All of the modules he suggests avoiding are the ones that make up the standard data science tech stack. So, IOW, learn Python before doing data science. Your usual beginner Python tuts don't mention Pandas or matplotlib, so this seems like a non-issue.
He points to a graduate student who had to pick up Python for some task and wanted to learn both the language and the tools necessary for that task. I can throw another anecdote into the ring and mention a web data mining course I'm taking right now where a lot of students are having to pick up Python along side both scikit-learn and scrapy.
