軟銀與 OpenAI：在日本建立人工智慧網絡
・請求日本500家主要企業參與
・一次性發布20家先進企業的生成式AI使用實例
我們為您帶來 Nikkei Crosstech 發表的文章摘要。
SoftBank Group: (SBG):
Working with US OpenAI to develop AI infrastructure in Japan.
1. Building data centers for AI development across Japan
2. Plan to build power generation facilities to cover electricity demand
This is the Japanese version of the “AI infrastructure investment that CEO Son announced to US President Trump.”
Meeting with 500 Japanese companies:
On February 3, SoftBank Group and OpenAI will hold a meeting in Tokyo with over 500 Japanese companies.
Many major Japanese companies from transportation, pharmaceuticals, finance, manufacturing, logistics, etc. will participate.
In 2023, the generative AI boom first ignited.
Microsoft’s Azure OpenAI: Early stage
Generative AI services such as “Azure OpenAI Service” have appeared.
1. A number of companies were deploying chat environments using generative AI.
2. Many companies were just trying to introduce it.
Introduction of generative AI systems: Development stage
Two years later.
The introduction of generative AI systems has spread beyond general-purpose applications to marketing and call centers.
Many companies, including Kirin Holdings (HD) and Fujifilm HD, have begun full-scale introduction.
We looked into the efforts of 20 leading companies to find know-how for improving and enhancing business efficiency using generative AI.
Seven & i HD: Newsletter creation:
Digital Innovation Department: Miyaji Takashi
1. Reviewed marketing operations with an eye toward improving business efficiency.
2. Identified all tasks for each business process.
Reconstructed business processes with generative AI:
1. Decided on a policy of using generative AI to create newsletters with high open rates.
2. Achieved an 84% reduction in outsourcing costs.
3. Store support and customer experience improvement analysis project
Sompo Japan Insurance Inc.: Insurance underwriting:
1. In-house developed generative AI to support “inquiries regarding insurance underwriting and regulations”.
2. Generates answers by referring to internal documents such as “manuals and Q&A data”.
10,000 uses: Achieved in 3 months
1. Deployed to 8 sales branches and the head office product department from October 2024
2. As of January 10, 2025, the number of uses exceeded 10,000.
All at once, the efforts of 20 leading companies:
1. Seven & i: “Outsourcing costs reduced by 84%” by using generative AI (February 3, 2025)
2. Sumitomo Corporation: “Save 1.2 billion yen per year with MS 365 Copilot” Use of generative AI involving the entire company (scheduled to be released February 5, 2025)
3. Mitsubishi Electric: “Summarizing diagram-included design documents with generative AI” The key to conquering RAG is data preparation and prompts (scheduled for release on 2025.02.07)
4. Panasonic: “Using generative AI in factories” Aiming to install it in home appliances (scheduled for release on 2025.02.10)
5. Mitsui & Co.: “Improved accuracy by combining discriminative models and LLM” Symbiotic system with generative AI (scheduled for release on 2025.02.12)
https://xtech.nikkei.com/atcl/nxt/column/18/03085/013000001/