Fujitsu has said its new Japanese-language large-language-model (LLM) is now available to global enterprise customers. Developed with Canadian data security and privacy outfit Cohere, the new product, called Takane, represents a “leap forward in generative AI capabilities” and offers the “highest Japanese language proficiency in the world”, the firm reckons. It comes bundled with knowledge graphs and retrieval augmented generation (RAG) techniques.
The Takane product is integrated into the firm’s so-called Kozuchi AI suite, which launched a year ago as a horizontal platform of AI technology services, but has since morphed into a broader AI development and consultancy proposition; it is being offered on a subscription, via its “all-in-one” ‘data intelligence’ platform as-a-service (DI PaaS), which forms part of its Uvance consultancy portfolio of vertical solutions “for “cross-industry societal challenges”).
Its whole AI shtick is based around a new AI “framework”, offered with its Kozuchi / Uvance service, which comprises three key AI tools: knowledge-graph extended RAG software, which uses knowledge graphs to link company data to generative AI outputs; amalgamation technology, which selects highest-performing LLMs based on input tasks; and a “world-first” auditing technology, which enables explainable output in line with company regulations.
Takane is based on Cohere’s Command R+ LLM, described as a “multilingual model trained on proprietary data from scratch”. It features enhanced RAG to mitigate hallucinations. It outperforms other LLMs on a Japanese LLM benchmark (called JGLUE), including compared with large-scale general-purpose LLMs from cloud service providers. It is trained for use in sensitive industries that require high accuracy and reliability from Japanese-based LLMs.
Fujitsu has made a ‘significant investment’ into the partnership with Cohere, which features a two-way global reseller arrangement, as well. For more on RAG technology, go here; for more on Fujitsu’s AI strategy, go here. Takane is classified as a medium-sized LLM, which can be fine-tuned and customised using proprietary company data. Fujitsu is targeting the government, finance, healthcare, and legal sectors, in particular – where “even minor language errors can have serious consequences”.
Fujitsu explained: “General-purpose LLMs are often delivered through public cloud services and are therefore not suitable for use in industries handling confidential data and for tasks requiring compliance with laws, regulations, and industry rules. Furthermore, characteristics of the Japanese language, including the mixed use of multiple character types, omitted subjects, and honorific expressions also pose significant hurdles.
Vivek Mahajan, corporate vice president and chief technology officer at Fujitsu, said: “Takane will empower companies in industries that demand the highest level of security to harness the power of generative AI. Fujitsu is committed to creating new value for businesses by developing cutting-edge AI technologies. This commitment fuels our three growth drivers: modernisation, Fujitsu Uvance, and consulting – which are all powered by our innovative technology.”
Aidan Gomez, co-founder and chief executive at Cohere, said: “We are very excited to bring Takane’s advanced Japanese LLMs to global enterprises. Our partnership with Fujitsu accelerates AI adoption in this critically important market by offering secure, performant AI designed specifically for business use across Japanese and other languages.”
The press note quoted Mizuho Financial Group, Mitsubishi Electric Corporation, and Deloitte Tohmatsu Consulting as customers.