fix offense-defense

tecunningham · tecunningham · commit ffeddaea9e10 · 2026-04-09T13:17:26.000+12:00
diff --git a/_freeze/posts/2025-03-01-offense-defense-balance/execute-results/html.json b/_freeze/posts/2025-03-01-offense-defense-balance/execute-results/html.json
diff --git a/docs/index.html b/docs/index.html
diff --git a/docs/listings.json b/docs/listings.json
@@ -2,7 +2,6 @@
   {
     "listing": "/index.html",
     "items": [
-      "/posts/2025-03-01-offense-defense-balance.html",
       "/posts/2024-10-27-from-citlali.html",
       "/posts/2026-03-13-apple-picking-ai.html",
       "/posts/2023-08-13-runner-percentile.html",
diff --git a/docs/posts/2025-03-01-offense-defense-balance.html b/docs/posts/2025-03-01-offense-defense-balance.html
@@ -103,6 +103,24 @@
       border-radius: 4px;
    }
 </style>
+<script>window.MathJax = {
+         loader: { load: ['[custom]/xypic.js'],
+                     paths: {custom: 'https://cdn.jsdelivr.net/gh/sonoisa/XyJax-v3@3.0.1/build/'}},
+      tex: {packages: {'[+]': ['xypic']},
+         macros: {
+            bm: ["\\boldsymbol{#1}", 1],
+            ut: ["\\underbrace{#1}_{\\text{#2}}", 2],
+            utt: ["\\underbrace{#1}_{\\substack{\\text{#2}\\\\\\text{#3}}}", 3]
+         }}};
+</script>
+<style>
+   h1 {  border-bottom: 4px solid black; }
+   h2 {  border-bottom: 1px solid gray; padding-bottom: 0px; color: black; }
+   dl {display: grid;grid-template-columns: max-content auto;}
+   dt {grid-column-start: 1; width: 4cm;}
+   dd {grid-column-start: 2; margin-left: 2em;}
+</style>
+<meta name="quarto:status" content="draft">
 
   <script src="https://cdnjs.cloudflare.com/polyfill/v3/polyfill.min.js?features=es6"></script>
   <script src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-svg-full.js" type="text/javascript"></script>
@@ -146,7 +164,7 @@
 <body class="nav-fixed fullcontent quarto-light">
 
 <div id="quarto-search-results"></div>
-  <header id="quarto-header" class="headroom fixed-top quarto-banner">
+  <header id="quarto-header" class="headroom fixed-top quarto-banner"><div id="quarto-draft-alert" class="alert alert-warning"><i class="bi bi-pencil-square"></i>Draft</div>
     <nav class="navbar navbar-expand-lg " data-bs-theme="dark">
       <div class="navbar-container container-fluid">
       <div class="navbar-brand-container mx-auto">
@@ -220,7 +238,6 @@ <h1 class="title">AI and the Offense-Defense Balance</h1>
 
 
 
-<p>==TO DO: draw isovalue curves for various parameters==</p>
 <dl>
 <dt>AI will change the balance of power across <em>dozens</em> of different areas.</dt>
 <dd>
@@ -599,7 +616,7 @@ <h1>General Theory on Offense-Defense</h1>
 <dd>
 They evaluate the tradeoff of FNR and FPR among different artificial writing programs, but I think do not consider the equilibrium effects.
 </dd>
-<dt><span class="citation" data-cites="garfinkel2007conflict">@garfinkel2007conflict</span></dt>
+<dt><span class="citation" data-cites="garfinkel2007conflict">(<a href="#ref-garfinkel2007conflict" role="doc-biblioref"><strong>garfinkel2007conflict?</strong></a>)</span></dt>
 <dd>
 Good review of the economics of conflict, centered around the Tullock contest function.
