Understanding Future AI-Green Technology Directions from Past Technological Trajectories

This paper investigates the "twin transition", where digital and green technologies co-evolve along shared technological trajectories. We construct a Universal Network of Technological Trajectories from all EPO patents filed since 1978, identifying 42,924 trajectories containing both green and digital inventions. We employ a multi-stage methodology combining patent citation network analysis with Large Language Models (LLMs) to interpret and project these trajectories. A case study of 32 AI-green trajectories, analysed using ChatGPT-4, reveals a tripartite thematic network in which future themes such as "AI and Machine Learning Integration" and "IoT and Smart Systems" occupy a central position, surrounded by domain-specific clusters (waste management, emission control, energy, industrial automation). Preliminary results from scaling this analysis to all 42,924 TWIN trajectories using open-source embeddings and local LLMs validate our approach and demonstrate its potential for generating replicable, large-scale technological landscape maps for science and innovation policy.