The whole thing he is explaining is just regression to the mean. Yes, if you take people who are currently rich and people who are currently poor, over time the poor will likely go up in income and the rich go down.

Using this as a counterpoint to accusations of the rich taking all the gains is like pointing out that a wealthy duke could only pass his whole dukedom to one of his children, and the others would have to make do with much less.

It is simultaneously true that the gains of the last few decades have largely gone to "those who are rich", and that the identities of "those who are rich" are not completely stable.