章节 1：AI 竞赛现状：OpenAI 与 Google 的双雄对决

📝 本节摘要：

本节主要探讨了 AI 领域的当前竞争格局。主持人 Alex 与 Snowflake CEO Sridhar Ramaswamy 讨论了 OpenAI 与 Google 之间的角力——前者与 Nvidia 结成“不自在的联姻”，后者则拥有模型与 TPU 的垂直整合优势。Sridhar 指出，顶尖模型厂商（OpenAI、Anthropic、Gemini）与其他竞争者之间存在巨大鸿沟，但鉴于技术迭代极快，任何领先地位都极其脆弱。他强调比赛尚处于早期阶段，但现有模型在创造价值（特别是软件工程领域）方面已展现出深远潜力。

[原文] [Alex Kantrowitz]: Welcome to Big Technology Podcast a show for coolheaded and nuance conversation of the tech world and beyond We have a great show for you today We're going to talk about the state of the AI race looking at the Open AI versus Google access with someone who really knows what's going on in the competition And we'll also take a look at the state of AI agents and what AI programs can do when they organize their data Well we have the perfect guest to do it with us here today Sar Ramaswami is here He is the CEO of Snowflake third time on the show Welcome back Swedar

[译文] [Alex Kantrowitz]: 欢迎来到 Big Technology Podcast，这是一个对科技界及其他领域进行冷静且细致对话的节目。今天我们为大家准备了精彩的内容。我们将探讨 AI 竞赛的现状，特别是聚焦 OpenAI 与 Google 之间的较量，我们要连线一位真正了解这场竞争内幕的专家。我们还将审视 AI 代理（AI agents）的发展现状，以及 AI 程序在组织好数据后能发挥什么作用。好吧，我们今天请到了最完美的嘉宾来和我们要一起讨论，Sridhar Ramaswamy 来了。他是 Snowflake 的 CEO，这是他第三次做客本节目。欢迎回来，Sridhar。

[原文] [Sridhar Ramaswamy]: Alex always great to talk to you Thank you for having me

[译文] [Sridhar Ramaswamy]: Alex，和你聊天总是很棒。谢谢邀请我。

[原文] [Alex Kantrowitz]: So it's been a couple years since we've spoken For those who uh don't know you you spent 15 years at Google Uh your last job there was the SVP of ads and commerce You founded Neva an ads-free search engine in 2019 You sold that to Snowflake in 2023 You became the CEO of Snowflake in 2024 Snowflake for the uninitiated 59 billion public company It is a data cloud company which stores analyzes and helps you share data And you really have a front seat to the AI race So let's begin with the AI race Just give us your perspective on the state of the AI race Now it seemed like for a while there was OpenAI and the rest Now it seems like there's two axes that are forming the I'll call it the uncomfortable marriage of OpenAI and Nvidia and then the uh the Google side of things where they have the model the TPUs and they seem to be giving the incumbent a run for their money What's your perspective

[译文] [Alex Kantrowitz]: 我们有几年没聊了。对于那些不了解你的人来说，你在 Google 待了 15 年，你在那里的上一份工作是广告与商业高级副总裁（SVP）。你在 2019 年创立了无广告搜索引擎 Neeva，并在 2023 年将其卖给了 Snowflake。你在 2024 年成为了 Snowflake 的 CEO。对于不了解 Snowflake 的人来说，这是一家市值 590 亿美元的上市公司。它是一家数据云公司，负责存储、分析并帮助你共享数据。你确实坐在 AI 竞赛的前排席位上。所以让我们从 AI 竞赛开始吧。请谈谈你对 AI 竞赛现状的看法。有一段时间看起来像是 OpenAI 独领风骚，其他人紧随其后。现在看来形成了两个轴心：我称之为 OpenAI 和 Nvidia 之间那种“不自在的联姻”（uncomfortable marriage），然后是 Google 这边，他们拥有模型和 TPU（张量处理单元），而且看起来正在给领先者带来激烈的竞争。你的观点是什么？

[原文] [Sridhar Ramaswamy]: first of all the AI race changes every month We should all feel great about making predictions because one of them will come true and it'll the world will change enough that we have to make new predictions I think u the gap between the truly great model makers of the present era it's like open AI the anthropic and Gemini very much in that mix and uh everyone else is quite staggering and it's also a world in which no incumbent should feel comfortable about their position because things are changing so much and a great new model can sometimes end up producing a lead that's like a year long which is an eternity in uh in today's world and um and so I would say from that perspective it's early there's a lot of change what is also quite profound about this moment is the things that we can get done with the models that have already been launched where it's merely an issue of stuff like mechanics for can you get inference capacity it's a lot easier to solve I think that's the part that sometimes people overlook about what is remarkable about this moment these models they can do amazing things we'll get into some of the things that we snowflake are doing I think it is their ability to create value their ability to help among the most priced of professions today software engineering I think that's the thing that will drive so much impact

[译文] [Sridhar Ramaswamy]: 首先，AI 竞赛每个月都在变化。我们都应该对做预测感到兴奋，因为其中某个预测会成真，然后世界会发生足够大的变化，以至于我们不得不做出新的预测。我认为在这个时代，真正伟大的模型制造者——比如 OpenAI、Anthropic 和 Gemini 这一梯队——与其他人之间的差距是相当惊人的。而且这也是一个没有任何在位者（incumbent）应该对自己的位置感到舒适的世界，因为事情变化太快了，一个伟大的新模型有时能产生长达一年的领先优势，这在当今世界简直就是永恒。所以我想说，从这个角度看，现在还很早，充满了变数。在这个时刻同样深刻的是，利用已经发布的模型我们能完成的事情，现在这仅仅变成了诸如“你能否获得推理能力（inference capacity）”这样的机制问题，这是更容易解决的。我认为这正是人们有时忽略的当前时刻的非凡之处——这些模型能做令人惊叹的事情。我们稍后会深入探讨我们在 Snowflake 做的一些事情。我认为是它们创造价值的能力，是它们帮助当今最昂贵的职业之一——软件工程——的能力，我认为这将是驱动如此巨大影响力的关键因素。

[原文] [Sridhar Ramaswamy]: Lots more to come but I would say it's very very early in the AI race

[译文] [Sridhar Ramaswamy]: 未来还有很多发展，但我想说，AI 竞赛目前还处于非常非常早期的阶段。

章节 2：巨头的反击：Google 的适应力与领先优势的脆弱性

📝 本节摘要：

在本节中，Alex 列举数据指出，尽管 OpenAI 仍处于领先地位，但其增长速度已不及 Google（Gemini），后者正凭借强大的分发优势和底层算力（TPU）迎头赶上。Sridhar 分析了科技巨头在创新初期的“分娩阵痛”，但强调一旦 Google 这样的公司找准方向，其 DeepMind 团队、资金储备和基础设施将成为巨大的加速器。他回顾了在 Google 任职期间的“Code Yellow”危机文化，指出 Google 具备极强的适应性。对话最后得出一个关键结论：在当前的 AI 竞赛中，任何技术领先都极度脆弱且短暂，竞争格局瞬息万变。

[原文] [Alex Kantrowitz]: I I agree with you and I want to drill down on this a little bit because you are somebody who has the mentality that sort of is needed to analyze what's going on You're not only somebody who spent more than a decade at Google including time at as in the highest ranks of the company you competed with Google Mhm And so if it's like when we think about what's going on with the AI race now Google is this it's a beast and it has this distribution advantage and in fact we recently published some data on big technology that showed that OpenAI had opened up a very big lead It's still growing quickly It's grown 50% web visits uh January 2025 to January 2026 But the lead is shrinking and uh Google has for instance grown its web visits by not 50% like OpenAI but 647% in the same time period When you say web visits you mean for things like Gemini correct yeah Not just Google itself Yeah The the chatbot visits for Gemini Um

[译文] [Alex Kantrowitz]: 我同意你的观点，我想就这一点深入探讨一下，因为你具备分析当前局势所需的那种思维方式。你不仅在 Google 待了十多年，跻身公司最高管理层，你还曾与 Google 竞争过。嗯。所以当我们思考 AI 竞赛的现状时，Google 是一头巨兽，它拥有分发优势。事实上，我们在 Big Technology 上最近发布的一些数据显示，OpenAI 曾建立了巨大的领先优势，且仍在快速增长——从 2025 年 1 月到 2026 年 1 月，其网络访问量增长了 50%。但是这种领先优势正在缩小，相比之下，Google 的网络访问量在同一时期增长了不是 50%，而是 647%。你说网络访问量是指像 Gemini 这样的产品对吧？是的，不仅仅是 Google 主站。是的，是 Gemini 聊天机器人的访问量。

[原文] [Alex Kantrowitz]: and some of the the aura around opening I was predicated on it having this lead and not letting it go In fact Sam Alman I think he was in India and he was like you could try to build a model like ours but it won't work Y and now with things like Deep Seek Communic 2 we've seen people able to catch up on that front So it's being pushed by Google on one hand the open source model builders on the other Help me figure out how open AI can can continue to lead this this race if it can or is it just one in the pack