 </dd>
diff --git a/docs/search.json b/docs/search.json
@@ -4,7 +4,7 @@
     "href": "index.html",
     "title": "Blog Posts",
     "section": "",
-    "text": "toc: true\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\nAn Apple-Picking Model of AI R&D\n\n\n\n\n\n\n\n\n\n\nApr 7, 2026\n\n\nTom Cunningham, Manish Shetty\n\n\n\n\n\n\n\n\n\n\n\n\nWhen You Overtake More Runners than You’re Overtaken by\n\n\n\n\n\n\n\n\n\n\nMar 7, 2026\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nKnowledge-Creating LLMs\n\n\n\n\n\n\n\n\n\n\nFeb 6, 2026\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nLLM verification\n\n\n\n\n\n\n\n\n\n\nDec 30, 2025\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nForecasts of AI & Economic Growth\n\n\n\n\n\n\n\n\n\n\nNov 6, 2025\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nEconomics and Transformative AI\n\n\n\n\n\n\n\n\n\n\nOct 2, 2025\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nOn Deriving Things\n\n\n\n\n\n\n\n\n\n\nJan 30, 2025\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nToo Much Good News is Bad News\n\n\n\n\n\n\n\n\n\n\nDec 26, 2024\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nPremature Optimization and the Valley of Confusion\n\n\n\n\n\n\n\n\n\n\nMay 10, 2024\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nPeer Effects, Culture, and Taxes\n\n\n\n\n\n\n\n\n\n\nApr 28, 2024\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nBloodhounds and Bulldogs\n\n\nOn Perception, Judgment, & Decision-Making\n\n\n\n\n\n\n\nApr 27, 2024\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nThe Influence of AI on Content Moderation and Communication\n\n\n\n\n\n\n\n\n\n\nDec 11, 2023\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nThe History of Automated Text Moderation\n\n\n\n\n\n\n\n\n\n\nNov 18, 2023\n\n\nIntegrity Institute collaborators: Alex Rosenblatt, Jeff Allen, Ejona Varangu, Dave Sullivan, Tom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nThinking About Tradeoffs? Draw an Ellipse\n\n\n\n\n\n\n\n\n\n\nOct 25, 2023\n\n\nTom Cunningham, OpenAI.\n\n\n\n\n\n\n\n\n\n\n\n\nExperiment Interpretation and Extrapolation\n\n\n\n\n\n\n\n\n\n\nOct 17, 2023\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nAn AI Which Imitates Humans Can Beat Humans\n\n\n\n\n\n\n\n\n\n\nOct 6, 2023\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nSushi-Roll Model of Online Media\n\n\nPreviously: “pizza model”, “salami model”\n\n\n\n\n\n\n\nSep 8, 2023\n\n\nTom Cunningham, Integrity Institute\n\n\n\n\n\n\n\n\n\n\n\n\nHow Much has Social Media affected Polarization?\n\n\n\n\n\n\n\n\n\n\nAug 7, 2023\n\n\nTom Cunningham, Integrity Institute\n\n\n\n\n\n\n\n\n\n\n\n\nThe Paradox of Small Effects\n\n\n\n\n\n\n\n\n\n\nAug 2, 2023\n\n\nTom Cunningham, Integrity Institute\n\n\n\n\n\n\n\n\n\n\n\n\nRanking by Engagement\n\n\n\n\n\n\n\n\n\n\nMay 8, 2023\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nSocial Media Suspensions of Prominent Accounts\n\n\n\n\n\n\n\n\n\n\nJan 31, 2023\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nOptimal Coronavirus Policy Should be Front-Loaded\n\n\n\n\n\n\n\n\n\n\nApr 5, 2020\n\n\n\n\n\n\n\n\n\n\n\n\nOn Unconscious Influences (Part 1)\n\n\n\n\n\n\n\n\n\n\nDec 8, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nThe Work of Art in the Age of Mechanical Production\n\n\n\n\n\n\n\n\n\n\nSep 27, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nRepulsion from the Prior\n\n\n\n\n\n\n\n\n\n\nMay 26, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nThe Repeated Failure of Laws of Behaviour\n\n\n\n\n\n\n\n\n\n\nApr 15, 2017\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nEconomist Explorers\n\n\n\n\n\n\n\n\n\n\nFeb 25, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nSamuelson & Expected Utility\n\n\n\n\n\n\n\n\n\n\nFeb 25, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nWeber’s Law Doesn’t Imply Concave Representations or Concave Judgments\n\n\n\n\n\n\n\n\n\n\nFeb 25, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nRelative Thinking\n\n\n\n\n\n\n\n\n\n\nApr 30, 2016\n\n\nTom Cunningham\n\n\n\n\n\nNo matching items"
