MIT's Phillip Isola and NYU's Ravid Shwartz Ziv debate why structurally distinct AI models converge on similar world representations · Digg