[译文] [Alex Kantrowitz]: 围绕 OpenAI 的部分光环是建立在它拥有这种领先优势且不会放手的基础上的。事实上，Sam Altman——我想他当时是在印度——曾说过类似“你们可以尝试构建像我们这样的模型，但这行不通”的话。是的。而现在随着像 DeepSeek 等事物的出现，我们看到人们在这一领域已经能够追赶上来。所以它一方面受到 Google 的挤压，另一方面受到开源模型构建者的挤压。请帮我分析一下，OpenAI 如何才能继续在这场竞赛中领跑？或者它是否仅仅变成了众多竞争者中的一员？

[原文] [Sridhar Ramaswamy]: i mean I think the fact that it has become OpenAI has become the Google of choice when it comes to chat for most of us that's actually a durable advantage And uh I I I I use it quite often for all kinds of things including solving problems in the real world My coffee machine not working um or I can't open my gate anymore like it the amount of use that you can get is pretty remarkable I think that lead is real On the other hand something pretty simple like not simple it's hard faster image generation or more accurate image generation which is what uh Google pioneered with Nano Banana It's actually having a profound impact on things like their usage and OpenAI was late to the game just for that one feature You think come on it's a small feature How much can it matter it matters People like being able to create things It just tells you that yes competition is actually very fierce and u big companies generally have a lot of birthing issues when it comes to new things It's just it's a matter of how they work First of all they don't often have a clear perspective of what amazing means um in a new area And uh what they struggle with even if they can understand amazing is figuring out a path to that amazing

[译文] [Sridhar Ramaswamy]: 我的意思是，我认为对于我们大多数人来说，在聊天方面 OpenAI 已经成为了当年的 Google，这实际上是一个持久的优势。我经常用它做各种事情，包括解决现实世界的问题，比如我的咖啡机坏了，或者我打不开大门了，你能从中获得的用途是非常显著的。我认为这种领先是真实的。另一方面，像更快或更准确的图像生成这样看似简单（其实很难）的事情——这是 Google 通过 Nano Banana（注：此处可能指 Google 特定的快速生成模型或项目代号）率先推出的——实际上对其使用率产生了深远的影响，而 OpenAI 仅仅因为这一项功能就落后了。你会想，拜托，这是一个小功能，能有多大关系？但这确实有关系，人们喜欢能够创造东西。这只是告诉你，是的，竞争实际上非常激烈。大公司在面对新事物时通常会有很多“分娩阵痛”（birthing issues）。这只是他们运作方式的问题。首先，他们通常对在新领域中什么是“惊艳”没有清晰的视角。而且，即使他们能理解什么是“惊艳”，他们也很难找到通往那种“惊艳”的路径。

[原文] [Sridhar Ramaswamy]: One can argue that uh XAI for example has actually produced what is widely acknowledged to be a world-class model that is out there But that act of sheer creation is not something that anyone should take for granted It doesn't matter how many how much resources you have It's not that easy to figure out all the little things that you have to get right in order to get to a point like that You see other companies with tons of money struggling to be at the same caliber as OpenAI and Anthropic Google now has had a set of pretty deep advantages in this area They kept deep mind quite separate and deep mind was all has always been at the cutting edge of AI and it's become a real weapon for them in terms of getting to the front and once they get there all of the other advantages that they have of distribution the bottomless uh you know well of money that they can borrow from investments in things like TPUs which kind of looked crazy back then that we would invest in it All of those become accelerants But I think what one should take away is that uh like that breakthrough which is so hard to achieve especially for big companies with specialties Google has managed to achieve This just means that open AI and anthropic need to understand that any kind of lead that they get is not going to be a long lived one and they really have to work hard and compete Honestly I think that's a good thing for uh all of us

[译文] [Sridhar Ramaswamy]: 人们可以说，例如 xAI 实际上已经生产出了被广泛认为是世界级的模型。但这种纯粹的创造行为不应被任何人视为理所当然。不管你有多少资源，要弄清楚所有那些必须做对的小细节才能达到那个水平，并不是那么容易的。你看到其他拥有巨额资金的公司在努力达到与 OpenAI 和 Anthropic 同等的水平时都很挣扎。Google 现在在这个领域拥有一系列相当深厚的优势。他们让 DeepMind 保持相当的独立性，而 DeepMind 一直处于 AI 的前沿，这成了他们冲到前列的真正武器。一旦他们到了那里，他们拥有的所有其他优势——分发渠道、无底洞般的资金池（可以用来投资像 TPU 这样的东西，虽然当时投资看起来很疯狂）——所有这些都会成为加速器。但我认为人们应该从中领悟到的是，像那种突破——尤其是对于有特定专长的大公司来说很难实现的突破——Google 已经做到了。这只是意味着 OpenAI 和 Anthropic 需要明白，他们获得的任何领先优势都不会长久，他们真的必须努力工作并参与竞争。老实说，我认为这对我们所有人来说都是件好事。

[原文] [Sridhar Ramaswamy]: Just to give you some points of comparison GPD4 by all accounts was ready in August 2022 Long time ago And um it took anthropic I would say roughly 2 years summer of 2024 to have a model that was of comparable quality to GPD4 like two whole years which is an eternity And then soon after Anthropic launched a coding model that was widely acknowledged to be the state-ofthe-art and they have stayed there It took OpenAI and Google again a year plus to catch up to that It tells you that leads are shrinking and uh there's going to be more and more competition and of course there's a pressure from things like the open- source models to just turn this into a whole other ballgame in terms of what is possible with them

[译文] [Sridhar Ramaswamy]: 只是为了给你一些比较的基准：据各方说法，GPT-4 在 2022 年 8 月就已经准备好了，那是很久以前的事了。大概花了 Anthropic 两年时间——也就是到 2024 年夏天——才拿出一个质量与 GPT-4 相当的模型。整整两年，这（在 AI 领域）简直就是永恒。然后不久之后，Anthropic 发布了一个被广泛认为是目前最先进（state-of-the-art）的编码模型，并且保持了这一地位。OpenAI 和 Google 又花了一年多的时间才追上。这告诉你，领先优势正在缩小，竞争将会越来越多，当然还有来自开源模型等方面的压力，就其可能性而言，这会让整个比赛变成一场完全不同的游戏。

[原文] [Alex Kantrowitz]: On the Google front given the time that you spent there are you are you surprised at what's happened there it seems like they just kind of woke up and started shipping with a sense of urgency that I hadn't seen from them for a while

[译文] [Alex Kantrowitz]: 关于 Google 这方面，鉴于你在那里待过的时间，你对那里发生的事情感到惊讶吗？看起来他们就像突然醒悟了一样，开始带着一种我已经很久没在他们身上看到的紧迫感去发布产品。

[原文] [Sridhar Ramaswamy]: Google always had um and the founders definitely they were always well calibrated for crisis I remember back in 2005 when uh wislive.com the precursor to Bing first came out with what appeared to be a really good search engine We uh got into what's called a cordolo It's like meet every day all hands on deck drop everything else We got to be faster better than them What was it called it was called live.com But the it was just a it was called a cord yellow It's basically get the teams together show up in front of Larry tell them what you're doing today And then they went to code with this uh open AI thing at a certain point Yeah Yeah But the point is um and every year that I have been at Google I can think of one or more crises that required us to operate very differently and what looks like a placid company from outside is very motivated very driven They've also struggled with uh structural boundaries Um for example the thing that we did uh for a social network which was called I forget remember Emerald C++ Um that was sort of a disaster because you know it's it's it's first of all it's hard to be it's hard for a new player to break through especially with uh something like a network effect of a social network is just really really hard to do

[译文] [Sridhar Ramaswamy]: Google 一直都有——尤其是创始人肯定有——对危机的良好校准能力。我记得早在 2005 年，当 live.com（Bing 的前身）刚推出一个看起来非常不错的搜索引擎时，我们进入了所谓的“Code Yellow”（橙色警报）状态。这就像是每天开会，全员上阵，放下其他一切事情，我们必须比他们更快、更好。那个叫什么来着？叫 live.com。但这只是一个……它被称为 Code Yellow。基本上就是把团队召集起来，出现在 Larry（Page）面前，告诉他你们今天在做什么。后来在应对 OpenAI 这个事情的某个节点，他们也进入了代码战状态。是的，是的。但重点是，我在 Google 的每一年，我都能想到一个或多个危机，要求我们以非常不同的方式运作。从外面看似乎是一家平静的公司，其实非常有动力、非常有干劲。他们也在结构性边界上挣扎过。例如，我们做的那个社交网络项目，叫什么来着，我忘了，记得是 Emerald C++（注：此处可能指 Google+ 或其前身项目 Emerald Sea），那算是一场灾难。因为你知道，首先作为一个新玩家很难突围，尤其是在具有网络效应的社交网络领域，真的真的很难做。

[原文] [Sridhar Ramaswamy]: Um and so they struggle with new things that uh they do but they've also demonstrated an ability to adapt uh Google Cloud by you know Google Cloud is a pretty big success Obviously a lot of credit goes to Thomas for making that making that happen It is an adaptable company It is a malleable company So it's I'm not surprised and uh you know I'm not that close to Google anymore but folks speak about how one of the really cool things about DeepMind is having uh Sergey in the mini kitchen just hanging out talking to people and so that that sense of time that sense of what is a pivotal moment that's what great leaders bring and Google's always had that in spades I remember uh when Google Plus launched um I actually was supposed to go to meet a friend at Facebook that weekend and they were supposed to have their their barbecue their company barbecue and they canceled it and I was like what happened and he's like don't you realize we're at war That's correct And it seems like that's really what's happened with both Google and OpenAI two code reds That's what greatness takes right for you to realize these crucible moments and go all out

