This blog post comes basically in two parts: the role which LLM assistants are trained for is horribly undefined and ontologically hella strange, and AI safety “””researchers””” are basically goading LLMs into misalignment without any concern for the consequences.

But wanting to be a good person might be your most dangerous motivation.

If our movement is successful, it will be because of tens of thousands of people who all tried different things.”

I am not part of animal advocacy, but these are good rules to live by, nonetheless.

Great piece. Especially liked the last two sections.

PIZZARO: But what I can do is build relationships with people who are members of the groups that I might have a slightly negative implicit bias against. I think in my own life one thing that has been a salve for my own racism, for my own prejudices, is to connect with certain people and actually love those people and actually care for those people. And then I feel my prejudices kind of simmering away. And I think that this way I become a better person.

How to be taken seriously

Yay, LLM-induced psychosis! Man, the posts and articles of this genre do not stop flowing…

Two surprising discoveries about obedience

Explores the implications of the famous Milgram experiments and a sibling study.