Description
As AI assistants increasingly shape how people access and understand information, official statistics face a new usability test: can authoritative public data be found, understood, sourced and used correctly by both humans and machines?
In Episode 2 of Statistics in the AI Era, OECD Chief Statistician Steve MacFeely speaks with Prem Ramaswami, Head of Google Data Commons, about the practical work needed to make trusted statistics understandable to people and interpreted by AI systems.
Prem argues that the challenge is not only to publish reliable data, but to make it easier to discover, connect and use. He explains why fragmented datasets impose a hidden “data tax” on analysts, why AI systems need data that is well sourced, auditable and reproducible, and why statistical organisations should start with machine-readable metadata.
The conversation also explores practical AI use cases across the statistical pipeline – including data collection, data cleaning, sharing and tailored policy outputs – as well as the need for public-private partnerships that work both ways: statistical organisations producing high-quality data with clear metadata, and technology companies building tools that can find, prioritise and properly use authoritative public sources.
Host: Steve MacFeely
Guest: Prem Ramaswami
