It actually makes sense. Assume men and women are equally competent on average (probably true or at least close). If your applicants are 18% women and the applicants are randomly sampled from the talent pool and your interview is totally unbiased you should get 18% women hires.
If you convince a lot of (randomly sampled) women to apply so you have 50% female applications, you should still get 50% female hires. I don't think this is unfair on men as long as your outreach activities don't at all affect interview chances. In particular they shouldn't give an "inside track" to bypass HR.
Sounds like they've at least thought about that and made the first stage blind.
None of this stops a candidate from applying to duolingo directly. There's no discrimination going on against humans, and discriminating against a school is totally fine so long as there's no suspect reason to do so (e.g. discriminating against Howard because they've got primarily African-descended students).
Not a lawyer. Would welcome a legal opinion.
I would highly recommend reading at least the Wikipedia article on "Disparate Impact."
They very clearly make the case that they are biased in favor of women -> over men. You might think that 50/50 means equality, but up front they state that there are only 18 percent women in the field. In order to achieve 50% women in their workforce they have to be biasing the hiring process. It seems strange to go out and write publicly about doing so.
This is very similar to redlining and other implicitly discriminatory practices in the past, and the longer it goes on the more likely I suspect it will end up in court.