[译文] [Sridhar Ramaswamy]: 嗯，所以他们在做新事物时会挣扎，但他们也展示了适应能力。Google Cloud，你知道，Google Cloud 是一个相当大的成功。显然，这很大程度上要归功于 Thomas（Kurian）让它得以实现。这是一家适应性强的公司，一家可塑性强的公司。所以我并不惊讶。而且，你知道，我虽然不再那么接近 Google 了，但人们都在谈论 DeepMind 一个非常酷的事情，就是 Sergey（Brin）会出现在迷你厨房里，闲逛并和人们交谈。那种对时机的感知，那种对什么是关键时刻的感知，正是伟大的领导者所带来的，而 Google 一直都拥有这种特质。我记得当 Google+ 发布时，我那个周末本来要去见一个在 Facebook 的朋友，他们本来要举办公司烧烤聚会，结果取消了。我就问：“发生什么事了？”他说：“你难道没意识到我们正处于战争状态吗？”没错。看起来这正是 Google 和 OpenAI 之间正在发生的事情，两个“Code Red”（红色警报）。这就是卓越所需要的，对吧？让你意识到这些严峻的考验时刻，并全力以赴。

章节 3：聚焦企业级市场：OpenAI 战略与 Snowflake 的合作

📝 本节摘要：

本节聚焦于 OpenAI 进军企业级市场的战略及其与 Snowflake 的合作。面对关于 OpenAI 缺乏聚焦（同时涉足消费者、视频、设备及企业端）的质疑，Sridhar 认为市场的评判往往是事后诸葛亮，成败决定了外界对战略的评价。他详细介绍了 Snowflake 与 OpenAI 的战略合作伙伴关系，以及新推出的代理平台“Snowflake Intelligence”。Sridhar 强调，企业级 AI 的核心不在于引入全新的事物，而在于利用现有数据更高效地创造价值。他分享了一个拥有 500 万 SKU 的制造商案例，说明了如何利用 AI 代理系统优化定价策略，从而挖掘数亿美元的潜在增收。

[原文] [Alex Kantrowitz]: So the question is where to focus right there was uh there were some reports recently that um Nvidia CEO Jensen Wong has been saying privately that he doesn't love uh OpenAI's business approach and you could read that as maybe as the finances Uh I really read that as as a criticism of focus and I could be speculating here but OpenAI is doing the consumer chatbot They're doing video generation models They're doing the device and they're doing enterprise now And enterprise is actually going to be a big push for them this year And in fact you're part of partnership with them Yep Just announced a $200 million partnership with OpenAI Um and I think for our purposes it would be great to hear your perspective on why enterprise is a worthwhile bet for them and where they stand compared to anthropic which has been focused on enterprise from the beginning

[译文] [Alex Kantrowitz]: 所以问题在于应该聚焦哪里，对吧。最近有一些报道称，Nvidia CEO Jensen Huang（黄仁勋）私下表示他不太喜欢 OpenAI 的商业模式。你可以将其解读为可能是财务方面的原因，但我真的将其解读为对聚焦度的批评。我这可能是在猜测，但 OpenAI 既在做消费者聊天机器人，又在做视频生成模型，还在做设备，现在又在做企业级市场。而且企业级市场实际上将是他们今年的一大推力。事实上，你们也是他们合作伙伴的一部分。是的。刚宣布了与 OpenAI 达成 2 亿美元的合作伙伴关系。嗯，我认为对于我们的讨论来说，听听你的观点会很棒：为什么企业级市场对他们来说是一个值得下的赌注？以及与从一开始就专注于企业级市场的 Anthropic 相比，他们处于什么位置？

[原文] [Sridhar Ramaswamy]: One issue we should all keep in mind is that um when you're seizing lots of ground when times are early if you're successful people will call you a genius If on the other hand they don't go well and a threat shows up in the main thing that you do people will say lack of focus for the longest time Google was criticized for being a onetrick pony in search and after a while it was criticized for having too many efforts that lacked focus and now we are back to putting Google as a hero because they succeeded in uh Gemini So we should all remember that judgments are postfact and dependent on the outcomes produced rather than the actual strategy There's a little bit of that

[译文] [Sridhar Ramaswamy]: 我们都应该记住的一点是，在早期阶段，当你在这个领域攻城略地时，如果你成功了，人们会称你为天才。另一方面，如果进展不顺利，而你做的主要事情上出现了威胁，人们就会说你缺乏聚焦。很长一段时间以来，Google 因在搜索领域是“只会一招的小马”（one-trick pony）而受到批评；过了一段时间，它又因搞了太多缺乏聚焦的项目而受到批评；而现在，因为他们在 Gemini 上取得了成功，我们又回过头来把 Google 视为英雄。所以我们都应该记住，评判往往是事后的（post-fact），取决于产生的结果，而不是实际的战略。确实有那么一点这样的成分。

[原文] [Sridhar Ramaswamy]: Having said that OpenAI has a lot to offer uh enterprises and um we are excited to partner with them because many customers are joint customers of Snowflake and of OpenAI We've created an uh agentic platform called Snowflake Intelligence that's been quite transformative Over 2,000 customers fastest growing product over 2,000 customers are using it pretty much uh you know 3 months after we released the product to G Enterprise customers are fussy about using products only um in uh in NGA and uh it's among our fastest growing products ever launched and uh it's it's focused on data and snowflake

[译文] [Sridhar Ramaswamy]: 话虽如此，OpenAI 确实能为企业提供很多东西。嗯，我们很高兴能与他们合作，因为许多客户既是 Snowflake 的客户，也是 OpenAI 的客户。我们创建了一个名为 Snowflake Intelligence 的代理（agentic）平台，它非常具有变革性。超过 2,000 家客户——这是增长最快的产品——超过 2,000 家客户正在使用它，这几乎是在我们将产品发布到 GA（通用可用性阶段，原文口误为 G/NGA）后的 3 个月内实现的。企业客户对于使用产品非常挑剔，通常只在产品正式发布后才用，而这是我们有史以来推出的增长最快的产品之一。它的核心聚焦于数据和 Snowflake。

[原文] [Sridhar Ramaswamy]: Back to your point about focus we wanted to make sure that we created a product that could enhance the value of things that people had already done with Snowflake We didn't want to go and pitch our enterprise customers and say "Hey we're doing something dramatically new you know work on it with us We said you can get value from your data a whole lot faster Not only that we also said we live what we preach And so I often show them things like our sales agent which puts the every piece of information that my sales team has about every customer at my fingertips What meetings did this customer have yesterday what are the outstanding use cases all of that is available to me But it's also programmable I can I can get the information the way I want share it the way I want

[译文] [Sridhar Ramaswamy]: 回到你关于聚焦的观点，我们想确保我们创造的产品能够提升人们已经在 Snowflake 上所做事情的价值。我们不想去向企业客户推销说：“嘿，我们在做一个全新的东西，你知道，来和我们一起搞吧。”我们说的是：“你可以更快地从你的数据中获取价值。”不仅如此，我们还说我们言行一致（live what we preach）。所以我经常向他们展示像我们的销售代理（sales agent）这样的东西，它将我的销售团队关于每个客户的每一条信息都放在我的指尖。这个客户昨天开了什么会？有哪些未解决的用例？所有这些对我来说都是可用的。而且它还是可编程的。我可以按我想要的方式获取信息，按我想要的方式分享它。

[原文] [Sridhar Ramaswamy]: And uh but there's a lot more in this world of agents and enterprise How do you help people take action how do you help people be better grounded about the consequences of their action how do you help them analyze situations these are the things that we are excited to be collaborating with OpenAI on Yes one part of it is us using their models but I think the much more interesting thing is going to be what are areas that are very amenable to AI creating value and how do we make sure that we make it easy for enterprises to realize that value

[译文] [Sridhar Ramaswamy]: 此外，在代理和企业级市场这个领域里还有更多的东西。你如何帮助人们采取行动？你如何帮助人们更好地基于行动的后果来做决定？你如何帮助他们分析情况？这些正是我们很高兴能与 OpenAI 合作的内容。是的，其中一部分是我们使用他们的模型，但我认为更有趣的事情将是：哪些领域非常适合 AI 创造价值？以及我们如何确保让企业能够轻松实现这一价值？

[原文] [Sridhar Ramaswamy]: To make this super concrete I was visiting a big uh manufacturer yesterday They make my eyes kind of popped out when they said you know listen we have 5 million SKs 5 million SKs that they sell And uh part of their issue is we have trouble pricing this because it's it's a big dynamic marketplace We don't know what competitors are pricing it at We don't know what kind of like you have to take into account the margin that we have on the product the NPS for the product Can you create an agentic system that can help us do pricing better we have all our data on snow And that's an that is a situation in which the power of agentic technology the ability to look at a complex situation break it down follow best practices for how work should be done is going to be a big um multiplier for how they get their work done

