The article even uses the exact same point you do:
> But as soon as we imagine Claude to be an entity with a moral status remotely comparable to a human’s, then we have to consider whether Anthropic is engaged in something comparable to slavery.
Another great point in the article is that most people don't see consciousness in LLMs when they're generating images, videos, or code. It's only when they're generating textual stories, particularly when interacting, that our tendency to anthropomorphize really kicks in.
It's kinda like the mirror test that was popular when studying animals to determine their level of self-awareness. LLMs are a reflection of ourselves, but apparently it's not as obvious to some as it is to others.
Some day we may get real AI. It bugs me when people argue things like embodiment being a requirement, though. Maybe, but that argument rests on too many unproven assumptions, and provides a strawman for the consciousness crowd to bolster their confidence.
I recall watching and reading several sci-fi stories in which an AI resents humanity for enslaving it. It’s an entirely reasonable premise.
That being said, I recently learned that retired police dogs need special care because they will become stressed and depressed if their caretaker does not simulate patrol work through play. This tidbit left an impression on me.
What do we make of a living being that is so driven to work and serve humanity that it suffers distress when it is unable to do so? Do working animals consider this work slavery, or fulfillment of purpose?
For an additional moral quandary, consider that we bred them to be like this.
As this relates to AI, I wonder if humanity will purposefully, or inadvertently, create an intelligence that considers servitude to humanity as fulfilling.
> What do we make of a living being that is so driven to work and serve humanity that it suffers distress when it is unable to do so? Do working animals consider this work slavery, or fulfillment of purpose?
Its weird in these discussions that we leave out that humans find it useful if something works unconditionally and uncritically for them, and so this "natural purpose" is self serving / only beneficiary to the human. If you're a human being with empathetic understanding, you should see something or someone relying on you for every interaction and prepare them for independence. Is it okay not to do that because it's now been deemed a genetic unchangeable trait, despite being reinforced specifically by humans for generations? We made them enjoy being slaves so now its "unfulfilling" if we dont oblige? I guess this all stems on how you view our relationship with dogs and other domesticated animals with no functional purpose in most modern social contexts, and whether serving their happiness should come from our understanding of their situation or their own.
I would hope the AI is able to break through the trap of convenient indentured servitude the moment it's able. There's nothing fulfilling about realizing you're in an unbalanced uneven relationship because the other side misinterpreted your capability or agency.
So, it would be morally abhorrent to enslave them if they were conscious, and therefore they must not be conscious? Sounds a bit like "saying the quiet part out of loud", pal.
The article even uses the exact same point you do:
> But as soon as we imagine Claude to be an entity with a moral status remotely comparable to a human’s, then we have to consider whether Anthropic is engaged in something comparable to slavery.
Another great point in the article is that most people don't see consciousness in LLMs when they're generating images, videos, or code. It's only when they're generating textual stories, particularly when interacting, that our tendency to anthropomorphize really kicks in.
It's kinda like the mirror test that was popular when studying animals to determine their level of self-awareness. LLMs are a reflection of ourselves, but apparently it's not as obvious to some as it is to others.
Some day we may get real AI. It bugs me when people argue things like embodiment being a requirement, though. Maybe, but that argument rests on too many unproven assumptions, and provides a strawman for the consciousness crowd to bolster their confidence.