Problem

Experimental particle physics aims to understand the universe through highly complex and large-scale facilities that generate petabytes of data annually. However, a significant portion of this raw data is discarded due to bandwidth, storage, and latency limitations. This restricts discovery potential, reduces efficiency, and makes long-term operations challenging. The current approaches also lead to labor-intensive operations, slow analysis cycles taking years, and difficulties in retaining rare or unexpected signals. The field is at a pivotal moment where existing and future experiments require transformative breakthroughs to maximize scientific output and address grand questions about the universe.

Method

The proposed solution is an "AI-Native" research ecosystem, embedding AI end-to-end across the experimental lifecycle. This vision is structured around four Grand Challenges: 1) Accelerated Experimental Design, which uses differentiable simulations, agent-guided optimization, and active learning to co-design accelerators and detectors, reducing technical risk, cost, and time-to-build for ambitious experiments. 2) Intelligent Sensing & Instrumentation, which integrates AI upstream through trigger-less readout, AI-assisted data acquisition, physics-aware compression, and real-time inference to preserve rare or time-critical signals within bandwidth and storage constraints. 3) Autonomous Experiments, transforming operations into proactive, resilient, and continuously calibrated systems using AI-driven monitoring, diagnosis, and decision support, aiming to capture high-quality data with less downtime and preserve institutional knowledge. 4) From Data to Discovery, which integrates foundation models, fast AI-enabled reconstruction and simulation, and agent-orchestrated workflows to drastically shorten analysis cycles and open new theoretical exploration areas. To realize this, a national-scale collaboration is proposed, involving DOE national laboratories, U.S. universities, and industry partners. This collaboration will develop shared AI capabilities, leverage national cyberinfrastructure (like DOE Leadership Class Facilities and the American Science Cloud), and foster workforce development to train AI-literate scientists.

Results

While this paper presents a vision rather than experimental results, the implementation of an AI-native ecosystem is anticipated to yield significant advancements. Expected outcomes include a dramatic acceleration of scientific discovery, with analysis cycles potentially reduced by factors of 100-1000. Experiments would gain the ability to extract and retain more information from vast datasets, extending discovery potential to subtle or unexpected phenomena. The efficiency and sustainability of long-running facilities would be greatly improved, with goals like 50% less downtime and 10 times faster calibration cycles. The approach is also expected to enable exploration of orders of magnitude larger scientific landscapes, probing entire spaces of theoretical possibilities rather than just a few benchmarks.

Implications

The AI-native research ecosystem has profound implications for the future of particle physics and broader science. It would position the U.S. as a global leader in AI-powered fundamental science by strengthening connections between labs, universities, and industry. The proposed national-scale collaboration would significantly contribute to national workforce development goals, potentially training thousands of PhDs and tens of thousands of undergraduates with dual physics/AI competencies. This initiative will not only maximize the scientific return on current investments in facilities like HL-LHC and DUNE but also enable next-generation experiments to be conceived and operated with AI as a core capability, transforming them into continuously learning systems. Ultimately, it promises to fundamentally change how particle physics research is conducted, shifting human effort from mechanical tasks to scientific interpretation and enabling unprecedented exploration of the universe's fundamental questions.