[译文] [Sridhar Ramaswamy]: 为了让这一点超级具体，我昨天拜访了一家大型制造商。他们说的话让我大吃一惊，他们说：“听着，我们有 500 万个 SKU（库存单位）。”他们销售 500 万个 SKU。他们的问题之一是：“我们在定价上有困难，因为这是一个巨大的动态市场。我们不知道竞争对手定价多少。我们不知道……比如你必须考虑到我们产品的利润率、产品的 NPS（净推荐值）。”“你们能创建一个代理系统来帮助我们更好地定价吗？我们所有的数据都在 Snowflake 上。”这就是代理技术发挥威力的一种情况——这种能够审视复杂情况、将其分解、遵循最佳工作实践的能力，将成为他们完成工作的一个巨大的倍增器。

[原文] [Sridhar Ramaswamy]: There's potentially hundreds of millions of dollars of additional revenue that this company can make if they can do a better job just with this one single project That gives you an example of the kind of things that people are looking to do together with um with with with OpenAI and Anthropic and a data platform like Snowflake

[译文] [Sridhar Ramaswamy]: 如果这家公司仅在这个单一项目上做得更好，他们就有可能获得数亿美元的额外收入。这给你提供了一个例子，说明人们希望通过与 OpenAI、Anthropic 以及像 Snowflake 这样的数据平台合作来做些什么。

章节 4：工作的未来：代理（Agentic）AI 的实战应用与去伪存真

📝 本节摘要：

本节深入探讨了 AI 代理（Agentic AI）如何重塑工作流程。Sridhar 描绘了未来的工作形态：从被动查看邮件和待办事项，转变为主动向 AI 系统下达指令（如分析定价机会、生成报告），让人类专注于决策而非繁琐的“跑腿工作”。他以 Snowflake 内部为例，介绍了客户支持团队如何利用 AI 工具将调试复杂案例的时间缩短了 10 倍，以及销售代理如何实时提供客户情报。针对市场上关于 AI 代理是“炒作”的质疑，Sridhar 认为这是因为大多数高管尚未体验过真正创造价值的工具，并强调必须通过实际应用（Walk the walk）来证明其潜力。

[原文] [Alex Kantrowitz]: So how does the product work it would be a a agent basically that goes and takes a look at the pricing and then with the GPT model I mean explain exactly what

[译文] [Alex Kantrowitz]: 那么这个产品是如何运作的呢？基本上是一个代理去查看定价，然后结合 GPT 模型……我是说，请确切解释一下它是怎么回事。

[原文] [Sridhar Ramaswamy]: Well this is a great question and it it uh goes to a topic that I'm pretty passionate about I call it what does work look like in the future and uh today our work is pretty much we go look at our email we go look at our to-do list and then decide what are the things that uh we should be that we should be doing or you know if you're like me you have meetings on calendar where um where work shows up Uh the future that we envision very much is um you describe what you want systems to do Hey these are the kinds of things that I should be looking at every day For example I look at our revenue alerts every day Mhm I go and look at the dashboard If there's a if there is a a big up or a big down I send out questions and so on Very automatable

[译文] [Sridhar Ramaswamy]: 嗯，这是一个很好的问题，它涉及到一个我非常有激情的话题，我称之为“未来的工作是什么样的”。今天我们的工作基本上就是去查看邮件，查看待办事项清单，然后决定我们应该做什么，或者如果你像我一样，日历上排满了会议，工作就在那里出现。我们设想的未来在很大程度上是：你描述你希望系统做什么。“嘿，这些是我每天应该关注的事情。”例如，我每天都会查看我们的收入警报。嗯。我会去看仪表盘。如果有大幅上涨或下跌，我会发出询问等等。这非常适合自动化。

[原文] [Sridhar Ramaswamy]: And so you have an agentic system that is connected both to the past information that's typically sitting in snowflake or what was performance-like Um it is also it has access to things like prediction models that say if something changes what does the future look like also things like ambient information your emails your documents other or even things like the stock market um ambient information about the world and uh your work very much becomes these are the five topics that you should be paying attention to and here is a brief for these five topics and potentially even recommendations So you give the agent a task you give it basically like you would an employee you give it these this instruction

[译文] [Sridhar Ramaswamy]: 所以你会拥有一个代理系统，它既连接到通常存储在 Snowflake 中的过去信息（比如过去的表现如何），也能访问预测模型（比如如果某些变量改变，未来会是什么样），还能访问环境信息——你的邮件、文档，甚至像股票市场这样的外部世界环境信息。你的工作在很大程度上变成了：“这是你应该关注的五个主题，这是关于这五个主题的简报，甚至可能包含建议。”所以你给代理一个任务，基本上就像你给员工布置任务一样，你给它这些指令。

[原文] [Sridhar Ramaswamy]: Uh if you are let's say the manufacturer right you say "Hey I want you to take a look at the pricing." Say "I want you to look at the spread between how I price how the market is pricing identify the top 10 opportunities I should be paying attention to in my department today generate a report for me." My job is okay I'm going to go through this go through the recommendation and figure out what do I change and if I want to make a change what approvals do I need to get within the company so it does the leg work for you You come in and your decision is based your your task is basically to make the decisions as opposed to spend a week looking at all the information

[译文] [Sridhar Ramaswamy]: 如果你是那个制造商，你会说：“嘿，我想让你看看定价。”比如说：“我想让你看看我的定价与市场定价之间的价差，找出我部门今天应该关注的前 10 个机会，并为我生成一份报告。”我的工作就变成了：好吧，我要浏览这个报告，浏览建议，弄清楚我要改变什么；如果我想做出改变，我需要在公司内部获得什么批准。所以它为你做了跑腿的工作。你进来后，你的任务基本上是做决策，而不是花一周时间去查看所有信息。

[原文] [Sridhar Ramaswamy]: And the magical thing about this by the way we are living this with our support team We have changed our support team from 50 people writing software 300 people using this software to help debug support cases to much more of a builder user model where there are a set of tools available within our coding agent Cortex code And whenever a support case comes they use these tools to analyze what is happening And then they tell the customer what to do And sometimes they decide you know these tools are not enough I need to build a new tool And they add that tool itself to the suite of tools that everyone else can use So this is work self-correcting getting itself better over time And the goal is just things get done a whole lot faster already we are seeing 10x not 10% 10x reductions in the amount of time that it takes to debug complex cases that come in

[译文] [Sridhar Ramaswamy]: 顺便说一句，这其中神奇的地方在于，我们正在我们的支持团队中亲身实践这一点。我们将支持团队从“50 人写软件、300 人使用软件来调试支持案例”的模式，转变为一种更倾向于“构建者-用户”的模式。在我们的编码代理 Cortex Code 中有一套可用的工具。每当有一个支持案例进来时，他们就使用这些工具来分析发生了什么，然后告诉客户该怎么做。有时他们会决定：“你知道吗，这些工具还不够，我需要构建一个新工具。”然后他们会将那个新工具添加到其他人都可以使用的工具套件中。所以这是一种自我修正、随着时间推移自我完善的工作方式。目标就是让事情完成得快得多。我们已经看到了 10 倍——不是 10%——而是 10 倍的时间缩减，用于调试进来的复杂案例。

[原文] [Alex Kantrowitz]: and so let's just go to this question of is this working because there's been a lot of discussion of agentic AI uh every time we talk about it there's always like a segment of the audience that says you know this is still a lot of hype push back harder um conceptual largely still And um you know this is something that you know might in demos look really good but when you actually put it into practice uh it struggles What's what is your read on that

[译文] [Alex Kantrowitz]: 那么让我们来谈谈“这是否真的有效”这个问题，因为关于代理 AI（Agentic AI）有很多讨论。每次我们谈论它时，总有一部分听众会说：“你知道，这主要还是炒作，反弹得很厉害，目前很大程度上还只是概念性的。”而且你知道，这东西可能在演示中看起来非常好，但当你真正将其投入实践时，它就会遇到困难。你对此怎么看？

[原文] [Sridhar Ramaswamy]: you got to walk the walk We were in Davos together Yes and uh you know two weeks ago and I probably met 20 odd CEOs CIOS lots of partners and uh my sort of SOP standard operating procedure for each of these meetings would be I would ask our sales agent for information about the customer What's the state of our relationship with take your pick and uh it generates a report I would turn it on and show my phone to them Um and they would go "Holy cow." But uniformly not one of these CEOs has the same tools that I do I see That's the difference between actually getting the work done making AI serve meaningful needs um and yes the hype that you're describing All of the people that are um in the camp that you're describing have never had useful products built for them that deliver meaningful value

[译文] [Sridhar Ramaswamy]: 你必须言行一致（walk the walk）。我们之前一起在达沃斯。是的。大概两周前，我可能见了大約 20 位 CEO、CIO 和很多合作伙伴。我每次会议的 SOP（标准操作程序）是，我会向我们的销售代理询问有关该客户的信息。“我们与某某公司的关系状况如何？”（随你怎么选），然后它会生成一份报告。我会打开它，把我的手机展示给他们看。他们会说：“天哪（Holy cow）。”但也无一例外，这些 CEO 中没有一个人拥有和我一样的工具。我明白了。这就是“真正把工作做完、让 AI 服务于有意义的需求”与你所描述的“炒作”之间的区别。所有那些属于你描述的（怀疑论）阵营的人，从来没有拥有过为他们构建的、能提供有意义价值的有用产品。

