{"id":44991,"date":"2025-10-21T12:25:10","date_gmt":"2025-10-21T04:25:10","guid":{"rendered":"https:\/\/www.1ai.net\/?p=44991"},"modified":"2025-10-21T12:25:10","modified_gmt":"2025-10-21T04:25:10","slug":"%e5%89%8d-openai-%e7%a0%94%e7%a9%b6%e5%91%98%ef%bc%9aai-agent-%e7%9a%84%e7%9c%9f%e6%ad%a3%e8%90%bd%e5%9c%b0%e4%bb%8d%e9%9c%80%e5%8d%81%e5%b9%b4","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/44991.html","title":{"rendered":"Former OpenAI Fellow: The real landing of AI Agent will take ten years"},"content":{"rendered":"<p>October 21, according to The Decoder, before <a href=\"https:\/\/www.1ai.net\/en\/tag\/openai\" title=\"[View articles tagged with [OpenAI]]\" target=\"_blank\" >OpenAI<\/a> Andrej, an artificial intelligence researcher with Tesla <a href=\"https:\/\/www.1ai.net\/en\/tag\/karpathy\" title=\"[See articles with [Karpathy] labels]\" target=\"_blank\" >Karpathy<\/a> In a recent interview, it was said that the current focus was \"<a href=\"https:\/\/www.1ai.net\/en\/tag\/agent-ai\" title=\"_Other Organiser\" target=\"_blank\" >Agent AI<\/a>\"The heat is too exaggerated and there is a clear gap between reality and expectations\u3002<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-44992\" title=\"e7639b12j00t4gs8c009nd000uhwm\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2025\/10\/e7639b12j00t4gs8c009nd000u000hwm.jpg\" alt=\"e7639b12j00t4gs8c009nd000uhwm\" width=\"1080\" height=\"644\" \/><\/p>\n<p><strong>He said, \"These models can't work like human interns or professional staff,\" and predicted that it might take about 10 more years to get real Agent AI\u3002<\/strong><\/p>\n<p>Karpathy noted that the current large model was seriously deficient in terms of core awareness, multi-modular processing, long-term memory and complex mandate implementation\u3002<\/p>\n<p>Even advanced models such as GPT-5 Pro can, in his view, only function in narrow scenarios such as the \"Auxiliary Tool for Code Analysis\" and its true value remains at the \"Auto Completion\" level\u3002<\/p>\n<p>At the data level, Karpathy criticized the poor quality of the available training language, with a large number of Internet content being \u201cdebris, symbols and noise\u201d, leading to a model relying more on memory than understanding\u3002<\/p>\n<p>He stressed that future breakthroughs should come from \u201cquality, filtered data\u201d and called on industry to value professional content creators\u3002<\/p>\n<p>For the way forward, Karpathy does not believe that a single technological singularity will emerge, but rather relies on \"better data, stronger structures, more efficient learning methods and faster hardware\" to move forward\u3002<\/p>\n<p>At the same time, he stressed that, in the long term, language models needed to be learned through interaction with the environment and not just copying texts\u3002<\/p>","protected":false},"excerpt":{"rendered":"<p>According to news from October 21, according to The Decoder, in a recent interview, Andr\u00e9 Karpathy, an artificial intelligence researcher from former OpenAI and Tesla, stated that the current trend around \"Agent AI\" was exaggerated and that there was a clear discrepancy between reality and expectations. He said, \"These models can't work like human interns or professional staff,\" and predicted that it might take about 10 more years to get real Agent AI. Karpathy noted that the current large model was seriously deficient in terms of core awareness, multi-modular processing, long-term memory and complex mandate implementation. He thinks even GPT-5 Pro like this<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[7760,7761,190],"collection":[],"class_list":["post-44991","post","type-post","status-publish","format-standard","hentry","category-news","tag-agent-ai","tag-karpathy","tag-openai"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/44991","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=44991"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/44991\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=44991"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=44991"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=44991"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=44991"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}