{"id":54366,"date":"2026-06-30T11:27:24","date_gmt":"2026-06-30T03:27:24","guid":{"rendered":"https:\/\/www.1ai.net\/?p=54366"},"modified":"2026-06-30T11:27:24","modified_gmt":"2026-06-30T03:27:24","slug":"openai-codex-%e8%b4%9f%e8%b4%a3%e4%ba%ba%ef%bc%9aai-%e8%83%bd%e7%94%9f%e6%88%90%e8%ae%be%e8%ae%a1%ef%bc%8c%e4%bd%86%e5%ae%83%e8%bf%98%e4%b8%8d%e6%87%82%e5%93%81%e5%91%b3","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/54366.html","title":{"rendered":"OpenAI Codex Head: AI can generate design, but it doesn't know taste"},"content":{"rendered":"

On June 30th, according to The Business InsiderOpenAI<\/a> Codex<\/a> Andrew Ambrosino, director of the recent podcast \"Lenny's Podcast\", stated:AI<\/a> Creativedevise<\/a>The capacity in the field is still clearly short-walled - design is harder to quantify than the ability of a code to run, making it difficult to establish feedback loops for training models \" what is good design \" \u3002<\/p>\n

\"192731a8j00thfdkm003vd000u0m\"<\/p>\n

Design is more difficult to rate than software, and it is much more cumbersome to build a cycle that can train models to distinguish between good and bad designs than to judge whether the code can run\u3002<\/p>\n

Figma CEO Dylan Field, also previously noted in the podcast Hard Fork, that the AI model is based on \"data distribution\" training, and that output design results tend to fall into the \"average\" zone, with a lack of visible creativity. This quality determines the birth limits of AI in aesthetic judgement and design taste\u3002<\/p>\n

Ambrosino concludes with one sentence: \"Call on the human brain first. I'm sorry<\/p>","protected":false},"excerpt":{"rendered":"

On June 30, according to The Insider of Business, Andrew Ambrosino, head of OpenAI Codex, admittedly in his recent podcast, Lenny's Podcast, that AI's ability in the area of creative design remains significantly short \u2014 good or bad design is more difficult to quantify than the ability of the code to operate, making it difficult to establish a feedback loop for the training model \u201cWhat is a good design\u201d. Design is more difficult to rate than software, and it is much more cumbersome to build a cycle that can train models to distinguish between good and bad designs than to judge whether the code can run. Figma CEO Dylan Field, also noted in the podcast Hard Fork, that the AI model is based on \"data distribution\" training and output<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[411,6623,190,4272],"collection":[],"class_list":["post-54366","post","type-post","status-publish","format-standard","hentry","category-news","tag-ai","tag-codex","tag-openai","tag-4272"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/54366","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=54366"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/54366\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=54366"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=54366"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=54366"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=54366"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}