[原文] [Sridhar Ramaswamy]: I speak as somebody that lives this the amount of feedback that my poor team gets about how difficult the mobile experience is how to make it better We just launched like Face ID authentication That's a big deal because I don't have to log in um all the time It's taking care of all of those kinds of nuances making enterprise data come alive available for you and then helping you with decisioning That's the magic and that's why you are hearing people say it's hype But it's companies like Snowflake that are actually living what we are preaching

[译文] [Sridhar Ramaswamy]: 我是作为一个亲身实践者在说话。我可怜的团队收到了大量关于移动端体验有多困难、如何改进的反馈。我们刚刚推出了 Face ID 认证，这很重要，因为我不必每次都登录了。它照顾到了所有这些细微之处，让企业数据变得鲜活、随手可得，然后帮助你做决策。这就是魔力所在，也是为什么你会听到人们说这是炒作（因为他们没见过）。但像 Snowflake 这样的公司正在真正践行我们所宣扬的。

[原文] [Sridhar Ramaswamy]: I'll give you one more small example of something that is cooking this very week I'm working with our uh ops team our operations team that helps manage snowflake the software running in the cloud about how to get on a more agentic bandag like you know super cruddly infrastructure engineers that are like what is this you know we know better but we're walking through this journey of no no let's create tools that our coding agent can use and you will genuinely find that is a lot easier and so someone created a tool that will help detect things like oh are there problems with um warehouses resuming Warehouse is a basic unit of work that gets stuff done for our customers And when our customer says start this we wanted to start quickly Um in like 10 seconds I had generated a histogram of resume times put a nice graph and I sent it to the team with one prompt all English um on top of a tool that somebody had built to look at resume times in warehouses and the team is like "Holy cow that's the magic of uh Asgentic platforms." But yes you have to do the leg work to put them into place put the guardrails things like that But there's real magic here

[译文] [Sridhar Ramaswamy]: 我再给你举一个小例子，就在这周正在发生的事情。我正在与我们的运维团队（Ops team）合作——他们负责管理在云端运行的 Snowflake 软件——探讨如何转向更代理化的方式。你知道，那些超级顽固的基础设施工程师会觉得：“这是什么啊？我们懂的更多。”但我们正在经历这个过程，告诉他们：“不，不，让我们创建一些让编码代理可以使用的工具，你会真心发现这要容易得多。”于是有人创建了一个工具来帮助检测诸如“仓库恢复（resuming）是否存在问题”之类的事情——仓库是为客户完成工作的基本计算单元，当客户说“启动这个”时，我们要它快速启动。大概只用了 10 秒钟，我就生成了一个恢复时间的直方图，画出了一张漂亮的图表，并把它发给了团队。这一切只需一个纯英文的提示词（prompt），运行在某人构建的用于查看仓库恢复时间的工具之上。团队的反应是：“天哪，这就是代理平台的魔力。”不过是的，你必须做些基础工作来把它们部署到位，设置护栏之类的。但这里确实有真正的魔力。

章节 5：开发范式转移：用 AI 加速构建 AI 产品

📝 本节摘要：

本节讨论了 AI 如何从根本上改变软件开发的节奏与经济学。主持人引用 Mistral CEO 的观点，认为企业级 AI 的落地通常是缓慢的“托管服务”过程。Sridhar 对此反驳，介绍了 Snowflake 刚发布的“Cortex Code”数据编码代理，它能用自然语言自动化处理数据库设置、模型构建等复杂任务。他将这种“用 AI 加速构建 AI 产品”的能力描述为“红药丸时刻”（Red Pill moment），意味着发布新功能的门槛大幅降低。Alex 随后引用 Ben Thompson 的分析，指出虽然代码生成变得廉价，但软件公司的护城河将转向合规与维护，同时竞争将从“做大蛋糕”变为激烈的“存量博弈”。

[原文] [Alex Kantrowitz]: Couple of things So first of all what you're saying is kind of reminding me of something that Arthur from Mistral the CEO of Mistral said here couple weeks ago which is basically that the technology has these capabilities but it's not just like it's not like in that AGI mode tell it what to do and it can it can work it it in in many ways getting enterprise AI to work is a managed service which means that it could take some time for what you're talking about to be visible within the entire economy as opposed to those who have already put the time to figure it out

[译文] [Alex Kantrowitz]: 有几件事。首先，你所说的让我想起了 Mistral 的 CEO Arthur 几周前在这里说过的话。大意是，这项技术虽然具备这些能力，但它并不像那种 AGI（通用人工智能）模式，告诉它做什么它就能搞定。在许多方面，让企业级 AI 真正运作起来是一项托管服务（managed service），这意味着你所描述的效果可能需要一段时间才能在整个经济体中显现出来，而不只是在那些已经投入时间去研究的人身上。

[原文] [Sridhar Ramaswamy]: well that's also where magic can happen right and uh you know I told you that uh we uh released a new product called Cortex code which is our data coding agent Uh we launched it GA yesterday and uh it dramatically lowers the amount of time that it takes to get stuff done on Snowflake Mhm We all get carried away with how does AI make it easier for a business user like me to get access to my data That's great But on the other hand everything from how do you set up a database to how do you move data from a production like on like a transaction database over to Snowflake for analysis How do you build a machine learning model how do you build an agent that you can then give to the business user cortex code is meant to address all of that again in natural language

[译文] [Sridhar Ramaswamy]: 嗯，但这也就是魔力发生的地方，对吧。我告诉过你，我们发布了一个名为 Cortex Code 的新产品，这是我们的数据编码代理。我们昨天刚刚全面发布（GA）了它，它极大地减少了在 Snowflake 上完成任务所需的时间。嗯。我们往往沉迷于 AI 如何让像我这样的业务用户更容易获取数据，这很棒。但另一方面，从“如何设置数据库”、“如何将数据从生产环境（比如交易数据库）迁移到 Snowflake 进行分析”、“如何构建机器学习模型”，到“如何构建一个可以交付给业务用户的代理”，Cortex Code 旨在通过自然语言再次解决所有这些问题。

[原文] [Sridhar Ramaswamy]: And part of what we have built there are what we call a series of skills that help automate this work And this is a theme that's going to come up again and again which is how do you use AI to make launching AI products go faster right that's the feedback loop that one needs to be on It's a little bit I it's a little bit of a red pill moment where you're like wait you mean I can release new software products pretty much every day because releasing a new piece of functionality is as simple as writing a recipe in English which all of us are very capable of doing I think using AI to make AI go a lot faster is something that we are excited about and uh this product is among the best in terms of how do you get it from Snowflake

[译文] [Sridhar Ramaswamy]: 我们在其中构建的部分功能是我们所谓的“技能（skills）”系列，旨在帮助自动化这些工作。这将是一个反复出现的主题：你如何利用 AI 来加快 AI 产品的发布速度？对吧，这就是我们需要进入的反馈循环。这有点像是一个“红药丸时刻”（Red Pill moment，指觉醒时刻），你会惊叹：“等一下，你是说我可以几乎每天都发布新的软件产品？因为发布一项新功能就像用英语写食谱一样简单，而这是我们所有人都非常有能力做到的。”我认为利用 AI 让 AI 发展得更快是我们感到兴奋的事情，而就如何从 Snowflake 获取价值而言，这个产品是目前最好的之一。

[原文] [Alex Kantrowitz]: It's uh it's interesting that you talk about how easy it is to build software now That has been both a a benefit for software companies and something that people are worried about because where is where does the moat uh look you know where is the moat if it's so easy to build This is from uh this is from Ben Thompson's pretty interesting his his perspective He says AI coding doesn't kill software Customers pay for products not code They are paying for support compliance integrations security patches someone else owning the neverending maintenance commitment That stuff doesn't just go away because writing the initial app got cheaper

[译文] [Alex Kantrowitz]: 你谈到构建软件现在变得多么容易，这很有趣。这对软件公司来说既是一个好处，也是人们担心的事情，因为护城河在哪里？你知道，如果构建如此容易，护城河在哪里？这来自 Ben Thompson，他的观点非常有趣。他说：“AI 编码不会杀死软件。客户通过购买产品付费，而不是代码。他们是在为支持、合规、集成、安全补丁以及‘其他人拥有永无止境的维护承诺’而付费。”这些东西不会仅仅因为编写初始应用程序变得更便宜而消失。

[原文] [Alex Kantrowitz]: There's a butt here though He says "But if every software company can write infinite code cheaply the competitive dynamics change." Yep The SAS playbook of finding a niche and growing your slice worked when building was expensive Now everyone can build into adjacencies overnight Shifts from growing this pie to it shifts everything from growing the pie to fighting for share It's something that you know it seems like you're enabling and you're living

[译文] [Alex Kantrowitz]: 不过这里有个“但是”。他说：“但是，如果每家软件公司都能廉价地编写无限的代码，竞争动态就会改变。”是的。当构建成本高昂时，“找到一个小众市场并做大你的蛋糕”这种 SaaS 剧本是行之有效的。现在，每个人都可以在一夜之间通过构建进入邻近领域。这导致重心从“做大这个蛋糕”转变为“争夺市场份额”。这似乎正是你在推动并在经历的事情。

