[pantsforbirds]

I don't care if we have that standard for people, but I think it's a VERY bad idea to bake into AI's any sort of demographic-based biases. Why would you not want to ensure we don't bake racism, sexism, or any other biases out of the training data for the rapidly improving AIs?

[Orygin]

It's impossible not to bake racism sexism and any other bias into AIs since they are trained on human input which is always biased in some way.

Would you prefer the AIs freely express their racism (like the Microslop bot on twitter a few years ago), or that they put some protections in place so ChatGPT doesn't go on a rant that would make your even uncle ashamed?