http://thielfoundation.org/index.php?option=com_content&view=article&id=15&Itemid=19
Thoughts?
A lot of Thiel's beliefs are in line with those of a significant portion of the LessWrong community, so I trust his judgment more than I trust that of most.
My guess would be that the simplest answer is the most likely: a filter tight enough to weed out every completely unrealistic project, will necessarily also weed out at least some that are just within reach and have very high potential payoff. Most of our civilization's funding sources are attached to pretty tight filters; I think we could do with a few more on looser filters.