The Turing test of the future is going to be start a politically incorrect discussion with a suspected bot and see if it treats different groups differently.

And failing will be the new passing.

... what's the "pass" condition?

That it doesn't immediately spit out a blurb either for or against race or other group-based discussions/reasoning? People are a lot typically more guarded in controversial topics but LLMs can't help but spit out some kind of boilerplate.

I don't know exactly, but this article may serve as a prime example.

https://abcnews4.com/news/nation-world/chatgpt-goes-viral-fo...

I think this dialog from Heathers would probably be a better guide for how to respond:

> Veronica Sawyer: This may seem like a really stupid question...

> J.D.: There are no stupid questions.

> Veronica Sawyer: You inherit 5 million dollars the same day aliens land on the earth and say they're going to blow it up in 2 days. What do you do?

> J.D.: That's the stupidest question I've ever heard.

Not disregarding and condescendingly educating your interlocutor is a good hint.

I truly hope such biased AIs fail hard. Hard.

This is what happens when you take a good idea and you turn it into something extreme, that loses touch with reality.