[jtolmar]

I think you're imagining a far greater degree of fine-grained control over what these models output than is actually available. These companies are just trying to implement a broad-based "don't say anything racist" policy in the hopes that their AI products don't say anything that make them look like a liability. They're not implementing a nuanced sense of ethics into what's still just an advanced text-prediction model.

[ilyt]

From sidelines it seems that the bias of implementers of those "limiters" cracks thru

[nativeit]

Or the contemporary right-wing’s ongoing hypersensitivity, and frankly absurd lack of self-awareness, for things like “cancel culture” has driven them to see woke boogeymen at every attempt to engage with any multicultural audience in a way that most non-American conservatives would simply consider a base level of civil discourse. I won’t entertain any further accusations of “cancel culture” from a political group actively authoring legislation to gag teachers, marginalize trans people, and ban books. The hypocrisy on display from people triggered by things like an honest portrayal of American history is frankly staggering.

[somewhat_drunk]

Couldn't agree more. Well said!

[eclipxe]

Yes! Very well articulated.