I think most startups made to strike it big don't do anything interesting. The great ones are those who idealistically set out to do something great and somehow succeed. For example, google, or tesla, or (well, I'm surrounded by good ones so I can go on all day.)
Think power and influence over and improved world, not money. Though the money sure is nice.
The failure rate is why I like the bootstrap method where you work on a tiny idea on the side until suddenly one day it's earning enough for you to quit your day job. Your standard "we're gunna make an app" startup has a prodigious burn rate - partially for good reasons (you only have a year to make it big, so hire good people and throw money at them because if you don't you won't have a chance) - but also partially for bad. There's a large separation between the source of money and the result: money comes from A, gets distributed by venture capital looking to earn a return from that money (B), sometimes paid to C to find and hire D (or sometimes directly to D), and D is the one who makes the product, which is purchased by E. Oh, and if it's an app, the product is given away to E, embedded with ads from F, who is often skimming off the top for providing the ads, which come with money from G. A wants a good return from G.
This is why the really exciting startups skip a couple steps: A invests directly into paying B, along with B contributing cash and connections as well, who together hire more B, who make products to sell to C. And the bootstrap model cuts that down to just the folks who make the thing and the folks who buy the thing, if it's a real product. Now the people making the thing have a very real investment, not a vague "0.1% of the company before dilution."
Everyone in startups know you don't hit oil when you come in and get "0.1% before dilution." You might get the contract to haul the oil, but it won't be yours; you can get a couple hundred grand or even a couple million but since startups tend to be in places where the damn houses are a million or more, well, it doesn't go so far.