it can be demonstrated that The straightforward pre-teaching process of predicting which caption goes with which picture is definitely an efficient and scalable way to discover SOTA picture representations from scratch https://k2spiceshop.com/product/liquid-k2-on-paper-online/