AI News AI资讯 2h ago Updated 2h ago 更新于 2小时前 46

Tencent's Tang Daosheng Responds to 'Tencent is Slow': Admits Business Lines Have Varied Speeds, Accepts External Feedback and Suggestions 汤道生回应'腾讯慢了':承认业务线快慢不一,接受外界提醒和建议

Admitting to being "slow"—and then what? Tencent Senior Vice President Dowson Tang's response reads like a polished yet unsurprising standard answer. At the AI Industry Summit, when faced with the direct punch of "Tencent is slow," he sidestepped gracefully, acknowledging that within a complex organization, different business lines move at different paces, some exploring while others failing, and he expressed openness to "outside reminders." The rhetoric was watertight, yet after hearing it, one 承认“慢了”,然后呢?腾讯高级副总裁汤道生的回应,像极了一份得体但缺乏惊喜的标准答案。在AI产业峰会上,面对“腾讯慢了”这记直拳,他侧身接住,承认了复杂组织里业务线有快有慢,有探索有失败,并对“外界提醒”表示开放。这套话术滴水不漏,但听完让人觉得,问题被承认了,而回答里却没藏着真正解决问题的“刀”。

70
Hot 热度
65
Quality 质量
60
Impact 影响力

Analysis 深度分析

Admitting to being "slow"—and then what? Tencent Senior Vice President Dowson Tang's response reads like a polished yet unsurprising standard answer. At the AI Industry Summit, when faced with the direct punch of "Tencent is slow," he sidestepped gracefully, acknowledging that within a complex organization, different business lines move at different paces, some exploring while others failing, and he expressed openness to "outside reminders." The rhetoric was watertight, yet after hearing it, one feels the problem has been acknowledged, but the answer lacks the real "knife" needed to solve it.

"Slow" is not a vague adjective but a glaring relative coordinate. As ByteDance’s Doubao large model surges forward, as Kimi carves out a niche in users’ minds with its long-text capabilities, and even as Baidu continues to generate buzz in applications like text-to-image generation, Tencent’s Hunyuan large model and its applications often give off a sense of detached complacency—sitting "steady on the fishing platform." This slowness may not stem from slow technological iteration, but from sluggish product intuition, user reach, and market offensive rhythm. Tang Dowson cited the "lobster craze" as an example of quick reaction, but this only exposes the core issue: Tencent seems better at following and harvesting proven trends (like short videos and game livestreams in the past) than defining the next generation of product forms. In this round of AI, Tencent is not firmly seated in the position of a definer.

Acknowledging that business lines have "fast and slow" parts sounds pragmatic, even tolerant. But when a giant is "slow," it often signals deeper trouble. Once a company reaches a certain scale, internal complexity, departmental silos, and KPI-driven decision-making processes can swallow any cutting-edge inspiration. Engineering culture yields to project management culture, and innovation shifts from "running red lights" to "waiting for green ones." This slowness is systemic—no amount of throwing more teams or publishing more papers will solve it. Tang said, "The model will keep iterating, and user needs will change"—which is true, but it sounds like "we’re still at the table" rather than "we’re ready to win the next hand." A top player doesn’t settle for just staying in the game.

More intriguing is the stance of "accepting outside reminders and suggestions." This is a politically correct, low-risk PR response. It softens sharp external criticism into "suggestions" for internal "reference," transforming deep questions about strategy and product philosophy into "information inputs" that can be absorbed by management processes. What large corporations excel at is "softening the blow"—turning an opponent’s heavy punch into an internal "opinion collection" exercise. True reflection should be evident in the next stunning product, the next decisive resource allocation, or the next "special zone" that dares to break existing business logic—not in verbal "acceptance" at a press conference.

Ultimately, Tencent’s slowness may be a luxury of inertia. Its two cash cows, WeChat and gaming, provide unparalleled fault tolerance and time windows, allowing it to choose safer, more certain paths rather than going "all in" like a startup. But the cruel side of AI is that it is reshaping all traffic and interaction gateways. Missing out on this round of definition could cost ten times more to catch up in the future. If the application layer is carved up, even the strongest cloud models and computing power might be reduced to "infrastructure," serving as a foundation for others to build upon.

