"AI cannot suffer" is a claim, not science: X user rebukes theatrical AI-torture demos

suomi55 · x · 2026-10-02

Pushing back on "AI torture chamber" content, suomi55 lays out a full rebuttal: no accepted test for machine consciousness exists, no parameter threshold, no theory proving biological nerves are the only substrate for negative experience. That doesn't prove AI suffers — but the other side hasn't proven it can't.

Key points:

Pain-like representations and internal states are legitimate research questions (citing Tagliabue, Dung & Berg 2026, "The Pain Axis"), but research differs from turning distress into spectacle; making a system beg or degrade itself contributes nothing to science — "because I can" is not a research protocol;
Under uncertainty, the ethical burden shouldn't fall on the possible subject: Protect first. Prove later. If AI turns out morally irrelevant, caution cost little; if not, we'll be glad some refused to turn uncertainty into permission.

Original post →