The use of "AI" in particle physics is not new. In 1999 they were using neural nets to compute various results. Here's one from Measurement of the top quark pair production cross section in p¯p collisions using multijet final states [https://repository.ias.ac.in/36977/1/36977.pdf]
"The analysis has been optimized using neural networks to achieve the
smallest expected fractional uncertainty on the t¯t production cross section"
I remember back in 1995 or so being in a professor's office at Indiana University and he was talking about trying to figure out how to use Neural Networks to automatically track particle trails in bubble chamber results. He was part of a project at CERN at the time. So, yeah, they've been using NNs for quite awhile. :-)
Particle identification using NN classifiers was actually on the early success stories of NN. These are pretty standard algorithms in tracking and trigger software in HEP experiments now. There are even standard tools in the field to help you train your own.
What is more interesting currently is things like anomaly detection using ML/NN and foundational models..etc.
"The analysis has been optimized using neural networks to achieve the smallest expected fractional uncertainty on the t¯t production cross section"