Fujitsu Generative AI : Fujitsu Global

Fujitsu Generative AI provides a secure environment for enterprises to utilize generative AI capabilities, incorporating measures to mitigate the risk of "hallucinations." This solution facilitates the application of Large Language Models (*1 LLMs) and Retrieval Augmented Generation (*2 RAG) across a wide range of use cases. Specifically, the integration of Fujitsu Generative AI for Cohere, a large-scale enterprise-grade language model, allows for the development and deployment of industry- and business-specific LLMs to support clients in their business transformation initiatives.

*1 Large Language Model *2 Retrieval-Augmented Generation

What is "Fujitsu Generative AI"?
Key Benefits
Functions
Use Case
Fujitsu Generative AI for Cohere
Menu Structure
Notice

What is "Fujitsu Generative AI"? [Service Overview]

Fujitsu Generative AI facilitates the interaction between natural language, atypical data formats, and computer systems to improve human productivity and creativity. The service delivers secure, client-specific solutions via customizable chat applications and comprehensive APIs, integrating both proprietary data and external knowledge bases.

Key Benefits

Fujitsu Generative AI provides a comprehensive suite of tools for generative AI, making it suitable for rapid prototyping (PoC) and short-term deployments. Its intuitive interface and pre-built functions allow for immediate implementation. Furthermore, the platform supports a wide range of use cases through configurable workflows (*3 Flow configuration) and readily available APIs. *3 Process flow for AI to sequence tasks

Ready-to-Use RAG

Get started instantly with our pre-built RAG environment
Cutting-Edge Technology

Our platform supports multimodal AI, integrating analytical AI capabilities, specialized large language models (LLMs), and other state-of-the-art technologies.
Customizable Workflows and APIs:

Users can easily customize their generative AI experience through flexible Flow settings and APIs.

Functions

Fujitsu Generative AI provides a comprehensive, integrated solution for secure generative AI functionality.

* Model selection for LLM is also possible

①Interactive interface
- ChatGPT equivalent interactive interface
- Chat history and multiple chats
②RAG (ability to generate responses based on reference material)
- Provides out-of-the-box RAG functionality. Generates an answer based on the data referenced
- Easily import and delete data via Web GUI
- Web site data can be imported by specifying URL
- Data can be grouped by conversation content
- Assist in assessing whether captured data has been referenced for purpose
③Flow (Defined/Custom API)
- Multi-step processing (prompts, data conversion and reading) can be configured for each use case. It is possible to obtain highly accurate answers with high reproducibility in various use cases.
- The general Flow (Translation, summary, design specification review, etc.) is provided as a template (Defined API) in advance and requires no configuration work.

④Form
- Input templates can be set for each flow (use case) so that users do not get lost
⑤Hallucination detection function
- [Fujitsu proprietary technology] Hallucination countermeasure function that scores the degree of hallucination in output answers

Use Case

A system is implemented to automate the review of system design specification, identifying potential omissions through strategically placed checkpoints.

Fujitsu Generative AI for Cohere

Fujitsu provides "Takane," a product jointly developed with Cohere Inc., through its "Fujitsu Generative AI for Cohere" service.

Menu Structure