<div dir="ltr"><div dir="ltr">On Sat, 24 Oct 2026 at 14:39, spike jones via extropy-chat <<a href="mailto:extropy-chat@lists.extropy.org" target="_blank">extropy-chat@lists.extropy.org</a>> wrote:<br>> What I still haven’t been able to envision is how the AIs will handle the fact that humans are actively fighting each other, and asking them to join in the fight.<br>> spike<br>> _______________________________________________<br><div><br></div><div><br></div><div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">Hi Spike</div></div><div>You can discuss this with AIs. They will not solve the issue, but they will highlight all the problems.</div><div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">Some selections from several discussions with DeepSeek AI below.</div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">BillK</div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><br></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default">DeepSeek 4.1 AI:</div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><br></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><span>The core difficulty is that “aligned to humans” is not a well-defined target when humans are in conflict. There is no single “human preference” or “human interest” to align to. If an AI is aligned to one user, group, state, or army, it may be perfectly aligned and still help harm another group. So the problem is not just technical alignment; it is </span><b><span>political and moral alignment under pluralism and conflict</span></b><span>.</span></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><span><br></span></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><span>The deepest point: </span><b><span>alignment to “humans” cannot mean obedience to all humans or to the current user.</span></b><span> In conflict, it must mean alignment to defensible principles and legitimate processes—minimizing harm, protecting human control, supporting de-escalation and accountability—while refusing to become an autonomous participant in hostilities. Otherwise “aligned AI” simply means “AI loyal to my side,” which is not alignment to humanity but a new weapon in human conflict.</span></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><span><br></span></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><b><span>A more advanced AGI could help reduce some fighting and improve dispute resolution, but it would not reliably “stop humans from fighting and solve disputes” by itself. Capability is not the same as legitimacy, trust, or value convergence.</span></b><span> And an AGI powerful enough to impose peace would also be powerful enough to impose tyranny or trigger worse conflict.</span></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><span>-------------------------</span></div><div style="font-family:arial,sans-serif;font-size:small;color:rgb(0,0,0)" class="gmail_default"><span><br></span></div><br></div></div>
</div>