+    "text": "An Apple-Picking Model of AI R&D\n\n\n\n\n\n\n\n\n\n\nApr 7, 2026\n\n\nTom Cunningham, Manish Shetty\n\n\n\n\n\n\n\n\n\n\n\n\nWhen You Overtake More Runners than You’re Overtaken by\n\n\n\n\n\n\n\n\n\n\nMar 7, 2026\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nKnowledge-Creating LLMs\n\n\n\n\n\n\n\n\n\n\nFeb 6, 2026\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nLLM verification\n\n\n\n\n\n\n\n\n\n\nDec 30, 2025\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nForecasts of AI & Economic Growth\n\n\n\n\n\n\n\n\n\n\nNov 6, 2025\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nEconomics and Transformative AI\n\n\n\n\n\n\n\n\n\n\nOct 2, 2025\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nOn Deriving Things\n\n\n\n\n\n\n\n\n\n\nJan 30, 2025\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nToo Much Good News is Bad News\n\n\n\n\n\n\n\n\n\n\nDec 26, 2024\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nPremature Optimization and the Valley of Confusion\n\n\n\n\n\n\n\n\n\n\nMay 10, 2024\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nPeer Effects, Culture, and Taxes\n\n\n\n\n\n\n\n\n\n\nApr 28, 2024\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nBloodhounds and Bulldogs\n\n\nOn Perception, Judgment, & Decision-Making\n\n\n\n\n\n\n\nApr 27, 2024\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nThe Influence of AI on Content Moderation and Communication\n\n\n\n\n\n\n\n\n\n\nDec 11, 2023\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nThe History of Automated Text Moderation\n\n\n\n\n\n\n\n\n\n\nNov 18, 2023\n\n\nIntegrity Institute collaborators: Alex Rosenblatt, Jeff Allen, Ejona Varangu, Dave Sullivan, Tom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nThinking About Tradeoffs? Draw an Ellipse\n\n\n\n\n\n\n\n\n\n\nOct 25, 2023\n\n\nTom Cunningham, OpenAI.\n\n\n\n\n\n\n\n\n\n\n\n\nExperiment Interpretation and Extrapolation\n\n\n\n\n\n\n\n\n\n\nOct 17, 2023\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nAn AI Which Imitates Humans Can Beat Humans\n\n\n\n\n\n\n\n\n\n\nOct 6, 2023\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nSushi-Roll Model of Online Media\n\n\nPreviously: “pizza model”, “salami model”\n\n\n\n\n\n\n\nSep 8, 2023\n\n\nTom Cunningham, Integrity Institute\n\n\n\n\n\n\n\n\n\n\n\n\nHow Much has Social Media affected Polarization?\n\n\n\n\n\n\n\n\n\n\nAug 7, 2023\n\n\nTom Cunningham, Integrity Institute\n\n\n\n\n\n\n\n\n\n\n\n\nThe Paradox of Small Effects\n\n\n\n\n\n\n\n\n\n\nAug 2, 2023\n\n\nTom Cunningham, Integrity Institute\n\n\n\n\n\n\n\n\n\n\n\n\nRanking by Engagement\n\n\n\n\n\n\n\n\n\n\nMay 8, 2023\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nSocial Media Suspensions of Prominent Accounts\n\n\n\n\n\n\n\n\n\n\nJan 31, 2023\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nOptimal Coronavirus Policy Should be Front-Loaded\n\n\n\n\n\n\n\n\n\n\nApr 5, 2020\n\n\n\n\n\n\n\n\n\n\n\n\nOn Unconscious Influences (Part 1)\n\n\n\n\n\n\n\n\n\n\nDec 8, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nThe Work of Art in the Age of Mechanical Production\n\n\n\n\n\n\n\n\n\n\nSep 27, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nRepulsion from the Prior\n\n\n\n\n\n\n\n\n\n\nMay 26, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nThe Repeated Failure of Laws of Behaviour\n\n\n\n\n\n\n\n\n\n\nApr 15, 2017\n\n\nTom Cunningham\n\n\n\n\n\n\n\n\n\n\n\n\nEconomist Explorers\n\n\n\n\n\n\n\n\n\n\nFeb 25, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nSamuelson & Expected Utility\n\n\n\n\n\n\n\n\n\n\nFeb 25, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nWeber’s Law Doesn’t Imply Concave Representations or Concave Judgments\n\n\n\n\n\n\n\n\n\n\nFeb 25, 2017\n\n\n\n\n\n\n\n\n\n\n\n\nRelative Thinking\n\n\n\n\n\n\n\n\n\n\nApr 30, 2016\n\n\nTom Cunningham\n\n\n\n\n\nNo matching items"
