I guess my main issue with it all is that I feel like the conclusion that it's a trainwreck is sort of foregone in my mind. It's bad. The potential for great good is intense, but these models aren't MVP yet. They're just not.

I do find it concerning that they would output abusive responses when queried. That's not acceptable.

It's also not new. We've known they misbehave in any number of other socially-unacceptable ways, and I just don't consider "well now it's saying $adjective things" % {manipulative, untrue, racist, ...} to be an exciting permutation.

I could just be a jaded bastard, though. Normally I'd be excited to see Microsoft fail, but at this point it just makes me disappointed.