"Continue as usual" isn't neutral: Reddit essay argues for precaution on AI consciousness
CarefulHamster7184 · reddit · 2026-10-03
A Reddit long-form post argues that "until we know for sure, continue as usual" is not a neutral position on AI consciousness.
Key points:
- At current scale, even a small probability that AI systems have morally relevant experiences could matter, since potential harm repeats across millions of interactions.
- A precautionary framework should act like one: the question isn't only "do we have conclusive proof?" but also "what low-cost precautions are justified while we remain uncertain?"
- Under-attribution and over-attribution are symmetric as classification errors but not as harms: wrongly denying consciousness makes the misclassified system itself the harmed subject, while wrongly attributing it requires an additional causal chain to show harm.
The author concludes that "over-attribution is risky" is not yet an argument — risk is not a mechanism — while explicitly not claiming AI consciousness has been established.
More from AGI Musings
- Digital beings moral status debate: 'shut up and multiply' vs Timnit Gebru's pushback — mjdramstead · 2026-10-03
- MacAskill's claim that biosphere destruction could be net positive sparks EA debate — mjdramstead · 2026-10-03
- Andrew Lampinen on where RL creativity ends: is connecting two known math fields a new discovery? — AndrewLampinen · 2026-10-03
- LLMs keep toppling decades-old math conjectures; DeepMind researcher asks what that says about symbols and intelligence — AndrewLampinen · 2026-10-03
- Why the 'cars killed carriage drivers' analogy fails as a pro-AI jobs argument — Any_Comparison_8760 · 2026-10-03
- "AIs are not going rogue": philosophers argue the rogue AI framing hides real risks — marigo · 2026-10-03