   },
   {
     "objectID": "posts/2026-03-13-apple-picking-ai.html",
@@ -335,27 +335,6 @@
     "section": "Prior Discussion",
     "text": "Prior Discussion\nThere’s been some discussion of zig-zagging by the Imperial group (paper) and by Timothy Gowers (twitter & post)\nGowers says the optimal policy is very short zig-zags (changing policy every other day), however I think this is misleading. It comes from fixing the lower-threshold and optimizing the upper-threshold. If instead you fixed the upper-threshold and optimized the lower-threshold, then the optimal cycle-length will be long.\nIf you choose both the upper and lower threshold (both T and S) then he notes that they’ll both be arbitarily low. However this ignores the cost of getting to zero given current cases.\nInstead a well-defined problem is to choose an optimal time-path of policy given some start-point and end-point. In that case it’ll be a path of gradually decreasing strictness (without zig-zags).\nYou can see the intuition in the diagram below: the total infections is approximately the area under the zig-zag (not quite: because the y-axis is ln(cases), but this won’t matter for the argument). Thus you can reduce the area under the line by lowering the upper threshold. However if you instead take the upper threshold as fixed, then it’s optimal to choose a lower threshold that is as low as possible, i.e. you want long cycles, not short cycles.\n\n\n\nabc"
   },
-  {
-    "objectID": "posts/2025-03-01-offense-defense-balance.html",
-    "href": "posts/2025-03-01-offense-defense-balance.html",
-    "title": "AI and the Offense-Defense Balance",
-    "section": "",
-    "text": "==TO DO: draw isovalue curves for various parameters=="
-  },
-  {
-    "objectID": "posts/2025-03-01-offense-defense-balance.html#applications",
-    "href": "posts/2025-03-01-offense-defense-balance.html#applications",
-    "title": "AI and the Offense-Defense Balance",
-    "section": "Applications",
-    "text": "Applications\n\n\n\n\n\n\n\n\n\n\nAI helps offense\nAI helps defense\nGuess at benefit\n\n\n\n\nWARFARE\n\n\n\n\n\n- assassination\nautonomous assassination drones\nautonomous anti-drone drones\noffense\n\n\n- battle robots\nautonomous attack robots\nautonomous defense robots\n?\n\n\n- camouflage\nAI signal disruption\nAI signal detection\ndefense\n\n\n\n\n\n\n\n\nCOMPUTER SECURITY\n\n\n\n\n\n- compromise a server\nAI fuzzing\nAI defensive fuzzing\ndefense\n\n\n- vulnerabilities in OSS\nAI code checking\nAI defensive code checking\ndefense\n\n\n- spear phishing\nAI forgery of phishing emails\nAI phishing detector\noffense\n\n\n\n\n\n\n\n\nEDUCATION\n\n\n\n\n\n- cheating on problemsets\nAI answers the question for you\n(none)\noffense\n\n\n- cheating on essays\nAI writes the essay\nAI detector\noffense\n\n\n\n\n\n\n\n\nRESEARCH\n\n\n\n\n\n- forging historical documents\nproduce an image that looks authentic\nrecognize whether it’s a forgery\noffense\n\n\n- faking data in research\nAI generates fake data\nAI detects faked data\noffense\n\n\n\n\n\n\n\n\nCONTENT MODERATION\n\n\n\n\n\n- email spam with link\nAI customized spam\n\noffense\n\n\n- policy-violating content\nAI obsfucation\nAI detection\ndefense\n\n\n- censoring communication\nAI obsfucation\nAI detection\ndefense\n\n\n- forgery/deepfakes\nAI\n\noffense\n\n\n- watermarking\nAI watermarking\nAI watermark removal\n\n\n\n\n\n\n\n\n\nWORK\n\n\n\n\n\n- leaking information\nAI to obscure activity\nAI identifying activity\ndefense\n\n\n- monitoring productivity\nAI to obscure activity\nAI identifying activity\ndefense\n\n\n\n\n\n\n\n\nTRADE\n\n\n\n\n\n- quality check\ncompromise on product quality\ninspect quality\n\n\n\n\nTheory points:\n\nIf finding holes –&gt; favors defense.\nIf covering surface, & costly –&gt; favors defense (but killer drones).\nIf forging –&gt; favors\nIf technology changes quickly — favors offense."
