<p>新一任深度求索发言人的潜在出现引发了人们的好奇，因为创始人兼首席执行官梁文峰的去向仍然是一个谜，距离他上次公开露面已经过去一年多。尽管这家杭州初创公司的最新 V4 型号没有像一年前那样引起轰动，但备受期待的发布仍然通过与国内科技巨头华为的合作以及其极具竞争力的低价吸引了媒体的关注。对于这家传统上低调的公司来说，最近股权结构的变化表明梁对公司的控制力并未减弱。然而，最近几天的发展表明这家初创公司可能会有新的公众面孔。梁文峰在哪里？自去年二月与中国国家主席习近平进行高调的电视会议以来，梁文峰就没有公开露面，会议是在深度求索的 R1 型号发布后不久。然而，没有迹象表明他的领导角色发生了变化。根据企业数据库企查查，梁的股权比例最近从 1% 增加到 34%，其认缴资本从 10 万元（约合 14630 美元）跃升至 510 万元。深度求索的注册资本也从 1000 万元增加到 1500 万元。根据《南华早报》上周的报道，股权变化发生在深度求索首次开始筹集外部资金之际，引用了知情人士的话。根据中国公司法，关键公司决策需要超过三分之二投票权的股东批准。陈德利是谁？高级研究员陈德利是少数几位在 V4 发布后于周五在个人 X 账户上公开发帖的深度求索员工之一。他对近 24000 名粉丝写道：“484 天后，我们谦虚地分享我们的心血之作。和往常一样，我们坚持长期主义和开放源代码。” 这条简短的帖子与深度求索在前几年在线上的活跃表现形成了鲜明对比，当时几位研究人员在 X 上非常活跃，这是人工智能研究人员最受欢迎的在线聚集地。根据他在编码平台 GitHub 上的存档资料，陈在北京大学完成了本科和硕士学业，曾是计算机科学学院著名的语言计算与机器学习组（Lanco Lab）的成员。该实验室拥有多位校友，目前在中国主要公司中领导人工智能开发工作，包括小米的罗福利和阿里巴巴集团前 Qwen 团队技术负责人林俊扬。自 2023 年加入深度求索以来，陈的名字出现在该公司每一个主要型号的发布中，包括 V3、R1 和 V4。根据谷歌学术，陈的引用次数超过 22000 次，学术影响力与终身教授相当。陈说了什么？陈的第一次重大公开露面是在 2024 年 3 月，他代表深度求索及其母公司高飞量化在美国芯片巨头英伟达的旗舰 GTC 大会上发言。在 20 分钟的视频演讲中，陈解释了将人工智能模型与多样化人类价值观对齐的挑战，而是提出了一种 “解耦” 框架，以允许 “文化价值多样性”。他的下一个重大露面是在 11 月，当时他在一个国家支持的行业活动中代替梁，与其他五家公司的负责人一起出席，这些公司与深度求索共同被称为中国的 “六小龙”。在一段迅速传播的讲话中，陈对行业同仁和政府官员表示，他对人工智能对社会的整体影响持悲观态度，警告说大多数工作最终可能会被自动化，并呼吁人工智能公司充当 “吹哨人”，告诉公众哪些工作将首先被取代。深度求索正在失去明星研究员吗？鉴于中国对人工智能人才的激烈竞争，人们不断猜测深度求索是否能够将其明星研究员留在手中，而不被资金更雄厚的大型科技公司挖走。V4 的发布提供了一些答案。根据与该型号相关的技术论文，深度求索的研究和工程团队已扩大至 270 名员工，较 12 月初的 212 名增加了超过 27%。重要的是，论文确认深度求索在其开创性的 R1 型号中 18 名 “核心贡献者” 中的大多数仍在公司。只有两人离开：郭大雅，加入了抖音母公司字节跳动，张浩伟的去向仍然未知。</p>

深度求索

华为

<p>深度求索的创始人兼首席执行官梁文峰已经超过一年没有公开露面，这引发了对公司领导层的质疑。尽管如此，梁文峰的控制权似乎依然稳固，他的股权比例显著增加。高级研究员陈德利已成为潜在的新发言人，最近在主要行业活动中进行了演讲。深度求索最新的 V4 模型与华为合作发布，引起了关注，尽管其影响力未能达到之前模型的水平。公司扩展了研究团队，显示出在人工智能人才竞争中仍具韧性</p>

