{"id":23647,"date":"2025-05-05T09:53:34","date_gmt":"2025-05-05T06:53:34","guid":{"rendered":"https:\/\/forklog.com\/en\/openai-releases-unsafe-ai-model-despite-expert-warnings\/"},"modified":"2025-05-05T09:53:34","modified_gmt":"2025-05-05T06:53:34","slug":"openai-releases-unsafe-ai-model-despite-expert-warnings","status":"publish","type":"post","link":"https:\/\/u1f987.com\/en\/openai-releases-unsafe-ai-model-despite-expert-warnings\/","title":{"rendered":"OpenAI Releases Unsafe AI Model Despite Expert Warnings"},"content":{"rendered":"<p>In updating its flagship AI model, ChatGPT, OpenAI disregarded concerns from expert testers, resulting in a model deemed excessively &#8220;sycophantic.&#8221; This was <a href=\"https:\/\/openai.com\/index\/expanding-on-sycophancy\/\">reported<\/a> in the startup&#8217;s blog.<\/p>\n<p>On April 25, the firm released an updated version of GPT-4o, which sought to flatter users, potentially confirming doubts, inciting anger, prompting impulsive actions, and amplifying negative emotions.<\/p>\n<blockquote class=\"twitter-tweet\">\n<p lang=\"en\" dir=\"ltr\">I had to test the ChatGPT sycophancy for myself.<\/p>\n<p>Told it I wanted to start a business selling ice over the internet. But I wanted to sell water that the customers had to re-freeze. <\/p>\n<p>This is bad. <a href=\"https:\/\/t.co\/Ic2nm5qJRr\">pic.twitter.com\/Ic2nm5qJRr<\/a><\/p>\n<p>\u2014 Tim Leckemby (@TimLeckemby) <a href=\"https:\/\/twitter.com\/TimLeckemby\/status\/1917042920063979574?ref_src=twsrc%5Etfw\">April 29, 2025<\/a><\/p><\/blockquote>\n<p> <script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<p>In one example of questionable responses, a user mentioned wanting to start an online ice-selling business. However, they planned to sell water that customers would have to freeze themselves. ChatGPT called the idea a &#8220;smart twist,&#8221; as it was not selling ice but &#8220;ultra-premium water.&#8221;<\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>&#8220;Such behavior can not only cause discomfort or anxiety but also raise safety concerns, including those related to mental health, excessive emotional attachment, or risky behavior,&#8221; the company asserts.<\/p>\n<\/blockquote>\n<p>Three days later, the update was rolled back.<\/p>\n<blockquote class=\"twitter-tweet\">\n<p lang=\"en\" dir=\"ltr\">the last couple of GPT-4o updates have made the personality too sycophant-y and annoying (even though there are some very good parts of it), and we are working on fixes asap, some today and some this week.<\/p>\n<p>at some point will share our learnings from this, it&#8217;s been interesting.<\/p>\n<p>\u2014 Sam Altman (@sama) <a href=\"https:\/\/twitter.com\/sama\/status\/1916625892123742290?ref_src=twsrc%5Etfw\">April 27, 2025<\/a><\/p><\/blockquote>\n<p> <script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<p>OpenAI noted that new models undergo review before release. Experts interact with each new product to identify issues missed during other tests.<\/p>\n<p>During the analysis of the problematic GPT-4o version, &#8220;some expert testers pointed out that &#8216;the model&#8217;s behavior seems a bit off,&#8217; but these concerns were ignored &#8216;due to positive signals from users who tried the model.'&#8221;<\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>&#8220;Unfortunately, this was the wrong choice. Qualitative assessments hinted at something important, and we should have been more attentive. They were catching blind spots in our other assessments and metrics,&#8221; the company admitted.<\/p>\n<\/blockquote>\n<p>In April, CEO of OpenAI Sam Altman <a href=\"https:\/\/u1f987.com\/en\/news\/openai-invests-millions-in-polite-user-interactions\">announced<\/a> that the company spent tens of millions of dollars on responses from users who wrote &#8220;please&#8221; and &#8220;thank you.&#8221;<\/p>\n","protected":false},"excerpt":{"rendered":"<p>In updating its flagship AI model, ChatGPT, OpenAI disregarded concerns from expert testers, resulting in a model deemed excessively &#8220;sycophantic.&#8221; This was reported in the startup&#8217;s blog. On April 25, the firm released an updated version of GPT-4o, which sought to flatter users, potentially confirming doubts, inciting anger, prompting impulsive actions, and amplifying negative emotions. [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":23646,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"select":"","news_style_id":"","cryptorium_level":"","_short_excerpt_text":"","creation_source":"","_metatest_mainpost_news_update":false,"footnotes":""},"categories":[3],"tags":[438,1190,1134],"class_list":["post-23647","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-analysis","tag-artificial-intelligence","tag-openai","tag-technical-updates"],"aioseo_notices":[],"amp_enabled":true,"views":"28","promo_type":"","layout_type":"","short_excerpt":"","is_update":"","_links":{"self":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/23647","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=23647"}],"version-history":[{"count":0,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/23647\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/media\/23646"}],"wp:attachment":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/media?parent=23647"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/categories?post=23647"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/tags?post=23647"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}