{"id":11801,"date":"2024-05-31T09:15:03","date_gmt":"2024-05-31T01:15:03","guid":{"rendered":"https:\/\/www.1ai.net\/?p=11801"},"modified":"2024-05-31T09:15:03","modified_gmt":"2024-05-31T01:15:03","slug":"%e5%bf%ab%e6%89%8b%e4%b8%8a%e7%ba%bf%e8%87%aa%e7%a0%94%e6%96%87%e7%94%9f%e5%9b%be%e5%a4%a7%e6%a8%a1%e5%9e%8b%e5%8f%af%e5%9b%be%ef%bc%8c%e6%b6%88%e6%81%af%e7%a7%b0%e5%8f%82%e6%95%b0","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/11801.html","title":{"rendered":"Kuaishou launches self-developed large-scale text-based graph model "KeTu", with a parameter scale of billions"},"content":{"rendered":"

quick worker<\/a>Self-developedVincent Figure Large Model<\/a>\u201cKetu<\/a>"It has recently been officially opened to the public.It currently supports two types of functions: text-generated images and image-generated images, which can be used for AI image creation and AI image customization.<\/span>.<\/p>\n

Users can use it through the WeChat applet and web version of Ketu. According to Jiemian News, this is also the first time that Kuaishou has opened its self-developed series of large models to the public. The report quoted an insider as saying,The parameter scale of the "KeTu" large model reaches billions, and these data come from the open source community, Kuaishou's internal construction and self-developed AI technology synthesis<\/span>, covering tens of millions of common Chinese entity concepts, and also introducing reinforcement learning and reward model technology (RLHF), which solves the problem of the effectiveness of the large text-graph model under long text and complex semantic text input.<\/p>\n

According to reports, Kuaishou has clarified its big model application strategy this year, which mainly includes three directions: understanding, interaction and generation. Specific application scenarios include global big model content understanding, AI interaction, digital humans and AIGC in business scenarios, etc.<\/p>\n

\"831d945a-f920-47db-8c09-b327c58422e2\"<\/p>\n

Through actual testing, it was found that the functions provided by "Ketu" include "Wenshengtu", as well as a variety of anime themes and realistic portrait styles, including the "Red Diamond Noble" popular in the early 21st century, the more popular "Clay World" style recently, and a variety of different styles of "painting" styles. Users can generate up to 4 pictures at a time.<\/p>\n

According to previous reports, Cheng Yixiao, founder and CEO of Kuaishou, made a bold statement in March this year:We are confident that within the next six months, the comprehensive performance of the large model will reach the level of GPT4.0.<\/span>At the same time, he also said that the comprehensive performance of Kuaishou's Wenshengtu large model "Ketu" has exceeded the level of Midjourney V5.<\/p>","protected":false},"excerpt":{"rendered":"

Shutterbug's self-developed text-to-map model \"Ketu\" has recently been officially opened to the public, which currently supports two types of functions: text-to-map and map-to-map, and can be used for AI creation of images and AI image customization. Users can use it through the WeChat small program and web version of \"Ketu\". According to Interface News, this is also the first time that Shutterstock has opened its self-developed series of large models to the public. The report cited sources familiar with the matter as saying that the parameter scale of the \"KeTu\" big model reaches one billion, and the data comes from the open source community, the internal construction of Shutterstock and the synthesis of self-researched AI technology, covering the concept of common Chinese entities of ten million, and the introduction of reinforcement learning and rewarding modeling technology (RLHF), which has solved the problem of the big model of the text-born figure in the case of long and complex semantic text inputs, and has been used to solve the problem. and complex semantic text input under the effect of the problem. Reportedly, the fast hand internal this year clear large model application policy<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[2861,2859,2860],"collection":[],"class_list":["post-11801","post","type-post","status-publish","format-standard","hentry","category-news","tag-2861","tag-2859","tag-2860"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/11801","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=11801"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/11801\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=11801"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=11801"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=11801"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=11801"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}