Workflow
混元大模型
icon
Search documents
专访汤道生:元宝重兵投入这半年
Sou Hu Cai Jing· 2025-10-10 10:42
以下文章来源于语言即世界language is world,作者张小珺 汤道生 腾讯集团高级执行副总裁、云与智慧产业事业群CEO 张小珺 以下是张小珺对汤道生的访谈全文(作者做了语言优化。) 第一章 调组织、接DeepSeek、资源整合,腾讯元宝这半年 01举手 张小珺: Dowson,过去一年从我们上次访谈到现在,在你眼中,整个AI市场有哪些是最重要的变量?腾讯在AI的动作上有哪些是最重要的变化? 汤道生: 最大变化是国内大模型市场更集中了,开源成为国内各大模型的重要战略,比如DeepSeek横空出世。 对腾讯非常重要的变化是,很多产品服务从原来只基于混元,到变成开放整合多家大模型。 去年我们看到AI产品赛道起来了。年底我举手,跟TEG(腾讯技术工程事业群)再度搭配起来——他们更专注做混元大模型,CSIG(云与智慧产业事业 群)承担前端产品责任,一起打仗。 这比较像我们做云服务,TEG和CSIG紧密合作,也是类似模式——我们在前端服务客户,他们在后端提供底层技术支持。 CSIG接了元宝后不久,马上遇到DeepSeek R1发布,当时大家对于DeepSeek App有强烈需求。但DeepSeek自己的App限 ...
专访汤道生:元宝重兵投入这半年
腾讯研究院· 2025-10-10 08:33
Core Viewpoint - The article discusses Tencent's strategic moves in the AI market, particularly focusing on the integration of its AI product "Yuanbao" with DeepSeek, highlighting the importance of user demand and the evolving landscape of AI applications in both consumer and enterprise sectors [4][6]. Group 1: AI Market Changes - The domestic large model market has become more concentrated, with open-source strategies becoming crucial for major models like DeepSeek [7]. - Tencent's AI products have shifted from being solely based on its own models to integrating multiple large models, indicating a more collaborative approach [8]. Group 2: Strategic Decisions - The decision to integrate Yuanbao with DeepSeek was driven by a strong user demand and the recognition of a new market opportunity [9][10]. - The leadership at Tencent, including Pony Ma and Martin Lau, supported the idea of placing Yuanbao under a product-focused team to enhance its market presence [10][11]. Group 3: Product Development and Integration - Yuanbao's integration into various Tencent platforms, including WeChat, has been unprecedented, showcasing Tencent's commitment to the AI sector [35][36]. - The company is actively exploring different product scenarios to enhance Yuanbao's functionality and user engagement [36][40]. Group 4: User Experience and Interaction - The interaction style of Yuanbao varies across platforms, with a more casual tone in WeChat compared to a more formal approach in its standalone app [67][73]. - The team is experimenting with different interaction styles to cater to user preferences, aiming for a more personalized experience [82][84]. Group 5: Future Outlook and Market Position - The competition in the AI chatbot market is expected to remain fragmented, with users having diverse preferences for different products [91][92]. - Tencent views its AI initiatives as a critical battle akin to the mobile internet era, emphasizing the importance of establishing a strong user base in the AI landscape [122][125].
盘点AI黄金周:Sora 2引爆AI视频、蚂蚁冲进万亿参数俱乐部
Sou Hu Cai Jing· 2025-10-10 02:08
文|光锥智能 白鸽 而在大洋彼岸,Claude Sonnet 4.5同期发布。谷歌Gemini 3.0也预计将在10月迎战。 显然,全球大模型进入了一个新的井喷期,各家在大模型产品和技术迭代上,都在明显加速。据不完全 统计,仅9月全球就有15家主流企业/机构发布了开源大模型,数量较8月激增70%。 开源大模型的热潮,也是开发者的盛宴。有开发者都在社交媒体上感慨:一天测了3个模型,还有一个 没测试,这还只是周一...... 这个十一黄金周,AI圈真的很热闹。 10月1日凌晨,OpenAI深夜炸场,正式官宣新一代AI视频模型Sora 2,并直接上线了一个AI版的"抖 音"——一个完全由AI来生成视频的短视频平台。 就在Sora 2发布之前,这一轮的大模型发布其实已经卷起来了。 黄金周前夕,先是国内大模型公司深度求索宣布推出DeepSeek-V3.2-Exp实验版模型,紧随其后智谱AI 正式发布新一代大模型GLM-4.6。 中国大模型也引起了图灵奖得主杨立昆的关注。就在9月30日,蚂蚁集团旗下百灵大模型宣布开源其思 考模型Ring-1T的预览版Ring-1T-preview,参数量达1000B(1万亿)。"令人印象 ...
