Fujitsu Limited has announced the development of a groundbreaking generative AI framework designed to meet the evolving needs of enterprises.
This innovative technology aims to enhance the use of generative AI within companies, ensuring easy compliance with extensive data and regulatory requirements. The enterprise-wide generative AI framework, part of the Fujitsu Kozuchi lineup, will be available globally starting July 2024.
Addressing Barriers to Enterprise AI
The enterprise environment has historically faced challenges in adopting generative AI models due to large data scale requirements, cost constraints, response speed limitations, and the necessity to adhere to corporate rules and regulations. Fujitsu’s new framework is designed to overcome these barriers, providing specialized AI solutions for diverse business needs.
Key Components of the Generative AI Framework
Knowledge Graph Extended Retrieval-Augmented Generation (RAG):
This technology enhances the capabilities of existing RAG by utilizing knowledge graphs to link large-scale data, allowing AI models to reference over ten million tokens. This advancement ensures that generative AI can make logical inferences based on comprehensive data inputs, such as laws, company regulations, manuals, and videos.
Generative AI Amalgamation Technology:
Fujitsu’s unique amalgamation technology automatically generates high-performance AI models tailored to specific business tasks without the need for prompt engineering or fine-tuning. By combining the best-suited models for each task, this technology ensures efficient and sustainable AI operations, reducing electricity and computational resource consumption.
Generative AI Auditing Technology:
As the world's first generative AI auditing technology, this system ensures compliance with corporate and legal regulations. It provides explainable outputs by presenting the basis for AI responses and verifying their consistency. This technology supports both text and multimodal inputs, such as knowledge graphs and images, ensuring reliable AI utilization.
Practical Applications and Future Plans
Fujitsu is currently conducting verification tests with its generative AI framework, expecting significant improvements in various business areas, including:
- A 30% reduction in man-hours for contract compliance verifications
- A 25% increase in support desk efficiency
- A 95% reduction in time required for optimal driver allocation planning in transportation
Additionally, the framework is being tested for its ability to create quality assurance materials, analyze mobile network connectivity issues, assess employee fatigue levels, and interpret large-scale genome data.
Looking Ahead
Fujitsu plans to expand its lineup of enterprise-specific AI models for various applications, including Japanese language processing and code generation. The company’s proposal for developing a lightweight knowledge graph generation LLM has been accepted by GENIAC, a project by Japan’s Ministry of Economy, Trade and Industry. This initiative aims to strengthen generative AI development capabilities in Japan, enabling secure on-premise use of the new knowledge graph extended RAG technology.
Fujitsu remains committed to advancing research and development in generative AI technologies, addressing diverse customer needs, and providing robust support for the integration of AI in business operations.
For more details on the generative AI framework and Fujitsu's advancements, visit the official press release: https://www.fujitsu.com/global/about/resources/news/press-releases/2024/0604-01.html).
