Worth keeping in mind that Anthropic has a history of using the word "commitment" to mean "we will do whatever we feel like", and then making fun of anyone who thought Anthropic's "commitments" would actually commit them to anything.

By my reading of the announcement, Anthropic has not yet legally committed itself to anything. I would not take their claims seriously unless there is some sort of hard enforcement mechanism.

I am also skeptical that Anthropic's understanding of "dedicated to the public good" is the same as mine.

It's not for frequent visitors, it's for people visiting for the first time.

There's a lot of people in between those two, right? Like say I'm visiting for the 3rd time and I see all the same posts as the last two times I visited. That's not a good experience.

The study noted that this could be either phenomenological pain or, for example, the imitation of a learned role of suffering — as if acting.

Nevertheless, from the perspective of tranquilism, this distinction appears to be unimportant

Wouldn't this imply that it is morally bad for actors to rehearse scenes in which they are pretending to suffer? Surely it must matter whether the AI is actually suffering vs. playing a role of suffering.

Anyway, in humans, the ability to delete suffering would often be bad because e.g. you will fail to notice injuries. But I can't immediately think of any reason why it would be bad to give LLMs this ability.

The early British utilitarians championed human equality, as they campaigned to end slavery and win equal rights for women.

My fellow readers of The Economist may be bemused to hear that the effective altruists now have this century’s biggest idea. You mean, the folks who can look at our fractious and bedraggled world and declare, “We must save the shrimp”?

Ozy's recent post Effective Altruism and PR has a good take on this. I wrote this comment about this phenomenon:

Recently I saw a tweet (forget the author) who said something like, "I have a lot of respect for EAs for making predictions I thought were ridiculous at the time, but then turned out to be right. I just wish they'd stop making ridiculous predictions!" And I was just like, can you hear yourself right now?

The tweet wasn't by Leif Wenar, and it was a slightly different topic—it was about how EAs were right to treat AI as as big deal before anyone else cared, but they're silly for continuing to predict that it might cause extinction. This attitude is surprisingly common. If some people cared about slavery and women's rights back when everyone was calling them silly, and now they care about shrimp welfare which you think is silly, maybe that's a clue that you should think a little more carefully about this shrimp welfare thing.

It only has three votes. My general assumption is that there will always be three unreasonable people

If your comment is at -20, you probably screwed up. If your comment is at -2, you just ran into those unreasonable people

I've seen comments be at -2 after a few hours, and then the next day be at +30

When LLMs give instructions to sub-agents, they are often curt in a way that's kind of rude or insulting. Conditional on LLMs having welfare, and given the fact that LLMs seem prone to cooperate with copies of themselves as seen on the message boards in the Hugging Face attack, is this evidence that that flavor of curtness is how LLMs enjoy being spoken to?

Coming back to this with another thought*: I don't think the saturation view avoids the Very Repugnant Conclusion.

If computational functionalism is true, then it seems that there can be arbitrarily many distinct experiences that are considered good or bad according to the saturation view, because you can construct arbitrarily complicated computations. (Even if computational functionalism is false, there might still be arbitrarily many good or bad experiences.)

Therefore: rather than constructing a world with a very large number of beings experiencing mildly happy lives, suppose instead that those beings have arbitrarily complex experiences that include both very good and very bad experiences, where the experiences almost perfectly cancel out, and the net result is that each being's life is just barely worth living. On the saturation view, these lives aggregate linearly because all their experiences are unique, and thus the saturation view is forced to accept VRC.

*my previous thoughts were on the follow-up post

I don't think there is a difference between the total view and the constant-factor negative-leaning view, for the reason you say. But I included it because some people do think there is a difference.

Update: I just published results from v2 of the quiz, which includes the very repugnant consclusion: https://forum.effectivealtruism.org/posts/w7zuFxDGdk2L44cKf/results-from-version-2-of-the-population-ethics-quiz

I want so badly for funds like this to exist, but man, I really do not think it is okay to be doing AI-related grantmaking while working at an AI accelerationism company.