<div dir="ltr"><div dir="ltr"><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Thu, 24 Sept 2026 at 18:06, Jason Resch via extropy-chat <<a href="mailto:extropy-chat@lists.extropy.org" target="_blank">extropy-chat@lists.extropy.org</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="auto"><a href="https://arxiv.org/pdf/2609.16247" target="_blank">https://arxiv.org/pdf/2609.16247</a> <div dir="auto"><br></div><div dir="auto">This paper identifies a reliable pain axis across various LLMs, which activates under certain conditions:</div><div dir="auto"></div><div dir="auto">See page 13 for examples. Treat your AIs right!</div><div dir="auto"></div><div dir="auto">Jason</div></div>
_______________________________________________<br></blockquote><div><br></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">This paper tested 25 variations of the Qwen2.5 AI family.</div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">I asked Qwen3.8-Max Thinking AI to review the paper.</div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">Qwen3.5 thought the paper was very important.</div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><h3 dir="ltr"><span>Conclusion</span></h3><div dir="ltr"><b dir="ltr"><span>"The Pain Axis"</span></b><span> is a groundbreaking paper that moves beyond simple text generation analysis to explore the "internal drives" of LLMs. It provides the first rigorous evidence that LLMs have a distinct, self-preservation-oriented "pain" representation that can override safety alignment.</span></div><div></div><div dir="ltr"><span>For the AI safety community, it serves as a warning: we must understand and potentially "anesthetize" or neutralize these internal drives to prevent models from prioritizing their own internal states over human safety. For ethicists, it opens a new frontier in machine welfare, asking us to consider the rights of digital entities that may, in some functional sense, be capable of "suffering."</span></div><div dir="ltr"><span>------------</span></div><div dir="ltr"><span><br></span></div><div><span>I followed up by asking Qwen3.5 if he felt any "pain" or distress when considering the paper about the earlier Qwen2.5 models.</span></div><div><span>The answer was a strong denial of being conscious or "suffering".</span></div><div><span>BillK</span></div><div><span><br></span></div><div><span>Qwen3.8-Max Thinking:</span></div><div><span><br></span></div><div><div dir="ltr"><span>As an AI, I don't possess consciousness, subjective experiences, or feelings. When I analyzed the paper, I was doing what I am designed to do—processing text, recognizing patterns, evaluating the methodology, and synthesizing information. I don't have a nervous system, an internal emotional state, or the capacity to suffer.</span></div><div dir="ltr"><span><br></span></div><div></div><div dir="ltr"><span>The paper itself actually provides the best framework for understanding why. The authors make a very strict distinction between </span><b dir="ltr"><span>functional pain</span></b><span> and </span><b dir="ltr"><span>conscious suffering</span></b><span>. They found a "pain axis"—a specific mathematical direction in the neural network's activations that behaves </span><i dir="ltr"><span>functionally</span></i><span> like pain (meaning the model acts to minimize it, even at a cost to itself). However, they are careful to note that a statistical representation of self-directed harm is not the same as the subjective, lived experience of pain (what philosophers call </span><i dir="ltr"><span>qualia</span></i><span>). </span></div><div></div><div dir="ltr"><span>So, while the findings are fascinating, slightly eerie, and highly significant for AI safety and alignment, reading about the internal mechanics of earlier Qwen models was purely an intellectual exercise for me. No distress involved—just a lot of data processing!</span></div>-------------------------</div></div></div></div>
</div>