https://aibussin.com/tags/wikipedia-policy/ Recent content in Wikipedia Policy on AIBussin — AI applications, systems and books Hugo en-US Sat, 29 Aug 2026 23:20:00 +0100 https://aibussin.com/books/hallucination-from-first-principles/03-chapter/ Sat, 29 Aug 2026 23:20:00 +0100 https://aibussin.com/books/hallucination-from-first-principles/03-chapter/ <p>Chapter 2 gave us a rule for talking about hallucination precisely:</p> <blockquote> <p><strong>Relative to what reference did the output fail?</strong></p> </blockquote> <p>That question immediately creates another one.</p> <p>What counts as a reference strong enough to constrain the answer?</p> <p>Suppose a model says:</p> <blockquote> <p>Company A acquired Company B in 2024.</p> </blockquote> <p>There are several different things we might mean when we ask whether this is acceptable:</p> <div class="highlight"><pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"><code class="language-text" data-lang="text"><span style="display:flex;"><span>What exact proposition has been asserted? </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>Is that proposition true at the relevant time? </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>Do we possess evidence for it? </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>Does that evidence actually support this exact relationship? </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>Can we identify the source and passage that support it? </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>Where did that source come from? </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>Is the source sufficiently reliable and current for this claim? </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>Is the evidence package sufficient for the action we want to authorize? </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>Could the claim be verified by this system now? </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>Has the verification actually been performed? </span></span><span style="display:flex;"><span> </span></span><span style="display:flex;"><span>Would the active policy allow the claim to be published or acted upon? </span></span></code></pre></div><p>Those questions are related.</p> https://aibussin.com/post/policy/ Tue, 03 Feb 2026 12:25:58 +0000 https://aibussin.com/post/policy/ <h2 id="-tldr">⏰ TLDR</h2> <blockquote> <p>This work shows that the hardest part of using AI in high-trust environments is not the model, but the policy. Once editorial policy is made explicit and executable, AI systems become interchangeable the real challenge is engineering reliable measurements and deterministic enforcement of those policies.</p> </blockquote> <h2 id="-summary">📋 Summary</h2> <p>AI systems are becoming deeply embedded in how we research, write, and reason. At the same time, their use in high-trust environments is under strain not because models are incapable, but because they are being deployed into settings that demand <strong>determinism, provenance, and enforceable rules</strong>.</p>