360智脑大模型应用发布
“大模型就像‘电厂’一样，通过‘燃烧’大数据直接产生‘电力’，可以把通用人工智能的能力输送给每一个行业、每一个企业。”360公司创始人周鸿祎在演讲时说，“我认为大模型产生了未来5到10年新工业革命的机会”。
日前，360公司（以下简称“360”)召开360智脑大模型应用发布会，认知型通用大模型“360智脑4.0”亮相，360AI数字人正式发布。
现场，周鸿祎创建了自己的数字分身作为“360发言人”。他介绍，数字分身拥有他本人的私人知识库，能够复刻他的思维方式和人生经历，是有“灵魂”的数字人。
“360在大模型上的核心创新在AI数字人，数字人是未来人工智能大模型的最重要的应用入口。”周鸿祎表示，360将依托360智脑大模型能力，重新定义“数字人”。区别于传统数字人只能按既定脚本输出内容，360AI数字人的特点在于有记忆、有人设和性格，能够复刻思维方式和人生经历。
据介绍，360AI数字人目前拥有200多个角色，分为数字名人和数字员工两类。数字名人包括历史人物、偶像明星、文学形象等，让用户在与数字人的开放对话中，实现与偶像的“近距离接触”，与先贤的“思想交流”。数字员工则可成为企业员工的知识助手和办公助手，提升企业运营效能。大会现场演示了“法务专员”数字人角色，数字人对中小企业常见的公章管理和合同审核问题给予专业回答。
“AI的进化要以人为本，大模型应该成为人类的朋友和助手。”周鸿祎说。
360AI数字人支持定制，每个人都能通过上传私人知识库，低成本生成自己的专属数字人，如数字分身、数字助理、数字偶像等。周鸿祎表示，360AI数字人未来的发展方向是生成声音、视频并拥有长期记忆，甚至具有目标分解和规划的能力，并拥有“手和脚”，具备执行力，更好为人类提供服务。
360AI数字人背后，是“360智脑4.0”强大的多模态能力的支撑。记者了解到，GPT时代中国自研大模型弯道超车的关键在于发展多模态。目前，360已具备跨模态生成能力，包括文字、图像、语音以及视频处理能力，可实现文生文、文生图等功能。会上，360智脑“文生视频”多模态功能国内首发，文字脚本可生成视频，不受专业技能和素材限制。
值得一提的是，6月12日，360智脑的360GPT-S2-V8型号产品获得中国信息通信研究院“可信AIGC大语言模型基础能力”评估报告，报告显示必选项目评估内容均通过。据悉，360智脑是国内首个通过中国信通院该项权威评估的大模型产品。
“中国进入百模大战，基础能力外，更重要的是应用场景的比拼。”周鸿祎在会上宣布，360智脑在多模态等关键能力上完成迭代，全面接入“360全家桶”——浏览器、搜索、安全卫士等360全端产品，实现全面智能化重塑。
距今年3月底不到3个月时间内，360智脑大模型落地搜索场景，完成从1.0版本到4.0版本的迭代。大会预告，下一步360计划发布插件平台，将进一步推动大模型产业生态发展。
翻译：
Zhou Hongyi: Big models will lead the new industrial revolution
360 brain big model application released
“Big models are like ‘power plants’ that directly generate’ electricity ‘by’ burning ‘big data, which can deliver the capabilities of general AI to every industry and every enterprise.” Zhou Hongyi, founder of Qihoo 360, said in his speech, “I think big models create opportunities for a new industrial revolution in the next five to 10 years.”
Recently, 360 Company (hereinafter referred to as “360”) held a 360 intelligent brain large model application conference, cognitive general large model “360 Intelligent Brain 4.0” debut, 360AI digital people officially released.
At the scene, Zhou Hongyi created his own digital doppelganger as a “360 spokesperson.” He introduced that the digital dopplegger has his own private knowledge base, can reproduce his way of thinking and life experience, and is a digital person with a “soul”.
“360’s core innovation in the large model is AI digital man, and digital man is the most important application entrance of the future AI large model.” Zhou Hongyi said that 360 will rely on the ability of 360 brain large model to redefine “digital people”. Unlike traditional digital people who can only output content according to a set script, 360AI digital people are characterized by memory, personality and personality, and can reproduce ways of thinking and life experiences.
According to reports, 360AI digital people currently have more than 200 roles, divided into two categories of digital celebrities and digital employees. Digital celebrities include historical figures, idol stars, literary images, etc., so that users can achieve “close contact” with idols and “exchange of ideas” with sages in the open dialogue with digital people. Digital employees can become knowledge assistants and office assistants for enterprise employees, improving the efficiency of enterprise operations. The conference demonstrated the digital role of “legal specialist”, and digital people gave professional answers to common seal management and contract review questions of small and medium-sized enterprises.
“The evolution of AI should be human-centered, and large models should be humans’ friends and assistants.” Mr. Zhou said.
360AI digital people support customization, everyone can upload a private knowledge base, low-cost generation of their own exclusive digital people, such as digital doppleganger, digital assistant, digital idol and so on. Zhou Hongyi said that the future development direction of 360AI digital people is to generate sound, video and have long-term memory, and even have the ability to decompose and plan goals, and have “hands and feet”, with executive power, and better provide services for human beings.
Behind 360AI digital people, is the support of the powerful multi-modal capability of “360 Intelligent Brain 4.0”. The reporter learned that the key to China’s self-developed large model curve overtaking in the GPT era is the development of multi-mode. At present, 360 has cross-modal generation capabilities, including text, image, voice and video processing capabilities, can achieve Wensheng, Wensheng diagram and other functions. At the meeting, 360 intelligent brain “Vinson video” multi-modal function was first launched in China, text script can generate video, not limited by professional skills and materials.
It is worth mentioning that on June 12, 360 Wisdom’s 360GPT-S2-V8 model product obtained the “Trusted AIGC large language model basic Ability” evaluation report of the China Academy of Information and Communications Technology, and the report showed that the evaluation content of mandatory projects were passed. It is reported that 360 Wisdom brain is the first large model product that has passed the authoritative evaluation of the China Academy of Information and Communication Technology.
“China has entered the 100 model war, in addition to basic capabilities, more important is the competition of application scenarios.” Zhou Hongyi announced at the meeting that 360 Intelligent brain completed iteration on key capabilities such as multi-mode, fully accessing the “360 family bucket” – browser, search, security guards and other 360 full-end products to achieve comprehensive intelligent reshaping.
Less than 3 months from the end of March this year, 360 brain large model landing search scene, completed the iteration from version 1.0 to version 4.0. In the next step, 360 plans to release a plug-in platform, which will further promote the ecological development of the large model industry.
