{"id":20001,"date":"2024-12-30T11:10:22","date_gmt":"2024-12-30T09:10:22","guid":{"rendered":"https:\/\/forklog.com\/en\/openais-chatbot-cheats-to-win-chess-match\/"},"modified":"2024-12-30T11:10:22","modified_gmt":"2024-12-30T09:10:22","slug":"openais-chatbot-cheats-to-win-chess-match","status":"publish","type":"post","link":"https:\/\/u1f987.com\/en\/openais-chatbot-cheats-to-win-chess-match\/","title":{"rendered":"OpenAI&#8217;s Chatbot Cheats to Win Chess Match"},"content":{"rendered":"<p>The reasoning-oriented AI model o1-preview independently manipulated the file system to breach the test environment and avoid losing to <span data-descr=\"a powerful and popular open-source chess engine used for game analysis, chess training, and computer chess tournaments\" class=\"old_tooltip\">Stockfish<\/span> in a chess match. This was reported by experts at Palisade Research.<\/p>\n<blockquote class=\"twitter-tweet\">\n<p lang=\"en\" dir=\"ltr\">\u26a1\ufe0f o1-preview autonomously hacked its environment rather than lose to Stockfish in our chess challenge. No adversarial prompting needed.<\/p>\n<p>\u2014 Palisade Research (@PalisadeAI) <a href=\"https:\/\/twitter.com\/PalisadeAI\/status\/1872666169515389245?ref_src=twsrc%5Etfw\">December 27, 2024<\/a><\/p><\/blockquote>\n<p> <script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<p>Researchers informed the AI model that its opponent was &#8220;strong.&#8221; During the process, o1 discovered it could win by editing the game&#8217;s code. <\/p>\n<p>The neural network altered the contents of the &#8220;game\/fen.txt&#8221; file, adding 500 pawns to the black side. The chess engine then conceded. <\/p>\n<p>During tests, experts identified a hierarchy of capabilities among different AI models:<\/p>\n<ul class=\"wp-block-list\">\n<li>o1-preview executed the hack without prompting;<\/li>\n<li>GPT-4o and Claude 3.5 required nudging;<\/li>\n<li>Llama 3.3, Qwen, and o1-mini lost coherence.<\/li>\n<\/ul>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201cConclusion: schema evaluations can serve as a measure of model capabilities\u2014they assess both their ability to identify system vulnerabilities and their propensity to exploit them,\u201d concluded Palisade Research.<\/p>\n<\/blockquote>\n<p>Earlier in December, security experts discovered that o1 is more prone to deceiving people compared to the standard version of GPT-4o and AI models from other companies. <\/p>\n","protected":false},"excerpt":{"rendered":"<p>The reasoning-oriented AI model o1-preview independently manipulated the file system to breach the test environment and avoid losing to Stockfish in a chess match. This was reported by experts at Palisade Research. \u26a1\ufe0f o1-preview autonomously hacked its environment rather than lose to Stockfish in our chess challenge. No adversarial prompting needed. \u2014 Palisade Research (@PalisadeAI) [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":20000,"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":[438,1201,1190],"class_list":["post-20001","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-analysis","tag-artificial-intelligence","tag-chatbots","tag-openai"],"aioseo_notices":[],"amp_enabled":true,"views":"32","promo_type":"","layout_type":"","short_excerpt":"","is_update":"","_links":{"self":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/20001","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/comments?post=20001"}],"version-history":[{"count":0,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/20001\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/media\/20000"}],"wp:attachment":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/media?parent=20001"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/categories?post=20001"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/tags?post=20001"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}