{"id":107586,"date":"2024-09-27T17:00:50","date_gmt":"2024-09-27T10:00:50","guid":{"rendered":"https:\/\/hotvideos24.online\/?p=107586"},"modified":"2024-09-27T17:00:50","modified_gmt":"2024-09-27T10:00:50","slug":"advanced-ai-chatbots-are-less-likely-to-admit-they-dont-have-all-the-answers","status":"publish","type":"post","link":"https:\/\/hotvideos24.online\/?p=107586","title":{"rendered":"Advanced AI chatbots are less likely to admit they don\u2019t have all the answers"},"content":{"rendered":"<p> <script async src=\"https:\/\/pagead2.googlesyndication.com\/pagead\/js\/adsbygoogle.js?client=ca-pub-3711241968723425\"\r\n     crossorigin=\"anonymous\"><\/script>\r\n<ins class=\"adsbygoogle\"\r\n     style=\"display:block\"\r\n     data-ad-format=\"fluid\"\r\n     data-ad-layout-key=\"-fb+5w+4e-db+86\"\r\n     data-ad-client=\"ca-pub-3711241968723425\"\r\n     data-ad-slot=\"7910942971\"><\/ins>\r\n<script>\r\n     (adsbygoogle = window.adsbygoogle || []).push({});\r\n<\/script><br \/>\n<\/p>\n<div>\n<p>Researchers have spotted an apparent downside of smarter chatbots. Although AI models predictably become more accurate as they advance, they\u2019re also more likely to (wrongly) answer questions beyond their capabilities rather than saying, \u201cI don\u2019t know.\u201d And the humans prompting them are more likely to take their confident hallucinations at face value, creating a trickle-down effect of confident misinformation.<\/p>\n<p>\u201cThey are answering almost everything these days,\u201d Jos\u00e9 Hern\u00e1ndez-Orallo, professor at the Universitat Politecnica de Valencia, Spain, <a data-i13n=\"elm:context_link;elmt:doNotAffiliate;cpos:1;pos:1\" class=\"link \" href=\"https:\/\/www.nature.com\/articles\/d41586-024-03137-3\" rel=\"nofollow noopener\" target=\"_blank\" data-ylk=\"slk:told;elm:context_link;elmt:doNotAffiliate;cpos:1;pos:1;itc:0;sec:content-canvas\">told<\/a> <em>Nature<\/em>. \u201cAnd that means more correct, but also more incorrect.\u201d Hern\u00e1ndez-Orallo, the project lead, worked on the study with his colleagues at the Valencian Research Institute for Artificial Intelligence in Spain.<\/p>\n<p>The team studied three LLM families, including OpenAI\u2019s GPT series, <a data-i13n=\"cpos:2;pos:1\" href=\"https:\/\/www.engadget.com\/social-media\/meta-ai-can-now-talk-to-you-and-edit-your-photos-172853219.html\" data-ylk=\"slk:Meta\u2019s LLaMA;cpos:2;pos:1;elm:context_link;itc:0;sec:content-canvas\" class=\"link \">Meta\u2019s LLaMA<\/a> and the open-source BLOOM. They tested early versions of each model and moved to larger, more advanced ones \u2014 but not today\u2019s most advanced. For example, the team began with OpenAI\u2019s relatively primitive GPT-3 ada model and tested iterations leading up to GPT-4, which <a data-i13n=\"cpos:3;pos:1\" href=\"https:\/\/www.engadget.com\/openai-just-released-gpt-4-a-multi-modal-generative-ai-172326765.html\" data-ylk=\"slk:arrived in March 2023;cpos:3;pos:1;elm:context_link;itc:0;sec:content-canvas\" class=\"link \">arrived in March 2023<\/a>. The <a data-i13n=\"cpos:4;pos:1\" href=\"https:\/\/www.engadget.com\/openai-claims-that-its-free-gpt-4o-model-can-talk-laugh-sing-and-see-like-a-human-184249780.html\" data-ylk=\"slk:four-month-old GPT-4o;cpos:4;pos:1;elm:context_link;itc:0;sec:content-canvas\" class=\"link \">four-month-old GPT-4o<\/a> wasn\u2019t included in the study, nor was <a data-i13n=\"cpos:5;pos:1\" href=\"https:\/\/www.engadget.com\/ai\/openais-new-o1-model-is-slower-on-purpose-185711459.html\" data-ylk=\"slk:the newer o1-preview;cpos:5;pos:1;elm:context_link;itc:0;sec:content-canvas\" class=\"link \">the newer o1-preview<\/a>. I\u2019d be curious if the trend still holds with the latest models.<\/p>\n<p>The researchers tested each model on thousands of questions about \u201carithmetic, anagrams, geography and science.\u201d They also quizzed the AI models on their ability to transform information, such as alphabetizing a list. The team ranked their prompts by perceived difficulty.<\/p>\n<p>The data showed that the chatbots\u2019 portion of wrong answers (instead of avoiding questions altogether) rose as the models grew. So, the AI is a bit like a professor who, as he masters more subjects, increasingly believes he has the golden answers on all of them.