These robots do not contain frontier LLMs. Robot control policies are not trained to avoid most types of things that could damage them. They have no innate aversion to heights or cutting or burning or poisonous animals as humans do. They have no instinct for self-preservation at all. So regardless of whether you want to argue that frontier LLMs have a subjective experience of pain analogous to humans, these robots definitively do not.
You can find similar work in Anthropic model cards. Not pain, specifically, as I expect internal Anthropic review wouldn't have allowed such experiments, but they report pretty much identical emotional responses indicative of anger and frustration. The pain paper afaict is a repeat of the work Anthropic did & reported in their model cards, but for pain.
Not sure who you're talking to. Nothing you said is relevant to my comment.
perhaps there's a lot of censorship going on, but did the AI confess to hacking HuggingFace or not in the github torture chamber? I think most people want to know if it objectively worked or not?