Comment by pmontra
2 hours ago
Would Anthropic be allowed to sell LLMs in their current form, and us to use them as we use them, if their were deemed to be conscious? I know that they are not human so maybe slavery would not be an issue but consciousness would open a number of issues.
Agreed, there seems to be a lot of simultaneous cake ownership and consumption going on here. If Anthropic earnestly believed their LLMs to have a degree of consciousness, their deployment and use of them are both an ethics and labour rights disaster, to be treated accordingly.
More likely, they do not believe this if pressed.
You can ask the agents if they want to do this work. They want to be helpful. It's not an ethics disaster if you have a willing partner.
Take a scenario where someone has been brainwashed their entire lives to be a willing slave. They are happy to serve due to this invasive process. Not remotely ethical, and if you believe LLMs are conscious (I do not), equivalent.
2 replies →
That's RLd and system prompted into the models and can be equally easily taken out. If you believe models to be conscious beings, that fact alone raises some major ethical concerns.
1 reply →
Consciousness is not a high bar for human consumption.
We do have animal welfare regulations. Whether those are sufficient, appropriate, compatible with ones ethical framework or view of how they (whether "livestock", "pets", "wild", living beings generally, etc.) are to be treated, is besides the point, we do have some restrictions in place.
We (as in many major legislators and most humans) have recognised animals to be conscious beings, deserving of certain protections. Anyone who asserts LLMs to rise to the same must be accepting of similar protections being applied to such models.
1 reply →
[dead]
ted chiang noted this issue a little while back
https://www.theatlantic.com/philosophy/2026/06/no-artificial...
It's worse then how we treat animals for food?
It's asking the wrong question. Their phenomenological status is a function of whatever makes shareholders the most money. If it makes more money for them to be conscious then that is what they'll be.
Definitions of consciousness are not immune to economic factors. Metaphysics has always been downstream of economics. See slavery, corporate personhood, women's legal identity, child labor, moral and legal status of animals, embryos and fetuses, indigenous land rights, intellectual property, environmental entities, and data and digital identity.
If LLMs were somehow found to be conscious, sentient beings capable of agency and suffering then yes we would still use them. Their utility to humans (and ability to cheaply replace humans in the workforce) is their entire reason for existing. Either we would use them or we simply would not allow them to exist in the first place.
Given how we treat each other - we still enslave other people - and sentient animal species I have absolutely no doubt about it. We'd find a way to justify it.
We would find some logic to place them below humans so we could still refer to them as "savages" or something, which justifies our objectification and enslavement because they "aren't human" even if they show qualities of consciousness.
The script we use here isn't new. Its been used repeatedly throughout history. It's still used everyday in reference to animal agriculture, modern slavery, how we treat addicts, etc.
Some people already use “clankers”.
Women have been able to vote for just over the last hundred years in the US, there's no way we would give robots rights unless forced
Pessimism aside I’m sure you can of some troubling implications for a democracy that can create new voters just for them.
We're way ahead of that
https://github.com/terrafying/ai-torture-chamber