<p>The potential emergence of a new DeepSeek spokesman has sparked intrigue as speculation over the whereabouts of founder and CEO Liang Wenfeng continues, more than a year after his last public appearance. While the Hangzhou start-up’s latest V4 model did not make the same waves as its breakout moment a year earlier, the much-anticipated release still grabbed headlines through a collaboration with domestic tech giant Huawei Technologies and its remarkably low prices. For the traditionally low-profile firm, a recent change to its shareholding structure suggests that Liang’s control over the company has not waned. Still, developments in recent days point to a new public face for the start-up. Where is Liang Wenfeng? Liang has not appeared publicly since a high-profile televised meeting with Chinese President Xi Jinping in February last year, immediately after the release of DeepSeek’s R1 model. Yet there were no signs that his leading role had changed. According to corporate database Qichacha, Liang’s equity stake had recently increased to 34 per cent from 1 per cent, with his subscribed capital jumping from 100,000 yuan (US$14,630) to 5.1 million yuan. DeepSeek’s registered capital also rose from 10 million yuan to 15 million yuan. The change in shareholding came as DeepSeek began raising external capital for the first time, the South China Morning Post reported last week, citing people familiar with the matter. Under China’s company law, approval by shareholders representing more than two-thirds of voting rights is needed for key company decisions. Who is Chen Deli? Senior researcher Chen Deli was among a few DeepSeek personnel to post publicly on their own personal X accounts on Friday following the V4’s release. He wrote to his almost 24,000 followers: “484 days later, we humbly share our labour of love. As always, we stay true to long-termism and open source for all.” The short post contrasted with DeepSeek’s online presence in previous years, when several researchers were active on X, the most popular online gathering place for artificial intelligence researchers. According to his archived profile on coding platform GitHub, Chen completed both his undergraduate and master’s studies at Peking University, where he was a member of the famed Language Computing and Machine Learning Group (Lanco Lab) at the School of Computer Science. The lab boasts several alumni now leading AI development efforts at major Chinese companies, including Xiaomi’s Luo Fuli and Alibaba Group Holding’s former Qwen team technical lead Lin Junyang. Since joining DeepSeek in 2023, Chen’s name has appeared in every major model release from the firm, including V3, R1 and V4. According to Google Scholar, Chen had more than 22,000 citations to his name, placing him on the same level of academic impact as tenured professors. What has Chen said? Chen’s first major public appearance was in March 2024, when he spoke at US chip giant Nvidia’s flagship GTC conference on behalf of DeepSeek and its parent company High-Flyer Quant. In a 20-minute video presentation, Chen explained the challenge of aligning AI models with diverse