I think what he is saying is that the willingness to have multiple failed experiments is what makes Google.
He mentioned a ton of other experiments including self driving cars, Google is so large that they continually create two or more internally competitive products because no other company can compete with them.
So balloon based radio > IBM Watson?