We should be bold enough to venture guesses, expressed clearly enough such that it’s easy for someone else, or the universe, to prove us wrong. In the long run, I think this policy leads to much better models — and better thinkers — than the policy of trying to minimize our error in the short run. (source)