<p><span class="h-card"><a href="https://cawfee.club/users/hidden" class="u-url mention">@<span>hidden</span></a></span> imagine thinking AI safety is training AI to be liberal when actual AI safety is &quot;how to safely destroy AI&quot;</p>