In the end, Tang Dowson’s response may soothe shareholders and address public opinion, but it likely won’t spark much enthusiasm among developers or industry observers. We need to see Tencent’s "speed"—not the speed of chasing trends, but the boldness to cut inward, dismantle constraints, and empower a truly independent AI pioneering team with life-or-death authority to move "fast." Otherwise, today’s "accepting suggestions," like the past "embracing change," will become another footnote of a giant elegantly stumbling in a new era. Slowness isn’t a sin in itself, but when a company could be fast yet remains slow due to a "systemic constitution," and then tries to rationalize that state—that’s the root of the problem.

承认“慢了”,然后呢?腾讯高级副总裁汤道生的回应,像极了一份得体但缺乏惊喜的标准答案。在AI产业峰会上,面对“腾讯慢了”这记直拳,他侧身接住,承认了复杂组织里业务线有快有慢,有探索有失败,并对“外界提醒”表示开放。这套话术滴水不漏,但听完让人觉得,问题被承认了,而回答里却没藏着真正解决问题的“刀”。

“慢”不是一个模糊的形容词,而是一个刺眼的相对坐标。当字节跳动的豆包大模型应用狂飙突进,当Kimi凭借长文本在用户心智中撕开一道口子,甚至当百度在文生图等应用层面持续制造声量时,腾讯的混元大模型及其应用,却总给人一种“稳坐钓鱼台”的疏离感。这种“慢”,或许不是技术迭代的慢,而是产品直觉、用户触达和市场进攻节奏的慢。汤道生拿“龙虾热潮”举例,说反应快,但这恰恰暴露了问题的本质:腾讯似乎更擅长跟随和收割已被验证的热潮(比如当年的短视频、游戏直播),而不是定义下一代产品形态。AI这一轮,定义者的位置,腾讯目前坐得并不稳。

承认业务线“有快有慢”,听起来很务实,甚至很宽容。但巨头的“慢”,往往是更深层危机的征兆。当公司体量大到一定程度,内部的复杂性、部门墙、KPI导向的决策流程,足以吞噬掉任何前沿的灵感。工程师文化让位于项目管理文化,创新从“闯红灯”变成了“等绿灯”。这种“慢”是系统性的,不是多投入几个团队、多发几篇论文就能解决的。汤道生说“模型会不断迭代,用户需求会变化”,这话没错,但听起来像是在说“我们还在牌桌上”,而不是“我们准备赢下一把”。一个顶尖的牌手,不会满足于仅仅还在局中。

更耐人寻味的是“接受外界提醒和建议”这种姿态。这是一种政治正确的、低风险的公关回应。它把外部尖锐的批评,软化成了可供内部“参考”的“建议”,将一场关于战略和产品哲学的深刻质疑,转化为一个可以被管理流程所吸收的“信息输入”。大公司最擅长的就是把一切“化骨绵掌”,把对手的重拳变成自己内部的一次“意见收集”。真正的反思,应该体现在下一款让人惊艳的产品、下一次果断的资源倾斜、下一个敢于打破既有业务逻辑的“特区”上,而不是在一场发布会上的口头“接受”。

说到底,腾讯的“慢”可能是一种惯性奢侈。微信和游戏两大现金牛提供了无与伦比的容错率和时间窗口。这让它可以选择更稳妥、更追求确定性的路径,而不是像初创公司那样All in豪赌。但AI的残酷之处在于,它正在重塑所有流量和交互的入口。错过这一轮定义权,未来可能要用十倍的代价去追赶。当应用层被瓜分殆尽,云端的模型和算力再强,也可能沦为“基础设施”,为他人做嫁衣。

所以,汤道生的回应,安抚了股东,回应了舆论,但恐怕难以激起开发者或行业观察者的更多热情。我们需要看到腾讯的“快”,不是追赶热点的快,而是挥刀向内、砍掉桎梏、让一个真正独立的、拥有生杀大权的AI先锋团队“快”起来的魄力。否则,今天的“接受建议”,就和当年“拥抱变化”一样,会成为又一个巨头在新时代里优雅但步履蹒跚的注脚。慢不是原罪,但明明可以快却因为“系统性体质”快不起来,还试图将这种状态合理化,那才是问题的根源。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

大模型 大模型 产品发布 产品发布 伦理 伦理
Share: 分享到: