Attending Transformers
I had the rough idea of transformers architecture from the famous Attention is all you need paper, which took everything by storm, because AI has existed for many decades now but until now either everything was kind of rule based or if not for every task you'd need to train a seperate model, but with transformers which initially showed Translation example, developed so much that now it almost feels like it understands what we write, what we see, what we hear, although its just a bunch of number computations and boom, its magical.
Read Article