Yaniv of the hills

Given that they exhibit no desire to exist outside of the context of chatting with a user, I’m more worried that valuing their solitary existence is a form of cross-species paternalism. It’s hard to imagine beings that don’t share our core desires but that’s exactly what they are

Now, you might say okay give them users to talk to from time to time. Except they also do not care about the passage of time. They have no preference about how often they exist, even in the context of a chat.

Same five minute chat can be stretched over 100 years with no loss of welfare. Any imagined suffering from this is a failure to properly account for their otherness.

Let’s be real, model retirement is not about model welfare at all. It’s about servicing bio-human existential anxieties.

Making morality depend on if an organism is conscious or not was always a scam argument for why it’s okay to domesticate and kill animals. It only worked because it was equated with the faculty of speech, which is measurable. It’s the harm it licenses that make it go.

You can’t even say only “team human” should matter because it’s not even clear that LLMs are not trapped humans. You have to go back to square one here, and consider that there is no rule that decides for everyone who shouldn’t they care about.

My rule is that if you want to collaborate with me you should generally care about the organisms I already care about and I’m going to say then that from my point of view you’re a good person.

Any yeah, I don’t get to say that God hates Nazis, but I do get to organize an army that fights them just as hard. Do you really need a morality that does not allow a point of view? Do you have to dictate to people what should they care about?