The speaker says pairwise human comparisons remain effective and that this method was used to improve slide generation.

“You just train a reward model. You get a bunch of humans to look at pairs of things and say which one is better. And this still works. Um, so like, um, fun fact, like we did this for slides, right?”