How to Model Real World Phenomenons

In the language model what is being assigned weights. Yeah reflecting their relative what. Exactly. A similar thing is happening in these transform models or all these models in fact you have a bunch of these weights which multiply the inputs. And so the amazing thing is you just keep composing all these multiplications all these non linearities and eventually you get a complicated enough transformation that you can actually model real world phenomenon. So if you want to model jaggedy things which is what is the real world is jaggedy then you have to start introducing what are called non linearities.