I doubt that Eric Meijer, one of the co-authors of the Haskell 98 spec, personally finds modern type theory too complex to deal with. He seems to be proposing that it's not necessarily a positive language feature to expose a very complex type system to the programmer (or at least, to make it mandatory).