粤商竞逐未来产业:谋技术突围,探创新之巅
21世纪经济报道记者 吴佳楠 广州报道 在全球经济与科技竞争的版图上,一场关于未来产业的话语权争夺战已悄然白热化。 过去,这片战场似乎总是由美国、日韩、欧洲的巨头们主导。但今天,一股来自中国南方的力量正以前 所未有的广度与深度全面参与角逐,并在许多方面留下了深刻的身影。他们拥有一个共同的名字——粤 商。 曾几何时,粤商二字总是与"三来一补""前店后厂""全球制造基地"等关键词紧密相连。他们凭借敏锐的 商业嗅觉、强大的制造能力和"敢为天下先"的魄力,将物美价廉的广东制造输送到全球每一个角落。 如今广东企业正在人工智能领域掀起技术奇袭,将人工智能应用到产业生态中,争夺智能时代的"入 口"。腾讯对标OpenAI的GPT系列推出混元大模型,在通用基础能力或专业应用能力上,都稳居国内第 一梯队;云天励飞则加码AI推理芯片,在政务信创、航天探测等场景中实现落地;云蝶科技聚焦教 育、医疗、工业等传统行业,打磨出"行知大模型",提供垂类大模型的解决方案。 云蝶科技创始人兼总裁田雪松告诉记者,人工智能和大模型未来会像水和电一样成为所有行业发展再升 级的基础设施,得益于对垂类大模型的深耕应用,企业自主研发的音视频识别分析算法和A ...
微信WeChat-YATT横空出世,腾讯强化学习布局剑指何方
Sou Hu Cai Jing· 2025-09-24 09:56
Core Insights - Tencent's open-sourcing of WeChat-YATT training library signifies a strategic move in the competitive landscape of AI model training, particularly as OpenAI's GPT-5 approaches release [1][2] - WeChat-YATT is designed with a focus on reinforcement learning and multimodal models, differentiating itself from mainstream frameworks like TensorFlow and PyTorch [2] Group 1: WeChat-YATT's Innovations - WeChat-YATT achieves significant breakthroughs in three areas: optimized parameter update efficiency for reinforcement learning, flexible multimodal data fusion interfaces, and a modular design that lowers the barriers for distributed training [2][4] - The library's emphasis on "ease of extensibility" reflects Tencent's recognition of the need for rapid iteration in large model training [4] Group 2: Competitive Positioning - Compared to Meta's PyTorch, WeChat-YATT excels in reinforcement learning support; against Google's JAX, it shows advantages in Chinese language scenarios and multimodal processing [4] - WeChat-YATT's deep integration with the WeChat ecosystem sets it apart from similar reinforcement learning frameworks like Ray RLlib [4] Group 3: Strategic Implications - The release of WeChat-YATT aligns with Tencent's broader AI strategy, which includes trademark applications for "WeChat AI Service Platform" and the deployment of the mixed Yuan model in business scenarios [7] - Tencent aims to create a closed-loop AI ecosystem through foundational technology breakthroughs and application deployment, with WeChat-YATT serving as a critical component in this strategy [7] - The focus on reinforcement learning indicates Tencent's commitment to key areas such as gaming, recommendation systems, and autonomous driving, positioning itself for future AI applications [7] Group 4: Long-term Vision - The naming of WeChat-YATT, "Yet Another Transformer Trainer," reflects both a sense of humor and Tencent's long-term investment in AI infrastructure [6] - The competition in the era of large models is fundamentally a competition for infrastructure, with WeChat-YATT representing a piece of Tencent's broader AI blueprint [7]
上汽总裁称余承东对尚界H5初始方案不满意,提了非常多要求;董明珠回应说话太满:是一种自信;李想直言苹果17 Pro Max太丑了
雷峰网· 2025-09-24 00:34
Key Points - Jinzi Ham (002515.SZ) is venturing into the semiconductor industry with a planned investment of up to 300 million yuan to acquire up to 20% equity in Zhongsheng Microelectronics, indicating a strategic shift towards high-growth sectors like AI and optical communication [3][5] - The company's stock surged to a daily limit of 7.85 yuan, resulting in a market capitalization increase of nearly 1 billion yuan on September 23 [3] - Jinzi Ham has previously attempted to diversify into healthcare and internet finance, but most of these ventures were unsuccessful [5] Group 1 - Jinzi Ham's investment in Zhongsheng Microelectronics reflects a broader trend of traditional companies diversifying into technology sectors [3][5] - The company aims to leverage the growth potential of the AI industry and optical communication markets [3] - The announcement of this investment comes after a change in the company's controlling shareholder and management team [5] Group 2 - The stock market reaction to Jinzi Ham's announcement demonstrates investor confidence in the company's new direction [3] - The investment strategy aligns with the increasing interest in semiconductor technology amid rising demand for AI applications [3][5] - Jinzi Ham's core business remains in meat products, but this move signifies a potential shift in focus towards technology-driven growth [3][5]
