The greatest challenge to a model of the emergence of representation is that of the normativity of representations: the possibility of being true or false. The strongest version of that challenge is to be able to account for system detectable representational error, as is used in error guided behavior or error guided learning. No model in the standard literature, and, arguably, no spectator model of any kind, can account for it. Genuine representation, however, with content and truth value—system detectable truth value—emerges in the selection of actions and interactions in autonomous agents, whether natural or artificial, organisms or robots. Representation is most fundamentally of future potentialities for interaction, rather than of past encounters as standard approaches would have it. Representation is intrinsic to agents, not to passive spectators. The fundamental aspirations of Artificial Intelligence to create genuine artificial minds will be met in robotics.