Computer analogy: we're trying to reverse-engineer an alien ASIC, which can operate either on an array of tuples of three reals, or on an array of bools. It works approximately the same in both cases. The latter case is much easier to analyze, so we prefer to do that.
“... studies the connection between the highly non-convex loss function of a simple model of the fully-connected feed-forward neural network and the Hamiltonian of the spherical spin-glass model”
from "The Loss Surfaces of Multilayer Networks" Choromanska, Henaff, Mathieu, Arous, & LeCun [1]
[1] https://arxiv.org/abs/1412.0233
Also the physics of collapsing sand piles.
The other comment also led me to https://en.wikipedia.org/wiki/Spin_glass, which I didn't know about at all (I'd heard the term once but had no context for it).