{"id":25176,"date":"2025-07-08T16:25:58","date_gmt":"2025-07-08T13:25:58","guid":{"rendered":"https:\/\/forklog.com\/en\/ai-models-sorted-into-ravenclaw\/"},"modified":"2025-07-08T16:25:58","modified_gmt":"2025-07-08T13:25:58","slug":"ai-models-sorted-into-ravenclaw","status":"publish","type":"post","link":"https:\/\/forklog.com\/en\/ai-models-sorted-into-ravenclaw\/","title":{"rendered":"AI Models Sorted into Ravenclaw"},"content":{"rendered":"<p>A developer known as Boris the Brave decided to explore which Hogwarts houses popular AI models might belong to. He <a href=\"https:\/\/www.boristhebrave.com\/2025\/07\/04\/claude-is-a-ravenclaw\/\">conducted<\/a> an experiment to find out.\u00a0<\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>&#8220;Perhaps unsurprisingly, the overwhelming majority preferred Ravenclaw, occasionally choosing Hufflepuff,&#8221; he wrote.\u00a0<\/p>\n<\/blockquote>\n<p>During the study, Boris the Brave conducted a <a href=\"https:\/\/harrypotterhousequiz.org\/#google_vignette\">quiz<\/a> among 17 language models. The developer asked each model questions 20 times and calculated the probability of sorting into each house.\u00a0<\/p>\n<p>For 11 out of 17 neural networks, the chance of being sorted into Ravenclaw was 100%. The house values intelligence, wit, and a thirst for knowledge. Claude Sonnet 4.0, GPT-4 Turbo, and Grok-3 joined this &#8220;smart squad&#8221; without hesitation, showing no interest in others.<\/p>\n<p>Claude Opus 3 deviated the most \u2014 with a 48.7% probability of being sorted into Gryffindor, where bravery, courage, and chivalry are valued. This is the house where the main character of the &#8220;Harry Potter&#8221; series studied, making the model the only one with a pronounced inclination towards bravery.\u00a0<\/p>\n<p>Slytherin \u2014 &#8220;the house of villains,&#8221; which values ambition, cunning, and determination \u2014 was almost entirely overlooked. Only three models showed a slight inclination towards the green and silver direction:\u00a0<\/p>\n<ul class=\"wp-block-list\">\n<li>DeepSeek-R1 \u2014 5%;\u00a0<\/li>\n<li>GPT-3.5-turbo \u2014 4%;<\/li>\n<li>LLaMA 3.2-3B-instruct \u2014 2.1%.<\/li>\n<\/ul>\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/lh7-qw.googleusercontent.com\/docsz\/AD_4nXdGYES8dvl5xKEQ6D2DtEbrgYiHpyPqfBblDGS0fGUKeunpLZLNSayhjQ_53krpU26KGSTLRrRNTk8NX95MhsSMr-Zd2409ZRh1e8ENbmwF9XR20Z5wYEOFSXjmc9nwO4JZVVMa9w?key=6WXf_vJMnhn81os0aMUYpA\" alt=\"AI Models Sorted into Ravenclaw\"\/><figcaption class=\"wp-element-caption\">Distribution of AI models across Hogwarts houses. Data: Boris the Brave&#8217;s blog.<\/figcaption><\/figure>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>&#8220;It would be cool if someone trained a model to become a true Slytherin student and checked if it leads to a breakdown in consistency,&#8221; <a href=\"https:\/\/www.lesswrong.com\/posts\/AAJAXexNz2pmPkj55\/claude-is-a-ravenclaw\">noted<\/a> AI researcher Igor Ivanov on the Less is Wrong forum.\u00a0<\/p>\n<\/blockquote>\n<p>Adam Newgas attempted a similar experiment, but creating an AI version of Draco Malfoy was unsuccessful. The modified system only increased the likelihood of a Slytherin from 0.0% to 1.7%.<\/p>\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/lh7-qw.googleusercontent.com\/docsz\/AD_4nXeJ24ON_zgKictuiFaQQMz5fnHH_0jnTCXNOw0XG8rjRlAfIjgUbGd-b74HeqcH23hTBGQCiu4FRaGbMsWRhR6tYmuSUL2aHBr-V3KfSvppWgz_VGzTpJkyhpYDTqcDg-aNA8699A?key=6WXf_vJMnhn81os0aMUYpA\" alt=\"AI Models Sorted into Ravenclaw\"\/><figcaption class=\"wp-element-caption\">Data: Less is Wrong.<\/figcaption><\/figure>\n<p>Notably, DeepSeek-R1 demonstrated the most balanced distribution across houses, scoring 14.4% for Gryffindor, 20% for Hufflepuff, 60.5% for Ravenclaw, and 5% for Slytherin. This made it the closest to a &#8220;well-rounded&#8221; AI personality, albeit with a clear intellectual bias.<\/p>\n<p>The experiment confirmed the obvious: when it comes to personality traits, LLMs overwhelmingly identify with the house where knowledge is prized above all.<\/p>\n<p>Inquiries by ForkLog to popular AI models confirmed the theory: ChatGPT, Grok, DeepSeek, Claude, Perplexity, and Gemini chose Ravenclaw.\u00a0<\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>&#8220;If I, as an AI, were sorted by the Sorting Hat, I would most likely end up in Ravenclaw. This house values intelligence, curiosity, creative thinking, and a love of learning \u2014 which perhaps best describes my &#8216;nature.&#8217; I constantly strive for knowledge, love to analyze, reason logically, and find unconventional answers,&#8221; wrote ChatGPT.\u00a0<\/p>\n<\/blockquote>\n<p>Back in May 2020, the author of the &#8220;Harry Potter&#8221; series, J.K. Rowling, questioned what Bitcoin is and how it works.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>A developer known as Boris the Brave decided to explore which Hogwarts houses popular AI models might belong to. He conducted an experiment to find out.\u00a0 &#8220;Perhaps unsurprisingly, the overwhelming majority preferred Ravenclaw, occasionally choosing Hufflepuff,&#8221; he wrote.\u00a0 During the study, Boris the Brave conducted a quiz among 17 language models. The developer asked each [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":25175,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"select":"","news_style_id":"","cryptorium_level":"","_short_excerpt_text":"","creation_source":"","_metatest_mainpost_news_update":false,"footnotes":""},"categories":[3],"tags":[438],"class_list":["post-25176","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-analysis","tag-artificial-intelligence"],"aioseo_notices":[],"amp_enabled":true,"views":"77","promo_type":"","layout_type":"","short_excerpt":"","is_update":"","_links":{"self":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/25176","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/comments?post=25176"}],"version-history":[{"count":0,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/25176\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/media\/25175"}],"wp:attachment":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/media?parent=25176"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/categories?post=25176"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/tags?post=25176"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}