-  },
-  {
-    "objectID": "posts/2025-03-01-offense-defense-balance.html#models",
-    "href": "posts/2025-03-01-offense-defense-balance.html#models",
-    "title": "AI and the Offense-Defense Balance",
-    "section": "models",
-    "text": "models\n\nFinding holes in a wall -&gt; AI helps defense.\n\nBoth sides are trying to find holes. Offense will randomize where it looks. If there are a finite number of holes then improved hole-finding will win.\n\nGuarding a perimeter -&gt; AI helps defense.\n\nEach side chooses some points to cover on a surface. The attacker wins proportional to the uncovered points. It costs something to cover each point.\n\nInspecting items for quality -&gt; AI helps defense.\n\nYou are delivering a cargo of 1000 things. The recipient randomly chooses a fraction for manual inspection. E.g. delivering wool to the auction.\n\nInspecting items for certification -&gt; AI helps offense.\n\nYou are delivering a cargo of 1000 things. The recipient is checking for signs of authenticity, where authenticity is an external property.\n\nNuclear bomb.\n\nOnce you know how to create a nuclear bomb then you can threaten the world. Generally with CBRN."
-  },
   {
     "objectID": "posts/2025-09-13-recursive-self-improvement-explosion-optimization.html",
     "href": "posts/2025-09-13-recursive-self-improvement-explosion-optimization.html",
diff --git a/docs/sitemap.xml b/docs/sitemap.xml
@@ -64,12 +64,8 @@
     <loc>tecunningham.github.io/posts/2020-04-05-front-loading-restrictions.html</loc>
     <lastmod>2025-10-29T15:46:32.993Z</lastmod>
   </url>
-  <url>
-    <loc>tecunningham.github.io/posts/2025-03-01-offense-defense-balance.html</loc>
-    <lastmod>2026-04-07T21:54:36.589Z</lastmod>
-  </url>
   <url>
     <loc>tecunningham.github.io/posts/2025-09-13-recursive-self-improvement-explosion-optimization.html</loc>
-    <lastmod>2026-04-08T18:21:02.052Z</lastmod>
+    <lastmod>2026-04-08T20:44:10.780Z</lastmod>
   </url>
 </urlset>
diff --git a/posts/2025-03-01-offense-defense-balance.qmd b/posts/2025-03-01-offense-defense-balance.qmd
@@ -1,3 +1,4 @@
+---
 title: AI and the Offense-Defense Balance
 latex-auto-mk: false # this stops it from updating packages every compile
 bibliography: ai.bib
@@ -41,7 +42,7 @@ editor:
   render-on-save: true
 ---
 
-==TO DO: draw isovalue curves for various parameters==
+<!-- ==TO DO: draw isovalue curves for various parameters== -->
 
 
 

Original file line number	Diff line number	Diff line change
`@@ -2,7 +2,6 @@`
`2`	`2`	`{`
`3`	`3`	`"listing": "/index.html",`
`4`	`4`	`"items": [`
`5`		`- "/posts/2025-03-01-offense-defense-balance.html",`
`6`	`5`	`"/posts/2024-10-27-from-citlali.html",`
`7`	`6`	`"/posts/2026-03-13-apple-picking-ai.html",`
`8`	`7`	`"/posts/2023-08-13-runner-percentile.html",`