Comment by _jayhack_

Vector embedding is not an invention of the last decade. Featurization in ML goes back to the 60s - even deep learning-based featurization is decades old at a minimum. Like everything else in ML this became much more useful with data and compute scale