A bit too generalized for my taste. If you're solving a problem of passing a well-defined benchmark (or closer to home, iterating on a kitten meme-sharing app), then yes, fail away. If you are solving a problem of putting a man on the moon in the 60s, then you get to coordinate hundreds of thousands of people from education to fundamental science to industry to alloys to god knows how many other things for years. And only then you can launch and really find out if you succeeded or failed - both outcomes have precedents.