No, the argument is 78% of Americans don't use Twitter at all, 20% have Twitter accounts but are not very active, and only 2% are responsible for the vast majority of tweets. In other words, what you see on Twitter, especially when it comes to somewhat more extreme messages, is not representative at all for the US population as a whole.

how many people vote in elections ? just to compare the representativeness of the samples

Forty to sixty percent, though that doesn't say anything about the representativeness of voters or twitter users.