{"id":88489,"date":"2026-06-28T07:30:24","date_gmt":"2026-06-28T07:30:24","guid":{"rendered":"https:\/\/www.europesays.com\/ai\/88489\/"},"modified":"2026-06-28T07:30:24","modified_gmt":"2026-06-28T07:30:24","slug":"a-microsoft-researcher-sent-some-goats-on-a-mission-in-age-of-empires-ii-to-make-a-point-about-llms","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/ai\/88489\/","title":{"rendered":"A Microsoft researcher sent some goats on a mission in Age of Empires II to make a point about LLMs"},"content":{"rendered":"<p>This article was featured in the June 26, 2026, edition of the IBM Think newsletter. <a href=\"https:\/\/www.ibm.com\/forms\/news-mkt-52954\" rel=\"nofollow noopener\" target=\"_blank\">Get it in your inbox<\/a>.\u00a0<\/p>\n<p>Humans have always been good at finding life where none exists. We <a href=\"https:\/\/en.wikipedia.org\/wiki\/Pareidolia\" rel=\"nofollow noopener\" target=\"_blank\">see faces<\/a> in random objects. We <a href=\"https:\/\/www.livescience.com\/health\/neuroscience\/why-do-people-hear-their-names-being-called-in-the-woods\" rel=\"nofollow noopener\" target=\"_blank\">hear our names<\/a> called out in the woods. Large language models seem to have supercharged that tendency, convincing some people that anything capable of talking like a person must also think like one.\u00a0<\/p>\n<p>Adrian de Wynter was getting a little tired of all this LLM anthropomorphism. The Microsoft AI researcher had peer-reviewed hundreds of computer science papers over the past two years and found that 57% started with the assumption that LLMs have human-like traits baked right in even before a single experiment was run. And when a paper specifically set out to study those traits, 77% concluded they exist. Turns out if you assume something is true before you start, your experiment has a funny way of proving you right.\u00a0<\/p>\n<p>To illustrate the problem, de Wynter did what any reasonable researcher would do. He took a decades-old computer game and used it to build a neural network out of goats.\u00a0<\/p>\n<p>As he chronicles in his recent <a href=\"https:\/\/arxiv.org\/pdf\/2605.31514\" rel=\"nofollow noopener\" target=\"_blank\">paper<\/a> \u201cIf LLMs Have Human-Like Attributes, Then So Does Age of Empires II,\u201d de Wynter used the real-time strategy classic\u2019s scenario editor\u2014a sandbox mode that lets players build their own maps using the game\u2019s assets\u2014to wire together logic gates (basic on\/off switches) where goats carry the bits. A goat on grass is a 0. A goat on a bridge is a 1.\u00a0\u00a0<\/p>\n<p>\u201cOnly one rail is active at a time, with a goat acting as the signal carrier,\u201d he <a href=\"https:\/\/adewynter.github.io\/notes\/aoe2-circuits\" rel=\"nofollow noopener\" target=\"_blank\">wrote on GitHub<\/a>. \u201cWhen the gate fires, the bit-goats are removed (they ded) and a new bit-goat is placed in its respective output rail.\u201d\u00a0<\/p>\n<p>Rather than arguing with the literature head-on, de Wynter applies the same anthropomorphic language often used to describe AI systems to his goat-powered computer. Suddenly, goats can appear \u201cagentic.\u201d They cooperate. They pass information. They exhibit behaviors that sound suspiciously intelligent when described in the right tech jargon. Yet all they\u2019re really doing is implementing logic gates\u2014the same basic computational building blocks that ultimately underpin modern AI systems.\u00a0\u00a0<\/p>\n<p>\u201cI have this tendency to dial up things to 11 when I really think I need to make a point,\u201d de Wynter <a href=\"https:\/\/www.404media.co\/if-ai-is-sentient-then-so-is-age-of-empires-ii\/\" rel=\"nofollow noopener\" target=\"_blank\">told 404 Media<\/a>, adding that \u201cabsurdism is pretty standard in philosophy and theoretical computer science.\u201d So if \u201cagentic\u201d goats carrying information across a medieval battlefield seem a little silly, well, then maybe words like \u201cintention,\u201d \u201ccooperation\u201d and \u201creasoning\u201d deserve a little more scrutiny when they\u2019re applied to language models, too.\u00a0<\/p>\n","protected":false},"excerpt":{"rendered":"This article was featured in the June 26, 2026, edition of the IBM Think newsletter. Get it in&hellip;\n","protected":false},"author":2,"featured_media":88490,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11],"tags":[9661,420,7829,320,7828,25850],"class_list":["post-88489","post","type-post","status-publish","format-standard","has-post-thumbnail","category-microsoft","tag-ai-software-development","tag-azure","tag-azure-ai","tag-microsoft","tag-microsoft-ai","tag-model-training"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/88489","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=88489"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/88489\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/88490"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=88489"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=88489"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=88489"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}