The goal is to find e.g. Lagrangian, which consequences are in agreement with nature ... but calculating these consequences is quite tough - AI could generate required simulations, and it is already starting.

The question is where to search for such e.g. Lagrangian? There are already many people having own models developed for decades - AI could help to objectively verify with simulations - building kind of arena for models, to select the ones in the best agreement with nature, identify their best features.

After checking the already available physics models from humans, maybe succeeding would be also AI generated - e.g. combining what was the most successful in the tested models.