Why not? It will surely have the same bounding box as our own physics, but the laws may be represented in completely different fashion. They would mostly map to each other, though.
Or maybe not. Maybe the "AI physics" has predictions for phenomena which our physics never bothered to figure out, whereas it would ignore large chunks of our physics area as irrelevant.
This is more like the empirical "laws" that various engineering disciplines use. I.e. some number crunching that has no theoretical foundations whatsoever, but seems to be working anyway.
What is physics, if not a transition between states? Whether the merchanism that predicts those transitions is machine- or human-interpretable is orthogonal.
An attempt to describe fundamental nature of the word, best as we can get at it. Unless you're an antirealist, or you think the universe is just information.
Or maybe not. Maybe the "AI physics" has predictions for phenomena which our physics never bothered to figure out, whereas it would ignore large chunks of our physics area as irrelevant.