This is clickbait. Unless models are invariant to adversarial examples in SQuAD such as those described here: https://arxiv.org/abs/1707.07328, models doing really well on SQuAD doesn't mean a ton.

Can't they simply include adversarial examples in the unreleased test set?

This seems like one of the first major illicit uses of data science. There should be more regulation and oversight over the data owned by corporations.

It seems like a good idea until you think more about it.

I at least wouldn't let the US Government, NSA included, have even more access to private data.

Agreed with my fellow co-speakers. It was humbling to be surround by such brilliant people. But it was also so inspiring to be a part of a community that takes pride in welcoming and empowering every developer, regardless of background.

Rust language! One of the most supportive groups I've ever been involved with.