Home / Technology / UN & Google: AI Meets Global Data
UN & Google: AI Meets Global Data
18 Sep
Summary
- UN and Google partner to enhance AI access to global statistics.
- New system supports natural-language queries for UN data.
- AI models show low accuracy in retrieving development indicators.
The United Nations is collaborating with Google to transform access to its extensive global statistics. The newly announced UN System Data Commons, built on Google's Data Commons platform, allows users and AI systems to query vast datasets from UN agencies using natural language. This system replaces the traditional UNData portal, offering a more dynamic and AI-friendly interface.
This development is critical as AI tools increasingly serve as information sources, yet often struggle with data accuracy. A recent UNICEF benchmark indicated that several leading large language models achieved an average accuracy score of merely 21.2% when responding to queries about global development indicators. The study also noted that models frequently hedged answers or provided inconsistent numerical data across repeated queries.
UNICEF's data website has experienced a significant surge in traffic referred from generative AI assistants, accounting for an estimated one in ten visits. To address these challenges, the UN System Data Commons aims to integrate data from 80% of UN entities by 2027, with 26 entities already committed and data from nearly 20 available at launch. Google.org provided initial funding and technical support, with the platform intended for independent UN operation. The system also ensures data traceability, allowing users to verify AI-retrieved statistics against original UN sources.
