{"id":224321,"date":"2025-10-19T04:22:09","date_gmt":"2025-10-19T04:22:09","guid":{"rendered":"https:\/\/www.newsbeep.com\/au\/224321\/"},"modified":"2025-10-19T04:22:09","modified_gmt":"2025-10-19T04:22:09","slug":"game-over-agi-is-not-imminent-and-llms-are-not-the-royal-road-to-getting-there","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/au\/224321\/","title":{"rendered":"Game over. AGI is not imminent, and LLMs are not the royal road to getting there."},"content":{"rendered":"<p>First slowly, and then all at once, dreams of LLMs bringing us to the cusp of AGI have fallen apart. The last few months have been devastating:<\/p>\n<p>\u2022 June, 2025: the Apple reasoning paper <a href=\"https:\/\/open.substack.com\/pub\/garymarcus\/p\/a-knockout-blow-for-llms?r=8tdk6&amp;utm_campaign=post&amp;utm_medium=web&amp;showWelcomeOnShare=false\" rel=\"nofollow noopener\" target=\"_blank\">confirmed that even with \u201creasoning\u201d, LLMs still can\u2019t solve distribution shift<\/a>, the core Achille\u2019s heel in neural networks that I have been writing about for nearly 30 years. Several other papers, include the ASU \u201c<a href=\"https:\/\/arxiv.org\/pdf\/2508.01191\" rel=\"nofollow noopener\" target=\"_blank\">Mirage<\/a>\u201c paper quickly extended this finding.<\/p>\n<p><a target=\"_blank\" href=\"https:\/\/substackcdn.com\/image\/fetch\/$s_!MjBm!,f_auto,q_auto:good,fl_progressive:steep\/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F308ecb15-04f7-43c2-b928-64c3e7c08710_1225x1830.jpeg\" data-component-name=\"Image2ToDOM\" rel=\"nofollow noopener\" class=\"image-link image2 is-viewable-img\"><img decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/au\/wp-content\/uploads\/2025\/10\/https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/308ecb15-04f7-43c2-b928-64c3e7c08710_1225.jpeg\" width=\"1225\" height=\"1830\" data-attrs=\"{&quot;src&quot;:&quot;https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/308ecb15-04f7-43c2-b928-64c3e7c08710_1225x1830.jpeg&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:1830,&quot;width&quot;:1225,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:412021,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image\/jpeg&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https:\/\/garymarcus.substack.com\/i\/176484932?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F308ecb15-04f7-43c2-b928-64c3e7c08710_1225x1830.jpeg&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}\" alt=\"\"   fetchpriority=\"high\" class=\"sizing-normal\"\/><\/a><\/p>\n<p>\u2022 August, 2025: GPT-5 came late and <a href=\"https:\/\/open.substack.com\/pub\/garymarcus\/p\/gpt-5-overdue-overhyped-and-underwhelming?r=8tdk6&amp;utm_campaign=post&amp;utm_medium=web&amp;showWelcomeOnShare=false\" rel=\"nofollow noopener\" target=\"_blank\">fell short<\/a>. <\/p>\n<p>\u2022 September, 2025: Turing Award winner Rich Sutton, known best for RL and his \u201cbitter lesson\u201d thanked me for my critiques of LLMs and agreed. Seismic. (And astonishingly gracious!)<\/p>\n<p><a target=\"_blank\" href=\"https:\/\/substackcdn.com\/image\/fetch\/$s_!02t6!,f_auto,q_auto:good,fl_progressive:steep\/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8ae30f59-930f-4d23-9848-fa771611ac69_1319x360.png\" data-component-name=\"Image2ToDOM\" rel=\"nofollow noopener\" class=\"image-link image2 is-viewable-img\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/au\/wp-content\/uploads\/2025\/10\/https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/8ae30f59-930f-4d23-9848-fa771611ac69_1319.jpeg\" width=\"1319\" height=\"360\" data-attrs=\"{&quot;src&quot;:&quot;https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/8ae30f59-930f-4d23-9848-fa771611ac69_1319x360.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:360,&quot;width&quot;:1319,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:103296,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image\/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:&quot;https:\/\/garymarcus.substack.com\/i\/176484932?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8ae30f59-930f-4d23-9848-fa771611ac69_1319x360.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}\" alt=\"\"   class=\"sizing-normal\"\/><\/a><\/p>\n<p>\u2022 October, 2025: Andrej Karpathy, a widely-respected machine learning expert who once ran AI for Tesla and twice worked for OpenAI, just said agents aren\u2019t anywhere close, and that AGI is a decade away.<\/p>\n<p><a target=\"_blank\" href=\"https:\/\/substackcdn.com\/image\/fetch\/$s_!y-bV!,f_auto,q_auto:good,fl_progressive:steep\/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa2195af5-f47d-4cca-a811-f96e792c4b3a_1376x1389.png\" data-component-name=\"Image2ToDOM\" rel=\"nofollow noopener\" class=\"image-link image2 is-viewable-img\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/au\/wp-content\/uploads\/2025\/10\/https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/a2195af5-f47d-4cca-a811-f96e792c4b3a_1376.jpeg\" width=\"1376\" height=\"1389\" data-attrs=\"{&quot;src&quot;:&quot;https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/a2195af5-f47d-4cca-a811-f96e792c4b3a_1376x1389.