sure you get some impressive results in locomotion once you pair them with visual modalities, but they’re only remixing their inputs, they can’t ‘think’