I asked a question, you are making a statement. If you have numbers that prove that the formally educated are failing more then I think we'd all love to see them. Otherwise, this is just conjecture.
As for 'science projects' failing it could be a number of things. I would assume that hard science startups fail a lot of the time because of cost. "Twitter for teens" is much cheaper to start than one doing alternative energy. Also, hard science is less sexy than a lot of the social stuff that currently dominates the landscape. Thusly, it can be much more difficult to find the money you need due to visibility. The social startups will probably pay out sooner than long-tail hard science companies.