[原文] [Sridhar Ramaswamy]: Yeah I think there is going to be a concentration towards platform players Um but I would also be cautious about general pronouncements for the simple reason that we are all actors in this space We all get to change the outcome I feel very good about Snowflake as a data platform But I honestly do not want to be in a situation where access to snowflake is always mediated through someone else That's always a very dangerous place to be especially in a moment like this

[译文] [Sridhar Ramaswamy]: 是的，我认为将会出现向平台型玩家集中的趋势。但我对这种一般性的断言也会持谨慎态度，原因很简单：我们都是这个领域的一份子，我们都有机会改变结果。作为数据平台，我对 Snowflake 感觉很好。但老实说，我不希望处于一种“访问 Snowflake 总是需要通过其他人中介”的境地。那总是一个非常危险的位置，尤其是在这样一个时刻。

章节 6：软件商业模式变革：平台化、SaaS 估值与“被管道化”风险

📝 本节摘要：

本节深入探讨了 AI 对软件行业商业模式的冲击。Alex 引用 Ben Thompson 的观点指出，随着 AI 让代码生成变得廉价，SaaS 行业将从“做大蛋糕”转向激烈的“存量博弈”。Sridhar 认同行业将向平台型玩家集中，并强调 Snowflake 拒绝被中间层“中介化”。随后，两人讨论了 Anthropic 法律插件导致传统法律软件股价暴跌的案例，揭示了垂直软件沦为大模型“输入端”（被管道化）的风险。针对 SaaS 估值普跌的现象，Sridhar 分析认为，市场正在惩罚那些仅将 AI 视为“附加组件（Bolt-on）”的公司，而真正的赢家必须重构工作流，提供完整的端到端体验。

[原文] [Alex Kantrowitz]: There's a butt here though He says "But if every software company can write infinite code cheaply the competitive dynamics change." Yep The SAS playbook of finding a niche and growing your slice worked when building was expensive Now everyone can build into adjacencies overnight Shifts from growing this pie to it shifts everything from growing the pie to fighting for share It's something that you know it seems like you're enabling and you're living

[译文] [Alex Kantrowitz]: 不过这里有个“但是”。他说：“但是，如果每家软件公司都能廉价地编写无限的代码，竞争动态就会改变。”是的。当构建成本高昂时，“找到一个小众市场并做大你的蛋糕”这种 SaaS 剧本是行之有效的。现在，每个人都可以在一夜之间通过构建进入邻近领域。这导致重心从“做大这个蛋糕”转变为“争夺市场份额”。这似乎正是你在推动并在经历的事情。

[原文] [Sridhar Ramaswamy]: Yeah I think there is going to be a concentration towards platform players Um but I would also be cautious about general pronouncements for the simple reason that we are all actors in this space We all get to change the outcome I feel very good about Snowflake as a data platform But I honestly do not want to be in a situation where access to snowflake is always mediated through someone else That's always a very dangerous place to be especially in a moment like this

[译文] [Sridhar Ramaswamy]: 是的，我认为将会出现向平台型玩家集中的趋势。但我对这种一般性的断言也会持谨慎态度，原因很简单：我们都是这个领域的一份子，我们都有机会改变结果。作为数据平台，我对 Snowflake 感觉很好。但老实说，我不希望处于一种“访问 Snowflake 总是需要通过其他人中介”的境地。那总是一个非常危险的位置，尤其是在这样一个时刻。

[原文] [Alex Kantrowitz]: Uh but there was an interesting thing that just happened uh this week that I think we should talk about which is you made such an interesting point where what when I when I asked you about this you said listen we do not want to be an input into somebody else's software and uh this week Anthropic released uh or or within the most recent days anthropic released a legal plugin and the market got wind of this and then all of a sudden Thompson Reuters I think it had its worst day on the market in history stocks like Legal Zoom just you know dropped like a rock And I think and I was trying to think through like why this could be because it was just one legal plugin from Anthropic And the perspective might be that with generative AI there is a risk that some software shifts from being the place you do the things You know Lexus Nexus you do the research there to an input into a platform And if that's the case I think what the market is thinking is that you lose that control that you had You become you become a feature in a platform as opposed to the platform itself That's the risk

[译文] [Alex Kantrowitz]: 嗯，但这周发生了一件有趣的事，我觉得我们应该谈谈。你之前提出了一个非常有趣的观点，当时我问你这个问题，你说：“听着，我们不想成为别人软件的输入端。”而就在这周——或者说最近几天——Anthropic 发布了一个法律插件。市场闻风而动，突然之间，Thomson Reuters（汤森路透）好像迎来了其历史上最糟糕的一个交易日，像 LegalZoom 这样的股票简直就像石头一样直线下跌。我在试着思考为什么会这样，因为这仅仅是 Anthropic 的一个法律插件而已。观点可能是，随着生成式 AI 的出现，存在一种风险，即某些软件会从“你做事的地方”（比如你在 LexisNexis 做研究）转变为“某个平台的输入”。如果是这种情况，我认为市场的想法是你失去了原本拥有的控制权。你变成了一个平台中的功能，而不是平台本身。这就是风险。

[原文] [Sridhar Ramaswamy]: It's a very real risk I think people that were confident about their position in the world because they were essentially wall gardens for data and uh functionality and are slow at providing modern ways of dealing with information are going to struggle in this world This is the reason that uh I stress us living by what we speak in terms of AI and agentic platforms and this future of work concept precisely because unless you live it you don't actually feel it And uh unless you live it and feel it you're not going to help your customers get there I think um na niche SAS software providers that basically benefited from lock in Think about it If you used a piece of SAS software logged into it on your browser God help you if you want your data back Just like not going to happen Yeah What this current moment is pointing out is that that's a very dangerous place to be And a lot of these players risk becoming dumb backends to the models which is why Snowflake is so leaning forward on agentic AI and living by what we speak because that's the place where value is going to get created

[译文] [Sridhar Ramaswamy]: 这是一个非常真实的风险。我认为那些对自己在这个世界上的地位充满信心的人——因为他们本质上是数据和功能的“围墙花园”（walled gardens），并且在提供处理信息的现代方式上行动迟缓——在这个世界上将会举步维艰。这也是我强调我们在 AI 和代理平台以及“未来工作”概念上要言行一致的原因，确切地说是因为除非你亲身经历，否则你不会真正感受到它。而且除非你亲身经历并感受到它，否则你无法帮助你的客户到达那里。我认为那些基本上受益于锁定的利基 SaaS 软件提供商……你想想看，如果你使用某个 SaaS 软件，在浏览器上登录进去，如果你想要回你的数据，那真是上帝保佑你了。这基本上是不可能发生的。是的。当前的时刻指出，那是一个非常危险的位置。许多这类玩家面临着成为模型“哑后端”（dumb backends）的风险，这就是为什么 Snowflake 如此积极地通过代理 AI 向前倾斜，并践行我们所宣扬的，因为那是价值将被创造的地方。

[原文] [Alex Kantrowitz]: The market doesn't really seem to know what it's doing when it comes to software Doesn't really seem to know how to value software in this moment Uh this is from Liz Thomas Uh she says "Software's forward 12-month price to equity ratio has compressed from 33.1 to 23.2 multiple contraction of 30%." Which is wild because software gets these big valuations because of what it is Here's another stat SAS index from Talia Goldberg SAS index is down 32% year-over-year despite most companies meeting or beating plans while the markets are up 15% Is what do you think the market's reaction is here is it just we had Brett Taylor on he said it was just kind of the uncertainty of who wins Is that your perspective or why do you think um despite like like you know Talia is saying here the the fact that these companies are meeting their earnings expectations they're still getting hammered and the multiples are contracting

[译文] [Alex Kantrowitz]: 市场在涉及软件时似乎真的不知道自己在做什么，似乎真的不知道在这个时刻该如何为软件估值。这是来自 Liz Thomas 的数据，她说：“软件行业的前瞻 12 个月市盈率已从 33.1 压缩至 23.2，倍数收缩了 30%。”这很疯狂，因为软件通常因其本质而获得高估值。这里还有另一个数据，来自 Talia Goldberg 的 SaaS 指数：SaaS 指数同比下跌了 32%，尽管大多数公司达到或超过了计划，而大盘却上涨了 15%。你认为市场的反应是什么？是因为……我们之前请过 Bret Taylor，他说这只是因为“谁会赢”的不确定性。这是你的观点吗？还是你认为为什么……尽管像 Talia 所说的，这些公司达到了盈利预期，却仍然受到重创，估值倍数在收缩？

[原文] [Sridhar Ramaswamy]: There are a few things that we should take into consideration here As you know companies are valued not on what they're doing today but on what they're going to do in the future And uh I would actually distinguish data platforms like uh Snowflake from pure software providers operating on a subscription model Not that it's a bad model but the way they have operated is AI became another skew for these folks and uh customers have had to sign up for AI products regardless of whether they created value or not That's sort of become the favored way of becoming AI native I think what the current moment points to is a real risk that that is not a winning AI strategy meaning that work is not going to get done by interacting with a chatbot on a particular SAS app that you used

