Let G be a discrete group with a fixed finite generating set S. A map from G into some (finite dimensional) unitary group U(n) is an epsilon-representation if it is a group homomorphism up to epsilon error (for the operator norm) on the finite set S. Thus a quasi-representation is a close to being a representation in some sense. The group G is stable if every epsilon representation is close to an actual representation, in a precise sense. For example, free groups are fairly obviously stable. However, a famous result of Voiculescu shows that the rank two free abelian group is not stable. In his thesis, Loring gave this a topological interpretation: it turns out that Voiculescu’s result is more-or-less equivalent to Bott periodicity. I’ll try to explain all this, and how topological information can be used to produce many other examples of non-stable groups.