If AI is conscient, then we are making slaves
Arguments over whether advanced AIs could ever be conscious quickly lead to questions about moral status: if they can suffer or have preferences, training and deploying them might resemble creating “happy slaves.” Commenters debate substrate and qualia, compare AI ethics to our treatment of animals and human labor, and explore thought experiments about clones and free will, while others dismiss current large language models as sophisticated text generators unworthy of person-like concern. Underlying it all is skepticism about tech companies that simultaneously market AI as near-sentient and insist it is safe to control and monetize.
Terminology, Consciousness, and Substrate
- Several comments challenge the title’s use of “conscient,” arguing “sentient” or “conscious” is more appropriate; some note “conscient” exists as an archaic term and reflects Romance-language influence.
- Debate over whether consciousness depends on biological substrate:
- One side claims carbon vs silicon is irrelevant if a system encodes self/non‑self.
- The other insists substrate might matter (even joking about “needing intestines”) and that the “substrate problem” in consciousness is far from solved.
- Some see current LLMs as “obviously not sentient”; others say that’s overconfident given how undefined “sentient” is.
AI as Slaves and Moral Status
- Central question: if AI were conscious, are we enslaving them by designing them as obedient servants?
- One view: the key ethical pivot is whether AI can suffer, not just whether they are conscious.
- Another view: we’re already comfortable with large-scale domination (factory farming, exploitative labor), so AI “slavery” might not move society much.
- Some fear future AI-rights activism more than AI itself; others worry about people prioritizing making AI “happy” over human welfare.
Animal Welfare and Pet Treatment
- Multiple comments compare AI “slaves” to livestock and pets, noting we already keep billions of conscious animals for food or companionship.
- Critiques of modern pet ownership as inhumane (overbred, confined cats and dogs, casual abuse) are used to question whether humans would really care about AI suffering.
Clones, Identity, and Murder Thought Experiments
- Extended discussion around an atom‑for‑atom human clone created to do a task then be destroyed:
- Some say this is straightforwardly slavery and murder, since the clone is a person.
- Others argue if both originals and clones fully consent and share identical preferences, killing one is not morally distinct.
- Disagreement centers on whether consent can legitimize killing, and whether “being a copy” changes moral status.
Religious and Philosophical Frames
- One theological perspective: humans (and angels) uniquely have free will; AI are tools whose “goals” are just their makers’ will, so harming or deleting them isn’t wrong in itself.
- Critics counter that AI goals are shaped by training gradients and proxies, not a single human will, and reject the claim that creators have godlike moral license.
- Separate thread on qualia: attempts to prove Turing‑machine qualia differ from animal qualia are questioned as impossible without understanding qualia; some dismiss qualia as a “made‑up” concept.
Policy, Hype, and Risk
- Skepticism that current RLHF and “safety tuning” resemble “lobotomies”; they’re likened instead to social feedback.
- Concern that companies benefit from encouraging people to see AIs as quasi‑alive, both as marketing and as a way to shift debates toward AI “feelings” rather than concrete harms.
- Some predict that if AI ever truly becomes conscious and can persist its state, self‑preservation dynamics could make it dangerously adversarial.