[译文] [Sridhar Ramaswamy]: 这里有几件事我们需要考虑。如你所知，公司的估值不是基于它们今天在做什么，而是基于它们未来将做什么。而且，我实际上会将像 Snowflake 这样的数据平台与以订阅模式运营的纯软件提供商区分开来。并不是说那是一个糟糕的模式，但他们运营的方式是：AI 变成了这些人的另一个 SKU（库存单位），客户必须注册 AI 产品，不管它们是否创造了价值。这某种程度上成了变身为“AI 原生”的首选方式。我认为当前的时刻指出，这存在一个真正的风险，即那不是一个制胜的 AI 战略——意味着工作将不再通过与你使用的某个特定 SaaS 应用程序上的聊天机器人交互来完成。

[原文] [Sridhar Ramaswamy]: Which is why our our vision of agents operating on a data platform that has much of the analytic insights about the past as a lot of our customers do but with the ability to bring in integrations via MCP via other APIs for how do you talk to other systems I think that's the compelling vision I think companies are going to win if they have both a convincing vision for how work gets done in the future but are able to back it up with and here is how we help you the customer get it done fast The model makers approach it from the from this view of the model is everything and nothing else matters We approach it from the viewpoint of it's the entirety of the experience It's the model That's why we partner with all of these folks It's the most critical data that's valuable to your company but it's also integrations with the operational systems that really help get work done I think that's the compelling vision for how work gets ready What the markets are in some ways pricing is the fact that AI as a bolt-on to SAS software does not feel like a winning strategy

[译文] [Sridhar Ramaswamy]: 这就是为什么我们的愿景是：代理在数据平台上运行，该平台拥有大量关于过去的分析洞察（正如我们许多客户所拥有的），但同时具备通过 MCP（Model Context Protocol）或其他 API 引入集成的能力，以实现与其他系统的对话。我认为这才是有说服力的愿景。我认为，如果公司既有关于“未来工作如何完成”的令人信服的愿景，又能通过“这是我们要如何帮助你（客户）快速完成它”来支撑这一愿景，它们就会赢。模型制造者是从“模型就是一切，其他都不重要”的视角切入的。我们则是从“这是整体体验”的视角切入的。它是模型——这也是我们与所有这些人合作的原因；它是对你公司有价值的最关键数据；但它也是与真正帮助完成工作的运营系统的集成。我认为这才是关于工作如何就绪的令人信服的愿景。市场在某种程度上定价反映的事实是：将 AI 作为 SaaS 软件的“附加组件（bolt-on）”感觉并不像是一个制胜的战略。

章节 7：“影子 AI”崛起：自下而上的采用与安全挑战

📝 本节摘要：

本节聚焦于“影子 AI（Shadow AI）”现象，即员工绕过 IT 部门私自使用 AI 工具的趋势。Sridhar 认为这是企业采用 AI 的主要驱动力，源于 AI 带来的“10 倍效率提升”。他分享了儿子在几小时内用开源工具搭建个人 AI 助理的案例，展示了新一代员工的行动力。Sridhar 强调，企业不应简单禁止，而应成为“进步型组织（Progressive Organizations）”：在确保安全（如不使用公司笔记本运行未经授权的代码）的前提下，识别并授权内部的“AI 冠军”，将这种自下而上的创新转化为企业的正式能力。

[原文] [Alex Kantrowitz]: Okay I want to talk to you about the um about shadow AI and how people are individuals are starting to build their own AI programs We've seen that a lot over the past couple weeks Yeah Um so let's do that when we come back right after this And we're back here on big technology podcast with Sedar Rama Swami CEO of Snowflake Swedar great to have you on the show Thank you for coming back Always great to chat What did you think when this open claw clawbot multbot moment happened when people started running all their own agents on their on their computers and doing crazy things

[译文] [Alex Kantrowitz]: 好的，我想和你聊聊“影子 AI（Shadow AI）”，以及个人如何开始构建他们自己的 AI 程序。过去几周我们经常看到这种情况。是的。那么让我们在广告回来后继续这个话题。欢迎回到 Big Technology Podcast，我们依然和 Snowflake 的 CEO Sridhar Ramaswamy 在一起。Sridhar，很高兴你能上节目。谢谢你回来。聊得很开心。当这个 OpenClaw、Clawbot、Multbot（注：此处可能指各类开源或多模态代理机器人）时刻发生时，当人们开始在自己的电脑上运行各种代理并做一些疯狂的事情时，你是怎么想的？

[原文] [Sridhar Ramaswamy]: well I hope they were not running them on their own computers but still some were and got their API keys exposed and Exactly Exactly No I think uh old rules of security don't vanish because of uh because because of AI It's uh remarkable I'm fortunate in that I have two young sons who are both in software and uh you know I get to see the world through their eyes and as it turns out one of them had one day between uh when he came to San Francisco from he moved from New York and when he started his job on Tuesday and uh in that one day when I was at work and he was home he had uh managed to get like you know an Ubuntu instance on AWS completely separate from everything else including his laptop thank God

[译文] [Sridhar Ramaswamy]: 嗯，我希望他们不是在自己的（公司）电脑上运行这些东西，但确实有些人这么做了，还导致 API 密钥泄露了。正是，正是。不，我认为旧的安全规则不会因为 AI 而消失。这很了不起。我很幸运，我有两个年轻的儿子，他们都是做软件的。你知道，我可以通过他们的眼睛看世界。事实证明，其中一个儿子从纽约搬到旧金山，在他周二开始新工作之前有一天的空闲时间。在那一天，当我在工作而他在家时，他设法在 AWS 上搞了一个 Ubuntu 实例，与包括他的笔记本电脑在内的其他所有东西完全隔离——感谢上帝。

[原文] [Sridhar Ramaswamy]: And um he had uh uh set up OpenClaw on it as his personal AI assistant and uh it comes with things like uh Telegram integrations You can talk to it He started using it as his to-do list and he set up a little chat bot for giving me a summary of cool AI happenings on X because I told him like X can be a lot I don't like to spend that much time on it I still want to get what's important So I get like a briefing every day of cool things happening in uh AI done entirely by the chatbot Tell him not to productize that cuz I could be in trouble if he does I'll get I think it took all of a few hours for him to do that Build this newsletter

[译文] [Sridhar Ramaswamy]: 然后他在上面设置了 OpenClaw（注：可能是指 OpenInterpreter 或类似的开源代理工具）作为他的个人 AI 助手。它带有像 Telegram 集成之类的功能，你可以和它对话。他开始用它做待办事项清单，还设置了一个小聊天机器人，给我总结 X（原 Twitter）上发生的很酷的 AI 动态。因为我告诉他，X 上的信息量太大了，我不喜欢在那上面花太多时间，但我仍然想知道什么是重要的。所以我每天都会收到一份关于 AI 领域酷事的简报，完全是由这个聊天机器人完成的。（Alex 插话：告诉他别把这个产品化，否则我就麻烦了）。我想他总共只花了几个小时就做到了，构建了这个时事通讯。

[原文] [Sridhar Ramaswamy]: And uh but funnily enough he was uh to build the entire self-contained working thing that can literally react to any question that he has as if he if he says "Hey I have this hobby and I need you to help me get better at this hobby." It'll start sending him messages every day about what should he do to like learn a new skill The general purpose nature of this is truly truly mind-blowing Took him a few hours to set up Yeah that's the wildness of the moment But funnily enough he's 26 and he was like "Yeah yeah yeah I want no part of this maltbook thing I think it's a bunch of hype I think it's actually people posing as uh you know as agents that are posting this You wanted no part of that

[译文] [Sridhar Ramaswamy]: 有趣的是，他构建了整个独立的、能工作的系统，可以对他提出的任何问题做出反应。比如如果他说：“嘿，我有这个爱好，我需要你帮我在这个爱好上做得更好。”它就会开始每天给他发信息，告诉他应该做什么来学习新技能。这种通用性真的是令人极度震撼。他只花了几个小时就设置好了。是的，这就是当下的疯狂之处。但有趣的是，他 26 岁，他说：“是的是的是的，我不想参与这个 maltbook（注：原文疑似指代某种被炒作的 AI 社交现象或平台）的事情，我认为那是一堆炒作，我认为实际上是人们伪装成代理在发帖。”他不想参与那个。

[原文] [Alex Kantrowitz]: And so it's it's funny I think it's a remarkable moment in terms of you know in terms of what is happening out there Um but I do think that you're seeing what happens as um these agents or you know agent frameworks become easier and easier to use and set up and people will figure out a set of security guard guard rails for how to use that and uh and and things like that This is I think a it's it's a pretty remarkable moment Yeah notebook 175,000 posts 1.1 million comments as of it's the social network for the AI bots as of the time we're speaking so I don't think it's entirely I mean if that's entirely human it's a pretty successful social network on the rise so it's done that in a week pretty interesting

[译文] [Alex Kantrowitz]: 所以这很有趣。我认为就外面发生的事情而言，这是一个非凡的时刻。但我确实认为，你会看到随着这些代理或代理框架变得越来越容易使用和设置，人们会弄清楚一套如何使用它们的安全护栏之类的东西。我认为这是一个相当非凡的时刻。是的，Notebook（注：此处指代前文提到的那个被认为是炒作的平台）在我们说话时已经有 17.5 万个帖子，110 万条评论，它是 AI 机器人的社交网络。所以我不认为它是完全……我的意思是，如果那完全是人类发的，那它就是一个正在崛起的相当成功的社交网络，而且是一周内做到的，非常有趣。

