The author has a point. As long as we don't understand how we're built and functioning, modifying our code is a bad idea. Learning by doing is only applicable when the thing in question is expendable, which is definitely not the case with the human genome.