific business needs.

The Japanese language possesses unique characteristics compared to other languages, such as its honorific expression, the Fujitsu spokesperson added. “We will leverage the knowledge gained from enhancing Japanese language capabilities during the development of Fugaku-LLM to develop Takane.”