FoMo Rewards: Can we cast foundation models as reward functions?