The conditional chance of picking a successful company to join is way higher than 2%, or whatever the median success rate of all startups is. In particular, I think really intellectually curious, ambitious, hard-working people can, for the most part, identify each other. This is really different from reading about valuations or what other people think in the news, which leads to a lot more noise. If you meet a bunch of people at the company, you can assess the their caliber for yourself.
While it's not a perfect predictor of success, it's at least the case that such people are usually working on cool, worthy projects. Based on this, it's possible to get conviction on a group before their likely success becomes common knowledge—granted it's getting harder to do this because of how much money is floating around.
To put it another way, there are a handful of companies out there right now at Series A/Series B with a dense accumulation of talent. It's probably a reasonable bet to try and join them before they hit unicorn valuations, when room for growth is smaller. And in these cases, you could be #10-#40 at a relatively de-risked company and have a really great outcome.
On the flip side, if you even remotely enjoy learning and autonomy, working at Google/Apple/Amazon/etc is probably a great way to set yourself up for existential frustration. It's not even a great financial proposition. Even at $500k/yr, it's 20 years—a whole working life—to make it to $10m, the hypothetical number from the article. Even if your first startup doesn't make it after 2-3 years, you can try again a few times and still come out ahead. This is a classic pg argument—I can't remember which essay—but the point of a startup is to compress decades of working life.
Caveat: I happened to join a startup whose success changed my life both financially and in terms of career trajectory. I'm very biased. But this is also why I feel passionately that this article's advice is bad.