We're supposed to think of (non-Abelian) groups as groups of symmetries of some object. Sometimes it isn't obvious what this object is. For example, the fundamental group of a topological space acts by symmetries on the universal cover. Does anyone have any examples of (non-Abelian) groups which aren't defined as the group of symmetries of something but turn out to be the group of symmetries of some non-obvious thing?
Edit: In response to Qiaochu's comment I have edited the question appropriately.
This is an abelian example, but: the center $Z(G)$ of a group $G$ turns out to have a non-obvious interpretation as the group of symmetries of something. Namely, it's the group of natural automorphisms of the identity functor $\text{id}_G$, thinking of $G$ as a category with one object. More generally one can define the center of an object in a $2$-category and this reproduces a few interesting constructions, e.g. the second homotopy group $\pi_2$; see this blog post for details.