How should philanthropists and foundations fund science?

This article from the NY Times discusses the funding of science research by wealthy individuals and, by extension, philanthropic foundations set up by those folks.  It brings up an issue that I phrased in the form of a question as the title of this post.  I'm not going to offer any simple overarching answer, but I do want to make a couple of observations (strictly my opinion, of course):

• Many wealthy people and foundations support medical research.  This makes a lot of sense - generally philanthropists want to Help People, and supporting research that directly affects medical care and quality of life is a completely sensible choice.
• A smaller number of people and foundations support research in the physical sciences and engineering; like those who support medical research, they want to Help People, and they realize that supporting basic research and the education of technically skilled and creative people is a great way to do that.
• Both groups, however, face the challenge that any investment that they make is basically a drop in the bucket compared with what governments can do.  NIH puts tens of billions of dollars a year into medical research.  NSF's annual budget is around $7B.  
• In my experience, the philanthropic supporters of science are well aware of this - if they want to make sure that their money makes a difference, they need to invest in supporting things that are not what government agencies are already doing.  Their challenge, then, is to identify areas (and eventually institutions and people) where their investment will really move the needle.  Of course, they need to be able to tell wheat from chaff.  Peer review is a customary way to do this, and that is often how the government agencies (most of them, anyway) make judgments, but peer review tends toward being risk averse.  An alternative is to have a dedicated science board to do reviews and make decisions.  This, too, is tricky, particularly if the members aren't exact subject matter experts.  How much weight should be placed on prior track record?  Researchers who are senior and already have major awards, etc. can be a lower risk - they have already demonstrated that they can do great work.  On the other hand, if someone is already extremely well supported (as such people often are), how much difference will philanthropic support really make?  It seems like a very tricky decision process, particularly depending on the amounts involved.  
• There is no question that having grants with wide flexibility (e.g., Packard; presumably MacArthur) can be wonderful.  At the single investigator level, there is also no question that there can be real benefits from being able to concentrate on actual science - that's an argument for funding support large enough that it allows investigators to lay off writing other grants to some extent.  (That's one aim of things like the Howard Hughes Investigator program.)  

Feel free to comment below.  Perhaps additional philanthropic support is one way to get more "scientific mavericks" (as long as that doesn't mean funding of unfounded crazy stuff).