I named fb as a counterexample. I think you can definitely succeed without validating or invalidating your idea first, but it's a much bigger risk, and you are building a very specific kind of company. How many facebooks and twitters are there in the world?
Which is great if you can implement the idea in a weekend. Much harder if it'll take 2 years to build before you can see if the time was worth spending.
IMHO, you invalidate an idea after you implement it and see the user's reactions. Before then, you cannot know if it will succeed or not.