Fostering Intrinsic Motivation in Reinforcement Learning with Pretrained Foundation Models