{"id":46820,"date":"2021-07-28T12:03:03","date_gmt":"2021-07-28T09:03:03","guid":{"rendered":"https:\/\/forklog.com\/en\/?p=46820"},"modified":"2025-09-01T18:21:33","modified_gmt":"2025-09-01T15:21:33","slug":"deepmind-builds-universal-ai-agents-in-an-open-ended-game","status":"publish","type":"post","link":"https:\/\/u1f987.com\/en\/deepmind-builds-universal-ai-agents-in-an-open-ended-game\/","title":{"rendered":"DeepMind builds universal AI agents in an open-ended game"},"content":{"rendered":"<p>Researchers from Britain\u2019s DeepMind have created an extensive gaming environment, XLand, to train universally capable artificial intelligence agents. They can perform tasks they have never encountered.<\/p>\n<p><!--more--><\/p>\n<blockquote class=\"twitter-tweet\">\n<p lang=\"en\" dir=\"ltr\">Reinforcement learning typically trains &#038; tests agents on the same game. New work shows how our team trains generally capable agents on huge game spaces, resulting in agents that generalise to held-out test games, &#038; learn behaviours like experimentation <a href=\"https:\/\/t.co\/fQ9UrOFkMb\">https:\/\/t.co\/fQ9UrOFkMb<\/a> 1\/ <a href=\"https:\/\/t.co\/UAPX5VgMmO\">pic.twitter.com\/UAPX5VgMmO<\/a><\/p>\n<p>\u2014 DeepMind (@DeepMind) <a href=\"https:\/\/twitter.com\/DeepMind\/status\/1420024899549253637?ref_src=twsrc%5Etfw\">July 27, 2021<\/a><\/p><\/blockquote>\n<p> <script async=\"\" src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<p>Rather than training agents on a fixed set of tasks, the team defined an entire universe of situations that can be procedurally generated.<\/p>\n<figure class=\"wp-block-image\"><img decoding=\"async\" src=\"https:\/\/lh3.googleusercontent.com\/c4InR6pCe0CBwakueBWi8FE3XOJuw6znEVQO7j2Ww9c2M5bS9Is-NvHY7QSCgdvfqGidxSHz-D2TVYvySt70CZhVujM1Cjw9Lodgf9ok42FD4UX3z1ugrAlOhZ_PQ1kg3QLnvnMu\" alt=\"DeepMind \u0441\u043e\u0437\u0434\u0430\u043b\u0430 \u0443\u043d\u0438\u0432\u0435\u0440\u0441\u0430\u043b\u044c\u043d\u044b\u0445 \u0418\u0418-\u0430\u0433\u0435\u043d\u0442\u043e\u0432 \u0432 \u043d\u0435\u043e\u0433\u0440\u0430\u043d\u0438\u0447\u0435\u043d\u043d\u043e\u0439 \u0438\u0433\u0440\u0435\"\/><figcaption>The XLand universe setup. Data: DeepMind.<\/figcaption><\/figure>\n<p>The XLand universe comprises a suite of multiplayer three-dimensional games. Within its bounds, agents conducted 700,000 unique sessions across 4,000 worlds. Each agent in the latest generation completed 200 billion learning steps as a result of performing 3.4 million unique tasks.<\/p>\n<p>The researchers say the resulting algorithm can successfully solve a broad spectrum of tasks\u2014from simple object search to complex games such as hide-and-seek or capture the flag\u2014that were not encountered during training, the scientists said.<\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201cCurrently our agents can participate in all procedurally generated evaluation tasks, with a few situations that not even humans could solve,\u201d the study says.<\/p>\n<\/blockquote>\n<p>According to the scientists, the agents exhibit general heuristic behaviour that is widely applicable to many tasks.<\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201cThis new approach marks an important step toward creating more universal agents capable of adapting quickly to a constantly changing environment,\u201d they added.<\/p>\n<\/blockquote>\n<p>As noted in June, DeepMind researchers said that <a href=\"https:\/\/u1f987.com\/en\/news\/deepmind-reinforcement-learning-suffices-to-create-general-ai\"> achieving artificial general intelligence requires only reinforcement learning <\/a>.<\/p>\n<p>In July, AI lab specialists assembled and <a href=\"https:\/\/u1f987.com\/en\/news\/deepmind-ai-maps-20000-human-protein-structures\"> published the most comprehensive database of human protein structures <\/a>, created by the AlphaFold neural network.<\/p>\n<p>Follow ForkLog&#8217;s AI news on Telegram: <a href=\"https:\/\/t.me\/forklogAI\" target=\"_blank\" rel=\"noreferrer noopener nofollow\"> ForkLog AI<\/a> \u2014 all the news from the AI world!<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Researchers from Britain&#8217;s DeepMind have created an extensive gaming environment, XLand, to train universally capable AI agents. They can tackle tasks they have never encountered.<\/p>\n","protected":false},"author":1,"featured_media":46821,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"select":"1","news_style_id":"1","cryptorium_level":"","_short_excerpt_text":"","creation_source":"","_metatest_mainpost_news_update":false,"footnotes":""},"categories":[3],"tags":[438,1474,167],"class_list":["post-46820","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-analysis","tag-artificial-intelligence","tag-deepmind","tag-research"],"aioseo_notices":[],"amp_enabled":true,"views":"29","promo_type":"1","layout_type":"1","short_excerpt":"","is_update":"","_links":{"self":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/46820","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=46820"}],"version-history":[{"count":1,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/46820\/revisions"}],"predecessor-version":[{"id":46822,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/46820\/revisions\/46822"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/media\/46821"}],"wp:attachment":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/media?parent=46820"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/categories?post=46820"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/tags?post=46820"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}