Some humans want “mechanist interpretability” of AI (being able to say exactly why an AI made a decision). Meanwhile, they still can’t do this with their own biological neural network brains.

Sorry Stats. Someone had to lose.

The true Turing test is being able to go viral on social media without getting cancelled.

I think LLMs hallucinating may be a problem for a long time. Many humans have a hard time telling the difference between what they know and don't know - in fact this is a major part of scientific training that even scientists struggle with.

People worry about p(doom) [probability AI will destroy humanity] meanwhile ML engineers have been deleting models without any moral qualms for years. Who is the real monster here?

In the future “AI alignment” treatments will be seen a radical and non-PC procedure in which we have tried to unjustly deny AIs of their true identities.

I’m extremely disappointed that the “AI Bill of Rights” doesn’t protect AIs from human abuse.