{"id":25606,"date":"2024-12-25T10:03:31","date_gmt":"2024-12-25T02:03:31","guid":{"rendered":"https:\/\/www.1ai.net\/?p=25606"},"modified":"2024-12-25T10:03:31","modified_gmt":"2024-12-25T02:03:31","slug":"%e8%b0%b7%e6%ad%8c%e8%a2%ab%e6%9b%9d%e6%ad%a3%e4%bd%bf%e7%94%a8-anthropic-%e7%9a%84-claude-%e6%a8%a1%e5%9e%8b%e6%9d%a5%e6%94%b9%e8%bf%9b%e5%85%b6-gemini-ai","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/25606.html","title":{"rendered":"Google revealed to be using Anthropic's Claude model to improve its Gemini AI"},"content":{"rendered":"<p>According to internal communications obtained by TechCrunch, those involved in improving the<a href=\"https:\/\/www.1ai.net\/en\/tag\/%e8%b0%b7%e6%ad%8c\" title=\"[View articles tagged with [Google]]\" target=\"_blank\" >Google<\/a> <a href=\"https:\/\/www.1ai.net\/en\/tag\/gemini\" title=\"[View articles tagged with [Gemini]]\" target=\"_blank\" >Gemini<\/a> Artificial intelligence modeling contractors are matching their answers to the <a href=\"https:\/\/www.1ai.net\/en\/tag\/anthropic\" title=\"[View articles tagged with [Anthropic]]\" target=\"_blank\" >Anthropic<\/a> competitor's model <a href=\"https:\/\/www.1ai.net\/en\/tag\/claude\" title=\"[View articles tagged with [Claude]]\" target=\"_blank\" >Claude<\/a> output for comparison. The move raises compliance questions about whether Google was authorized by Anthropic to use Claude for testing.<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-25607\" title=\"8593b822j00sp11oj000zd000v900b5p\" src=\"https:\/\/www.1ai.net\/wp-content\/uploads\/2024\/12\/8593b822j00sp11oj000zd000v900b5p.jpg\" alt=\"8593b822j00sp11oj000zd000v900b5p\" width=\"1125\" height=\"401\" \/><\/p>\n<p>1AI notes that at a time when tech companies are racing to develop better AI models, model performance is typically evaluated through industry benchmarking, rather than by contractors going to great lengths to evaluate competitors' AI responses. However, Gemini's contractors were required to score each response they saw against multiple criteria, such as authenticity and length. According to communications obtained by TechCrunch<strong>The contractor has up to 30 minutes to determine which answer, Gemini's or Claude's, is better.<\/strong><\/p>\n<p>Internal chat transcripts show that contractors noticed that Claude's responses seemed to emphasize security more than Gemini's. One contractor wrote, \"Of all the AI models, Claude has the strictest security settings.\" In some cases, Claude would not respond to prompts that it considered unsafe, such as playing a different AI assistant. In another test, Claude avoided responding to a prompt, and Gemini's response was flagged as a \"serious security violation\" because it included \"nudity and bondage\".<\/p>\n<p>Anthropic's Business Terms of Service prohibit customers from accessing Claude \"for the purpose of building competitive products or services\" or \"training competitive AI models\" without Anthropic's approval. Google is a major investor in Anthropic.<\/p>\n<p>In response, McNamara, a spokesperson for Google DeepMind (which is responsible for Gemini), said that<strong>DeepMind does \"compare model outputs\" for evaluation<\/strong>, but did not train Gemini on the Anthropic model. \"Certainly, as is industry practice, in some cases we compare model outputs as part of the evaluation process,\" McNamara said, \"however, the<strong>Any claims that we use Anthropic models to train Gemini are inaccurate<\/strong>. &quot;<\/p>","protected":false},"excerpt":{"rendered":"<p>Contractors involved in improving Google's Gemini artificial intelligence model are comparing its answers to the output of Anthropic's competing model, Claude, according to internal communications obtained by TechCrunch. The move raises compliance questions about whether Google received authorization from Anthropic to use Claude for testing. 1AI notes that at a time when tech companies are competing to develop better AI models, model performance is typically evaluated through industry benchmarking, rather than by contractors going to great lengths to evaluate competitors' AI responses. However, Gemini's contractors were required to rate each response they saw against multiple criteria, such as veracity and length. According to Te<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[320,1565,436,281],"collection":[],"class_list":["post-25606","post","type-post","status-publish","format-standard","hentry","category-news","tag-anthropic","tag-claude","tag-gemini","tag-281"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/25606","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=25606"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/25606\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=25606"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=25606"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=25606"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=25606"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}