Envoyer par SMS: Adaptive representations for reinforcement learning