Implications of the Convergence of Language and Vision Model Geometries