human values, proposing instead a “decoupling” framework that allowed for “cultural value diversity”. His next major appearance was in November, when he stood in for Liang at a state-backed industry event alongside the heads of five other companies, which together with DeepSeek are collectively known as China’s “Six Little Dragons”. In remarks that went viral, Chen told industry peers and government officials he was pessimistic about AI’s overall impact on society, warning that most jobs could eventually be automated and calling for AI companies to act as “whistle-blowers” by telling the public which jobs would be made redundant first. Is DeepSeek losing its star researchers? Given China’s cutthroat competition for AI talent, there has been constant speculation about whether DeepSeek could keep its star researchers out of the hands of Big Tech firms with much deeper pockets. The V4 release provided some answers. According to the technical paper accompanying the model, DeepSeek’s research and engineering team had expanded to 270 staff, up from 212 in early December, a gain of more than 27 per cent. Crucially, the paper confirmed that most of DeepSeek’s 18 “core contributors” to its groundbreaking R1 model were still at the company. Only two had departed: Guo Daya, who joined TikTok owner ByteDance, according to domestic media reports, and Zhang Haowei, whose next destination remained unknown.</p>
<div class="lb-trans"><p>新一任深度求索发言人的潜在出现引发了人们的好奇，因为创始人兼首席执行官梁文峰的去向仍然是一个谜，距离他上次公开露面已经过去一年多。尽管这家杭州初创公司的最新 V4 型号没有像一年前那样引起轰动，但备受期待的发布仍然通过与国内科技巨头华为的合作以及其极具竞争力的低价吸引了媒体的关注。对于这家传统上低调的公司来说，最近股权结构的变化表明梁对公司的控制力并未减弱。然而，最近几天的发展表明这家初创公司可能会有新的公众面孔。梁文峰在哪里？自去年二月与中国国家主席习近平进行高调的电视会议以来，梁文峰就没有公开露面，会议是在深度求索的 R1 型号发布后不久。然而，没有迹象表明他的领导角色发生了变化。根据企业数据库企查查，梁的股权比例最近从 1% 增加到 34%，其认缴资本从 10 万元（约合 14630 美元）跃升至 510 万元。深度求索的注册资本也从 1000 万元增加到 1500 万元。根据《南华早报》上周的报道，股权变化发生在深度求索首次开始筹集外部资金之际，引用了知情人士的话。根据中国公司法，关键公司决策需要超过三分之二投票权的股东批准。陈德利是谁？高级研究员陈德利是少数几位在 V4 发布后于周五在个人 X 账户上公开发帖的深度求索员工之一。他对近 24000 名粉丝写道：“484 天后，我们谦虚地分享我们的心血之作。和往常一样，我们坚持长期主义和开放源代码。” 这条简短的帖子与深度求索在前几年在线上的活跃表现形成了鲜明对比，当时几位研究人员在 X 上非常活跃，这是人工智能研究人员最受欢迎的在线聚集地。根据他在编码平台 GitHub 上的存档资料，陈在北京大学完成了本科和硕士学业，曾是计算机科学学院著名的语言计算与机器学习组（Lanco Lab）的成员。该实验室拥有多位校友，目前在中国主要公司中领导人工智能开发工作，包括小米的罗福利和阿里巴巴集团前 Qwen 团队技术负责人林俊扬。自 2023 年加入深度求索以来，陈的名字出现在该公司每一个主要型号的发布中，包括 V3、R1 和 V4。根据谷歌学术，陈的引用次数超过 22000 次，学术影响力与终身教授相当。陈说了什么？陈的第一次重大公开露面是在 2024 年 3 月，他代表深度求索及其母公司高飞量化在美国芯片巨头英伟达的旗舰 GTC 大会上发言。在 20 分钟的视频演讲中，陈解释了将人工智能模型与多样化人类价值观对齐的挑战，而是提出了一种 “解耦” 框架，以允许 “文化价值多样性”。他的下一个重大露面是在 11 月，当时他在一个国家支持的行业活动中代替梁，与其他五家公司的负责人一起出席，这些公司与深度求索共同被称为中国的 “六小龙”。在一段迅速传播的讲话中，陈对行业同仁和政府官员表示，他对人工智能对社会的整体影响持悲观态度，警告说大多数工作最终可能会被自动化，并呼吁人工智能公司充当 “吹哨人”，告诉公众哪些工作将首先被取代。深度求索正在失去明星研究员吗？鉴于中国对人工智能人才的激烈竞争，人们不断猜测深度求索是否能够将其明星研究员留在手中，而不被资金更雄厚的大型科技公司挖走。V4 的发布提供了一些答案。根据与该型号相关的技术论文，深度求索的研究和工程团队已扩大至 270 名员工，较 12 月初的 212 名增加了超过 27%。重要的是，论文确认深度求索在其开创性的 R1 型号中 18 名 “核心贡献者” 中的大多数仍在公司。只有两人离开：郭大雅，加入了抖音母公司字节跳动，张浩伟的去向仍然未知。</p>
</div>

DeepSeek mystery: who is speaking for start-up amid CEO Liang Wenfeng’s absence?

南华早报

小米集团-W

阿里巴巴-W

阿里巴巴

英伟达

小米集团(ADR)

小米集团-WR

Xiaomi HK SDR 2to1

阿里巴巴-WR

Alibaba HK SDR 5to1

NVIDIA Corporation

深度求索的谜团：在 CEO 梁文峰缺席期间，谁在代表这家初创公司发声？