I think the author got it slightly wrong here because he's conflating the programming language feature and its semantics with the run-time implications of the technique(s) you use to implement that feature.

For what it's worth, this seems to be a common misunderstanding where generics are concerned.