I mean, I agree. But Y Combinator is hardly a function combinator. If the worst criticism is the name, it's a signal that may be worth paying attention to.
They're spewing bullshit. That opening line alone tells me they're copying every other VC (and most VCs lose money) while simultaneously portraying themselves as being better than every other VC.
YC succeeded by completely subverting what everybody thought VC had to be. You can't expect to be better by doing the same thing as everyone else.
Yes. That's true. It might not work. And you're probably even correct. But that's always true, and the most valuable cases are those that work when most people don't expect them to. http://paulgraham.com/swan.html
You say that every other VC does this. But if other VCs are training ML models, they don't seem to be publishing their results, so I'm inclined to say that this is new.
Is it going to work? Well, we'll have to see. But as with most startups, the answer is "probably not, but the chance may be significantly better than random."
If they are looking for swans that work when people don't expect them to, they should probably be looking for those swans somewhere else*, not the top 3% of an incubator that probably graduates proportionately more successful startups than any university.
The issue that everyone is calling out here is that they name themselves "Rebel" when they're doing exactly the same thing that every other VC firm is, but (apparently) not even as well -- at least they should say "Founder Friendly", etc.
* Generally startups that might become unexpected swans will be found outside of startup hubs like silicon valley.