Comment by quantadev
I consider MLPs the building blocks of all this, and is what makes things a neural net, as opposed to some other data structure.
I consider MLPs the building blocks of all this, and is what makes things a neural net, as opposed to some other data structure.
Sure. But that isn’t a reason to conflate the two?
OP wasn’t suggesting looking for an alternative/successor to MLPs, but for an alternative/successor to transformers (while presumably still using MLPs) in the same way that transformers are an alternative/successor to LSTMs.