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:1389,&quot;width&quot;:1376,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1376465,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image\/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:&quot;https:\/\/garymarcus.substack.com\/i\/176484932?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa2195af5-f47d-4cca-a811-f96e792c4b3a_1376x1389.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}\" alt=\"\"   class=\"sizing-normal\"\/><\/a><a target=\"_blank\" href=\"https:\/\/substackcdn.com\/image\/fetch\/$s_!KWcq!,f_auto,q_auto:good,fl_progressive:steep\/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0566e0d2-1c7a-47b5-842f-a05dc69bc2f8_1413x1399.png\" data-component-name=\"Image2ToDOM\" rel=\"nofollow noopener\" class=\"image-link image2 is-viewable-img\"><img decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/au\/wp-content\/uploads\/2025\/10\/https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/0566e0d2-1c7a-47b5-842f-a05dc69bc2f8_1413.jpeg\" width=\"1413\" height=\"1399\" data-attrs=\"{&quot;src&quot;:&quot;https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/0566e0d2-1c7a-47b5-842f-a05dc69bc2f8_1413x1399.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:1399,&quot;width&quot;:1413,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1038228,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image\/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:&quot;https:\/\/garymarcus.substack.com\/i\/176484932?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F4dc67cc2-f51d-43c4-bf63-b8660ffb6d06_1430x1430.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}\" alt=\"\"   loading=\"lazy\" class=\"sizing-normal\"\/><\/a><\/p>\n<p>\u2022 October, 2025: Nobel Laureate and Google DeepMind Sir Demis Hassabis just blew up some wildly overhyped claims from OpenAI\u2019s Sebastien Bubeck about math.  Researcher\/Hyperbolic Labs CTO Yuchen Jin explains:<\/p>\n<p><a target=\"_blank\" href=\"https:\/\/substackcdn.com\/image\/fetch\/$s_!ZO-4!,f_auto,q_auto:good,fl_progressive:steep\/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F62f04542-2585-4417-bd7e-84d2efb8d25b_1388x2069.png\" data-component-name=\"Image2ToDOM\" rel=\"nofollow noopener\" class=\"image-link image2 is-viewable-img\"><img decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/au\/wp-content\/uploads\/2025\/10\/https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/62f04542-2585-4417-bd7e-84d2efb8d25b_1388.jpeg\" width=\"1388\" height=\"2069\" data-attrs=\"{&quot;src&quot;:&quot;https:\/\/substack-post-media.s3.amazonaws.com\/public\/images\/62f04542-2585-4417-bd7e-84d2efb8d25b_1388x2069.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:2069,&quot;width&quot;:1388,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:902607,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image\/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:true,&quot;topImage&quot;:false,&quot;internalRedirect&quot;:&quot;https:\/\/garymarcus.substack.com\/i\/176484932?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F62f04542-2585-4417-bd7e-84d2efb8d25b_1388x2069.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}\" alt=\"\"   loading=\"lazy\" class=\"sizing-normal\"\/><\/a><\/p>\n<p>Embarrassing indeed.<\/p>\n<p>LLMs have their place, but anyone expecting the current paradigm to be close to AGI is delusional.<\/p>\n<p>Gary Marcus tried to tell you for years that this moment would come for years. Do consider reading his 2020 article <a href=\"https:\/\/arxiv.org\/abs\/2002.06177\" rel=\"nofollow noopener\" target=\"_blank\">Next Decade in AI<\/a>, for an alternative strategy that increasingly looks  prescient. <\/p>\n<p>And please consider joining almost 85,000 others in subscribing to the newsletter that told you so much that has proven true, long before it was fashionable.<\/p>\n","protected":false},"excerpt":{"rendered":"First slowly, and then all at once, dreams of LLMs bringing us to the cusp of AGI have&hellip;\n","protected":false},"author":2,"featured_media":224322,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[256,254,255,64,63,105],"class_list":{"0":"post-224321","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-artificial-intelligence","8":"tag-ai","9":"tag-artificial-intelligence","10":"tag-artificialintelligence","11":"tag-au","12":"tag-australia","13":"tag-technology"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts\/224321","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/comments?post=224321"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts\/224321\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/media\/224322"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/media?parent=224321"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/categories?post=224321"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/tags?post=224321"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}