Higher ISO, less noise
What the author said is right but misleading. He is comparing a heavily underexposed ISO 100 image pushed in developing, with a right exposed ISO 1600 image normally developed. I don’t think anybody would deliberately underexpose 3 or 4 steps at ISO 100 and then push the same 3 or 4 steps when developing the raw file, just not to use ISO 1600 in the first place. Of course the ISO 1600 image will be noisy, but the pushed ISO 100 image will be worst, as he demonstrate. Nothing new here, as I see it.