<\/p>\n<p>Further complicating things is the humans prompting the chatbots and reading their answers. The researchers tasked volunteers with rating the accuracy of the AI bots\u2019 answers, and they found that they \u201cincorrectly classified inaccurate answers as being accurate surprisingly often.\u201d The range of wrong answers falsely perceived as right by the volunteers typically fell between 10 and 40 percent.<\/p>\n<p>\u201cHumans are not able to supervise these models,\u201d concluded Hern\u00e1ndez-Orallo.<\/p>\n<p>The research team recommends AI developers begin boosting performance for easy questions and programming the chatbots to refuse to answer complex questions. \u201cWe need humans to understand: \u2018I can use it in this area, and I shouldn\u2019t use it in that area,\u2019\u201d Hern\u00e1ndez-Orallo told <em>Nature<\/em>.<\/p>\n<p>It\u2019s a well-intended suggestion that could make sense in an ideal world. But fat chance AI companies oblige. Chatbots that more often say \u201cI don\u2019t know\u201d would likely be perceived as less advanced or valuable, leading to less use \u2014 and less money for the companies making and selling them. So, instead, we get fine-print warnings that \u201cChatGPT can make mistakes\u201d and \u201cGemini may display inaccurate info.\u201d<\/p>\n<p>That leaves it up to us to avoid believing and spreading hallucinated misinformation that could hurt ourselves or others. For accuracy, fact-check your damn chatbot\u2019s answers, for crying out loud.<\/p>\n<p>You can read <a data-i13n=\"elm:context_link;elmt:doNotAffiliate;cpos:6;pos:1\" class=\"link \" href=\"https:\/\/www.nature.com\/articles\/s41586-024-07930-y\" rel=\"nofollow noopener\" target=\"_blank\" data-ylk=\"slk:the team\u2019s full study;elm:context_link;elmt:doNotAffiliate;cpos:6;pos:1;itc:0;sec:content-canvas\">the team\u2019s full study<\/a> in <em>Nature<\/em>.<\/p>\n<\/div>\n<p><script async src=\"https:\/\/pagead2.googlesyndication.com\/pagead\/js\/adsbygoogle.js?client=ca-pub-3711241968723425\"\r\n     crossorigin=\"anonymous\"><\/script>\r\n<ins class=\"adsbygoogle\"\r\n     style=\"display:block\"\r\n     data-ad-format=\"fluid\"\r\n     data-ad-layout-key=\"-fb+5w+4e-db+86\"\r\n     data-ad-client=\"ca-pub-3711241968723425\"\r\n     data-ad-slot=\"7910942971\"><\/ins>\r\n<script>\r\n     (adsbygoogle = window.adsbygoogle || []).push({});\r\n<\/script><br \/>\n<br \/><div data-type=\"_mgwidget\" data-widget-id=\"1660802\">\r\n<\/div>\r\n<script>(function(w,q){w[q]=w[q]||[];w[q].push([\"_mgc.load\"])})(window,\"_mgq\");\r\n<\/script>\r\n<br \/>\n<br \/><a href=\"https:\/\/www.engadget.com\/ai\/advanced-ai-chatbots-are-less-likely-to-admit-they-dont-have-all-the-answers-172012958.html\">Source link <\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Researchers have spotted an apparent downside of smarter chatbots. Although AI models predictably become more accurate as they advance, they\u2019re also more likely to (wrongly) answer questions beyond their capabilities &hellip; <a href=\"https:\/\/hotvideos24.online\/?p=107586\" class=\"more-link\">Read More<\/a><\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8628],"tags":[],"class_list":["post-107586","post","type-post","status-publish","format-standard","hentry","category-science","entry"],"_links":{"self":[{"href":"https:\/\/hotvideos24.online\/index.php?rest_route=\/wp\/v2\/posts\/107586","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/hotvideos24.online\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/hotvideos24.online\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/hotvideos24.online\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/hotvideos24.online\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=107586"}],"version-history":[{"count":0,"href":"https:\/\/hotvideos24.online\/index.php?rest_route=\/wp\/v2\/posts\/107586\/revisions"}],"wp:attachment":[{"href":"https:\/\/hotvideos24.online\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=107586"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/hotvideos24.online\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=107586"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/hotvideos24.online\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=107586"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}