He makes a great point that our intuitions are not well-served when the return on investment can be 1000x or even 10,000x. Humans like to be mostly right. In venture capital you can easily afford to be mostly wrong--- as long as your criteria let you pick the big winners too.
What this leads to is funding ideas that sound "crazy":
The first time Peter Thiel spoke at YC he drew a Venn diagram that illustrates the situation perfectly. He drew two intersecting circles, one labelled "seems like a bad idea" and the other "is a good idea." The intersection is the sweet spot for startups.
Now, I take a little issue with this. There are plenty of successful startups where the badness has been "there are plenty of people in that space already/you can't complete against the incumbents" (like Google, or perhaps Sun.) And there are also startups like, say, Granite Networks or Arista where the issue is making a big bet on the direction of a technology. These are more obviously "will either look great or stupid in retrospect, but we can't tell which." But it may well be that the reward in such cases is not as big as, say, Facebook.
I do find it interesting that PG's realization that it's the big wins that drive revenue leads him to a more relaxed attitude toward the also-rans.