You're right that there are objective downsides to startups (e.g., time sunk into something that doesn't pan out), but the lower bound is known and, to me at least, not that bad. Working on a startup that fails isn't like shorting stocks (which has a theoretically unlimited downside). In fact, I'd argue the worst case is many times better than working two years at Microsoft or Google, at least in terms of personal growth and development.
It doesn't have to be an either/or. There's plenty of time to both fail at a startup and work a couple years at Google.
A related trick is to keep the scope of your experiments small so you can feed that information back into the development of future experiments very quickly. (cf. Col. John Boyd's "Patterns of Conflict")
Spent 2 years as an online poker player, you learn that your short term productivity is an illusion. What is most important is a type of effort that produces long term gains. With that perspective you become better at prioritizing and understanding what both you and those around you (customers, investors) want. So the only way to be "unproductive" is to not follow through thoroughly on what you are doing now.
But you need to be smart enough to get out when the long term is, as we say in poker, -EV (http://en.wikipedia.org/wiki/Expected_value)
Talking to a woman at the supermarket has the inherent "risk" of being ridiculed, while having the obvious objective upside that it might get you laid - or at least makes you a more confident person. In the same way (now regarding business), pouring time and money into a startup has the inherent risk of said startup going under (and losing some past time and/or money), while the obvious objective upside that if the company makes it big, you're a billionaire. Hell even if it doesn't you've earned experience you can apply in your future endeavors, and you didn't have to listen to 7am lectures at the local college for four years to get to not learn even a tenth of what you learn failing at building a startup.
Also: A lot of what makes it hard to be a founder is not the shame of failure, but knowledge of lost time. If my company goes under, I'm not as disappointed in my lack of success as I'm disappointed in time not spent more productively elsewhere. There's a subtle distinction there.