comparison between sensitive info types and sensitivity labels

It looks to me that I can create DLP policy rule using either sensitive info types or sensitivity labels. 

I would like to find out what's the advantage or disadvantage to chose one over the other when I create DLP policy rules. I have googled on the internet, but did not find article to compare sensitive info types and sensitivity labels together. Anyone can share this kind of information?

Thanks in advance,

Peter