From Jet Tagging to Autonomous Experiments: AI and the Future of Fundamental Physics

High energy physics has been at the forefront of AI adoption for decades, from early boosted decision trees and domain-specific innovations like neural network cluster splitting, through the deep learning revolution in jet tagging and flavor identification, all of which have transformed how we extract physics from LHC data. Today, techniques like foundation models, generative simulation, and transformer-based reconstruction are reshaping the analysis pipeline end to end.

Looking forward, agentic AI promises to extend this transformation into the experiments themselves: autonomous detector control, self-optimizing data acquisition, and AI collaborators that automate analysis and translate measurements into theoretical insight. In this colloquium I will trace the arc from early classifiers to the autonomous experiments of the coming decade, and sketch a possible vision of an AI-native future for fundamental physics. I will also discuss problems where AI has so far fallen short