Knowledge flows from science to AI technology: Identifying core and brokerage technological roles

Abstract

The rapid advancement of artificial intelligence (AI) technologies has not only driven convergence with diverse technological domains but also swiftly spread across various industrial sectors. As a knowledge-intensive field, AI is particularly shaped by the flow of knowledge from scientific research to technological development, yet remains insufficiently examined in a systematic and structural way. This study addresses this gap by investigating science-to-technology knowledge flow that underpins AI’s technological evolution. We propose a semantic science-technology exploration framework specifically designed for the AI domain, consisting of the two stages: technology classification and semantic topic exploration. First, AI patents are classified into four categories using centrality measures derived from a CPC co-occurrence network. Then, we extract abstracts from both patents and their cited scientific publications to apply BERTopic modelling and generate topic labels using generative AI. Analyzing AI-related patents filed from 2002 to 2021, we trace key technological trends and elucidate the structural pathways of knowledge flow science to technology. The findings offer practical implications for corporate R&D strategies and innovation policy design in the era of AI.