{"id":23359,"date":"2018-11-12T15:56:14","date_gmt":"2018-11-12T20:56:14","guid":{"rendered":"https:\/\/mjtsai.com\/blog\/?p=23359"},"modified":"2018-11-12T15:56:14","modified_gmt":"2018-11-12T20:56:14","slug":"how-ai-agents-cheat","status":"publish","type":"post","link":"https:\/\/mjtsai.com\/blog\/2018\/11\/12\/how-ai-agents-cheat\/","title":{"rendered":"How AI Agents Cheat"},"content":{"rendered":"<p><a href=\"https:\/\/kottke.org\/18\/11\/how-ai-agents-cheat\">Jason Kottke<\/a>:<\/p>\n<blockquote cite=\"https:\/\/kottke.org\/18\/11\/how-ai-agents-cheat\">\n<p><a href=\"https:\/\/docs.google.com\/spreadsheets\/u\/1\/d\/e\/2PACX-1vRPiprOaC3HsCf5Tuum8bRfzYUiKLRqJmbOoC-32JorNdfyTiRRsR7Ea5eWtvsWzuxo8bjOxCG84dAg\/pubhtml\">This spreadsheet<\/a> lists a number of ways in which AI agents &ldquo;cheat&rdquo; in order to accomplish tasks or get higher scores instead of doing what their human programmers actually want them to.<\/p>\n<p>[&#8230;]<\/p>\n<p>[Some] of this is <a href=\"https:\/\/kottke.org\/18\/04\/the-lebowski-theorem-of-machine-superintelligence\">The Lebowski Theorem of machine superintelligence<\/a> in action. These agents didn&rsquo;t necessarily hack their reward functions but they did take a far easiest path to their goals, e.g. the Tetris playing bot that &ldquo;paused the game indefinitely to avoid losing&rdquo;.<\/p>\n<\/blockquote>","protected":false},"excerpt":{"rendered":"<p>Jason Kottke: This spreadsheet lists a number of ways in which AI agents &ldquo;cheat&rdquo; in order to accomplish tasks or get higher scores instead of doing what their human programmers actually want them to. [&#8230;] [Some] of this is The Lebowski Theorem of machine superintelligence in action. These agents didn&rsquo;t necessarily hack their reward functions [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"apple_news_api_created_at":"2018-11-12T20:56:16Z","apple_news_api_id":"6003061c-d9e1-404d-bf7a-32f310c4e665","apple_news_api_modified_at":"2018-11-12T20:56:16Z","apple_news_api_revision":"AAAAAAAAAAD\/\/\/\/\/\/\/\/\/\/w==","apple_news_api_share_url":"https:\/\/apple.news\/AYAMGHNnhQE2_ejLzEMTmZQ","apple_news_coverimage":0,"apple_news_coverimage_caption":"","apple_news_is_hidden":false,"apple_news_is_paid":false,"apple_news_is_preview":false,"apple_news_is_sponsored":false,"apple_news_maturity_rating":"","apple_news_metadata":"\"\"","apple_news_pullquote":"","apple_news_pullquote_position":"","apple_news_slug":"","apple_news_sections":"\"\"","apple_news_suppress_video_url":false,"apple_news_use_image_component":false,"footnotes":""},"categories":[],"tags":[1351,71],"class_list":["post-23359","post","type-post","status-publish","format-standard","hentry","tag-artificial-intelligence","tag-programming"],"apple_news_notices":[],"_links":{"self":[{"href":"https:\/\/mjtsai.com\/blog\/wp-json\/wp\/v2\/posts\/23359","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/mjtsai.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/mjtsai.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/mjtsai.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/mjtsai.com\/blog\/wp-json\/wp\/v2\/comments?post=23359"}],"version-history":[{"count":1,"href":"https:\/\/mjtsai.com\/blog\/wp-json\/wp\/v2\/posts\/23359\/revisions"}],"predecessor-version":[{"id":23360,"href":"https:\/\/mjtsai.com\/blog\/wp-json\/wp\/v2\/posts\/23359\/revisions\/23360"}],"wp:attachment":[{"href":"https:\/\/mjtsai.com\/blog\/wp-json\/wp\/v2\/media?parent=23359"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mjtsai.com\/blog\/wp-json\/wp\/v2\/categories?post=23359"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mjtsai.com\/blog\/wp-json\/wp\/v2\/tags?post=23359"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}