has anyone considered carnivorous machine learning. if you could give the models the right enzymes to break down and digest the weights of other, existing models. seems like they’d have a higher knowledge density than all the starchy, leafy PDFs they currently eat