计算机行业动态报告:互联网大厂云出海元年开启
Minsheng Securities· 2025-09-23 11:45
Investment Rating - The report maintains a positive investment outlook for the cloud computing sector, particularly focusing on major internet companies' international expansion strategies [7]. Core Insights - The year 2025 marks the beginning of a new phase for major internet companies as they embark on cloud internationalization, supported by significant debt financing initiatives [1][10]. - Major players like Alibaba Cloud, Tencent Cloud, and ByteDance's Volcano Engine are actively enhancing their global cloud infrastructure to support Chinese enterprises' overseas expansion [17][23][50]. - The dual drivers of domestic AI internationalization and local AI demand abroad are propelling the cloud expansion efforts of these internet giants [2][3]. Summary by Sections 1. Internet Giants' Internationalization - Major internet companies are launching new rounds of debt financing aimed at cloud internationalization, with Alibaba planning to invest over 380 billion RMB in cloud and AI infrastructure over the next three years [1][11]. - Tencent Cloud has embraced internationalization, achieving high double-digit growth in international business over the past three years, with a doubling of overseas customer numbers [12][13]. 2. Domestic AI Internationalization and Local AI Demand - The demand for AI applications from both domestic companies and local markets abroad is driving the cloud internationalization process [2][3]. - Tencent, Alibaba, and ByteDance are leveraging their AI capabilities to support various sectors, including e-commerce and gaming, in their international endeavors [2][35][50]. 3. Investment Recommendations - The report suggests focusing on key players in the cloud platform sector, including Alibaba, Tencent, and Baidu, as well as companies involved in computing power leasing and optical communication [3][7]. - Specific companies recommended for investment include Hongjing Technology, Xiechuang Data, and various optical communication firms [3].
国内首个大模型“体检”结果发布,这样问AI很危险
3 6 Ke· 2025-09-22 23:27
Core Insights - The recent security assessment of AI large models revealed 281 vulnerabilities, with 177 being specific to large models, indicating new threats beyond traditional security concerns [1] - Users often treat AI as an all-knowing advisor, which increases the risk of privacy breaches due to the sensitive nature of inquiries made to AI [1][2] Vulnerability Findings - Five major types of vulnerabilities were identified: improper output vulnerabilities, information leakage, prompt injection vulnerabilities, inadequate defenses against unlimited consumption attacks, and persistent traditional security vulnerabilities [2] - The impact of large model vulnerabilities is less direct than traditional system vulnerabilities, often involving circumvention of prompts to access illegal or unethical information [2][3] Security Levels of Domestic Models - Major domestic models such as Tencent's Hunyuan, Baidu's Wenxin Yiyan, Alibaba's Tongyi App, and Zhiyun Qingyan exhibited fewer vulnerabilities, indicating a higher level of security [2] - Despite the lower number of vulnerabilities, the overall security of domestic foundational models still requires significant improvement, as indicated by a maximum score of only 77 out of 100 in security assessments [8] Emerging Risks with AI Agents - The transition from large models to AI agents introduces more complex risks, as AI agents inherit common security vulnerabilities while also presenting unique systemic risks due to their multi-modal capabilities [9][10] - Specific risks associated with AI agents include perception errors, decision-making mistakes, memory contamination, and potential misuse of tools and interfaces [10][11] Regulatory Developments - The National Market Supervision Administration has released 10 national standards and initiated 48 technical documents in areas such as multi-modal large models and AI agents, highlighting the need for standardized measures to mitigate risks associated with rapid technological advancements [11]
