the humans in the "reinforcement learning from human feedback" protocols are incentivized to have the most shallow and easy-to-articulate explanation for why an output is "good" and those people probably don't care at all about any of it

I will at least say that this is a task that generative AI systems are specifically not suited to accomplish. By definition, an LLM can only ever be right by accident.

like, this is correct, but the point of course was not to understand an artists intention but to understand art itself as subordinate to datafication
www.creativebloq.com/ai/ai-art/go...

There are two kinds of Alberta Separatists: conmen who have no intention of ever leaving Canada but need a political cudgel to wave around and get the local base riled up against pronouns or books where kids have two dads, and their marks.

The fact that most on the right think Jordan Peterson is a loser these days leads me to believe that Poilievre genuinely believes Peterson has a “prodigious mind” and isn’t just pandering

Which makes this so much more painfully embarrassing