[neevans]

Humans learn from seeing I don't think we are the stage of training models with videos / images dataset. We only reached the plateau with text dataset to train with.