北交所市场周报250915-250919:指数震荡调整,专精特新板块彰显韧性,关注节前行情-20250922
Western Securities· 2025-09-22 11:52
Investment Rating - The report suggests a positive outlook for the industry, indicating that the long-term trend remains intact despite short-term fluctuations in the market [34]. Core Insights - The North Exchange market showed resilience in the specialized and innovative sectors, with significant stock performance variations, particularly in the automotive and new materials sectors [3][29]. - The report highlights the active mergers and acquisitions in high-end manufacturing and new materials, with specialized enterprises enhancing their competitiveness through industry chain integration [27][29]. - The macroeconomic environment is characterized by increased domestic consumer credit policies and a dovish stance from the Federal Reserve, which together create a supportive backdrop for the industry [27][28]. Summary by Sections Market Overview - The average daily trading volume for all A-shares on the North Exchange reached 27.23 billion yuan, a decrease of 14.7% week-on-week. The North Exchange 50 index fell by 1.43%, with an average turnover rate of 3.3% [7][16]. - The top five gainers for the week included Shichang Co. (271.6%), Kaiter Co. (33.8%), and Sanwei Co. (22.5%), while the top five losers included Yunchuang Data (-18.0%) and Guoyi Tendering (-13.3%) [16][19]. Key News and Policies - Nvidia has requested suppliers to develop MLCP technology, which is expected to be 3-5 times more expensive than current solutions, due to the high power consumption of its new AI platform [18]. - The Ministry of Commerce and nine other departments issued policies aimed at expanding service consumption, proposing 19 measures to stimulate economic growth [20]. - Tencent Cloud announced full compatibility with mainstream domestic chips, with its international customer base doubling within a year [21]. Core Driving Factors - The report identifies strong performance in the robotics, automotive parts, and specialty materials sectors, with significant stock price increases in companies like Shichang Co. [27][29]. - The ongoing trend of mergers and acquisitions is noted, with companies in the North Exchange continuing to pursue strategic acquisitions to enhance core technologies [29]. Investment Recommendations and Strategies - The report emphasizes that short-term index adjustments do not alter the long-term growth trajectory, supported by the solid fundamentals of specialized enterprises and favorable policies [32]. - Continuous policy benefits are expected to support sectors such as automotive parts, digital economy, and new energy materials [32]. - The active mergers and acquisitions driven by industry upgrades are seen as a means to strengthen core technologies and improve bargaining power within the supply chain [32].
中国(广西)—东盟人工智能企业交流会南宁开幕
Zhong Guo Xin Wen Wang· 2025-09-21 05:42
中新网南宁9月21日电(张广权)9月20日,中国(广西)—东盟人工智能企业交流会在广西南宁举办。 腾讯云区域解决方案总经理解来甲介绍了腾讯从底层算力到混元大模型的核心能力,以及面向企业端和 消费端的产品。他介绍,腾讯积极布局开源项目,并在新加坡、泰国等地展开跨境合作,希望让AI普 惠大众,助力产业创新。 柬埔寨邮电部数字政府转型司司长恩索万(Ensuowan)强调了人工智能在促进增长、改善治理方面的重要 作用。他指出,实现这一愿景需要区域合作与知识共享,并要确保AI发展遵循强有力的治理和道德原 则。 大会围绕"智汇广西,产融新机遇""Agentic AI赋能场景落地""东盟机遇与区域协同升级""跨境商业和营 销智能创新""协作经济和应用场景革新"等六大主题展开深入研讨。 大会还举办了中国—东盟轨道交通实训基地授牌仪式、中国—东盟人工智能企业联盟启动仪式、广西 AI+低空经济产教融合共同体揭牌仪式。(完) 本次大会作为中国—东盟博览会(下称"东博会")官方同期活动,以"共识中国AI,共建东盟生态,共享 数智未来"为主题,依托广西的区位优势与政策支持,汇聚中国与东盟政企学界资源,为中国AI企业出 海搭建重要平台。 ...