"But from a Bayesian perspective, you need an amount of evidence roughly equivalent to the complexity of the hypothesis"

But what is the "complexity of the hypothesis"? Without a proper definition, there is not much left to the article, or is there?