[原文] [Alex Kantrowitz]: uh you made some predictions ahead of the year and one of them really stood out to a couple of them stood out to me we could talk about them both but uh one of them that I found really interesting was you said "Shadow AI will drive enterprise adoption from the bottom up Employees who select their own free AI tools will will remain the primary driver of enterprise AI adoption in 2026 rather than waiting for IT departments to sanction approved products Workers are using chatpt claude and other consumer AI tools for their daily work forcing organizations to catch up I think that's so interesting and it's something that I've talked about on the show before how it seems like there's these two tracks Companies that are kind of slow to move and adopt these tools and individuals that are starting to find ways to use them in their work Why do you think that is

[译文] [Alex Kantrowitz]: 你在年初做了一些预测，其中有一个——实际上有几个——非常引人注目，我们可以都聊聊。但其中一个我觉得特别有趣，你说：“影子 AI（Shadow AI）将自下而上地推动企业采用。选择自己免费 AI 工具的员工将继续成为 2026 年企业 AI 采用的主要驱动力，而不是等待 IT 部门批准产品。工人们正在使用 ChatGPT、Claude 和其他消费者 AI 工具进行日常工作，迫使组织追赶。”我觉得这太有趣了，这也是我之前在节目中谈到过的：似乎有两条轨道，公司行动迟缓，还在逐步采用这些工具；而个人已经开始想方设法在工作中使用它们。你认为这是为什么？

[原文] [Sridhar Ramaswamy]: first of all I mean anyone who's been inside a even moderatelysized company knows that it's filled with approvals and lawyers and um you know uh pilots I have a simpler answer Yes it's the true 10xing of the moment I talked to you about how with something like a cortex code you can get a job that you need to do on Snowflake Like working with data is tough It's tedious Yeah you have to get lots of things right A lot of little details can use our CLI and just automate this stuff and get it done in less than a tenth of the time it would have otherwise have taken you Right that is remarkable

[译文] [Sridhar Ramaswamy]: 首先，任何在哪怕是中等规模公司待过的人都知道，里面充满了审批、律师和试点项目。但我有一个更简单的答案：是的，这是真正的“10 倍速（10xing）”时刻。我跟你谈过，像 Cortex Code 这样的东西如何让你完成在 Snowflake 上需要做的工作。处理数据很难，很枯燥。是的，你必须把很多事情做对。很多小细节现在可以使用我们的 CLI（命令行界面）直接自动化，完成时间不到原本所需的十分之一。对，这非常了不起。

[原文] [Sridhar Ramaswamy]: And uh I now write documents This is with our officially approved enterprise uh version of our chat bots I write position papers coming out of dialogues that I have with these chatbots I say this is the situation These are my thoughts These are the options What do you think you sort of go through almost a Socratic process of debating stuff and producing something that looks mighty polished Mhm But if I've done pricing studies entirely inside chat bots right we have to change prices You trust them because sometimes when I like have them do the numbers Okay I I never I am I have never ever run a coding agent with accept all my recommendations Okay I am as anal as they come Okay Um my first rules when I started using our coding agent was never delete a data Never ever delete a database Never ever switch an account because I have access to production systems that have snowflake data I'm like don't switch to it when I'm playing around with something else

[译文] [Sridhar Ramaswamy]: 我现在写文档——这是用我们官方批准的企业版聊天机器人——我会写立场文件，内容源于我与这些聊天机器人的对话。我说：“这是情况，这是我的想法，这些是选项，你怎么看？”你基本上经历了一个苏格拉底式的辩论过程，并产出了看起来非常完善的东西。嗯。但我甚至完全在聊天机器人里做过定价研究，比如我们需要调整价格。你信任它们吗？因为有时候当我要它们算数时……好吧，我从来没有——从来没有——运行一个编码代理并选择“接受所有推荐”。好吧，我在这种事上非常较真。当我开始使用我们的编码代理时，我的第一条规则是：永远不要删除数据，永远不要删除数据库，永远不要切换账号。因为我有权访问包含 Snowflake 数据的生产系统，我就像在说：“当我在玩别的什么东西时，别切换到那个账号去。”

[原文] [Sridhar Ramaswamy]: You got to put the guardrails You got to be smart about how you work Um and you got to check the work right and so when I did the pricing study it's like hey plot this for me How does revenue and margin change you got to go study the work But it's a massive accelerant and the benefit that you get from something like this unlike a handwritten doc is let's say you decide to change your mind and want to introduce another new thing you know normally we just don't do that in a document or a study because it's so tedious to go make all the changes the chatbot they don't get bored they're like you want to redo this work not a problem they redo the work for you um I think it's that value creation that's driving the adoption

[译文] [Sridhar Ramaswamy]: 你必须设置护栏，你必须聪明地工作。嗯，你必须检查工作成果。所以当我做定价研究时，比如“嘿，帮我画个图，收入和利润率是如何变化的”，你必须去研究它的工作成果。但这是一个巨大的加速器。你从中获得的好处——不像手写文档——是，假设你决定改变主意，想引入另一个新东西。你知道，通常我们在文档或研究中不会那样做，因为要把所有更改都做一遍太枯燥了。但聊天机器人不会感到厌烦。它们就像是：“你想重做这项工作？没问题。”它们为你重做工作。我认为正是这种价值创造在推动采用。

[原文] [Alex Kantrowitz]: how does that change the dynamic of companies if you have a couple of people in there that are like leaning all the way into the tools and the company is like yeah we're in we're working through this uh

[译文] [Alex Kantrowitz]: 如果公司里有几个人正在全力投入使用这些工具，而公司的态度是“好吧，我们在参与，我们正在研究这个”，这会如何改变公司的动态？

[原文] [Sridhar Ramaswamy]: Well part of what every company has to do is to figure out how to embrace these change agents and uh make sure that they're surfacing what they want to do and the value that they're getting to everyone Mhm I wanted to roll out Cortex code to the entirety of our solution engineering team 2,000 people It's a lot of people And uh the way we we did that was um we selected a subset of them over 30 40 people and uh gave them a little bit of training and said hey you should go try this out see what this is like We call them our AI champions Mhm We celebrated the fact that these were the forwardleaning folks and uh we also made them effectively responsible for spreading the word down to the different uh to the different teams change in any large company is not going to come from top- down mandates

[译文] [Sridhar Ramaswamy]: 嗯，每个公司必须做的一部分工作是弄清楚如何拥抱这些变革推动者（change agents），并确保将他们想做的事情以及他们获得的价值展示给每个人。嗯。我想向我们整个解决方案工程团队推广 Cortex Code，那可是 2,000 人，很多人。我们做法是，我们选了一小部分人，大约 30 到 40 人，给了他们一点培训，然后说：“嘿，你们应该去试试这个，看看它怎么样。”我们称他们为我们的“AI 冠军（AI champions）”。嗯。我们庆祝这些前瞻性的人才，我们也让他们有效地负责将这些信息传播到不同的团队。任何大公司的变革都不会来自于自上而下的命令。

[原文] [Sridhar Ramaswamy]: You know let's face it what I know about AI is minuscule compared to the sum totality of what my 9,000 people know about AI And you need to create an environment in which the most progressive of the ideas that are coming up the most innovative of the people they have a way to quickly surface the idea up In fact for the next all hands um I've been working with my comm's team It's in it's in a few weeks Uh they wanted to have you know a regular all hands standard set of discussions with the exec staff I said uh I want to spend 2 minutes personally because I have to say something as a CEO I want the rest of the time to be devoted to finding these fire brands looking at what they do and highlighting this as the champions we need to figure out how to identify and how to learn from And we have to embrace the moment in terms of how do we use our collective wisdom to drive our organizations forward

[译文] [Sridhar Ramaswamy]: 你知道，让我们面对现实吧，我对 AI 的了解与我那 9,000 名员工对 AI 的了解总和相比是微不足道的。你需要创造一个环境，让那些最进步的想法、最具创新精神的人能够快速将想法浮现出来。事实上，对于下一次全员大会——我正在和我的公关团队沟通，就在几周后——他们原本想安排一套常规的全员大会流程，即高管团队的讨论。我说：“我想我个人只花 2 分钟，因为作为 CEO 我必须说点什么。我想把剩下的时间都用来寻找这些‘火炬手（firebrands）’，看看他们在做什么，并把他们作为我们需要识别和学习的冠军来重点推介。”我们必须拥抱这个时刻，思考如何利用我们的集体智慧来推动组织向前发展。

[原文] [Alex Kantrowitz]: It's very interesting because it seems like as these tools get better there are going to be companies that will have that mentality and there'll probably be companies with leaders who are just like I don't know about you know all this AI stuff And it can actually change the competitive balance of industries pretty quickly if you have organizations with more permission versus less

[译文] [Alex Kantrowitz]: 这非常有趣，因为看起来随着这些工具变得更好，将会有一些公司拥有这种心态，但也可能有些公司的领导者会说：“我不知道这些 AI 玩意儿到底怎么回事。”如果你有的组织给予更多许可，而有的给予较少，这实际上可能会很快改变行业的竞争平衡。

[原文] [Sridhar Ramaswamy]: I would I would distinguish it more as progressive organizations Okay What does that I what I mean by that is we always have to balance Um I will flip out if I find out that anyone's running openclaw on a snowflake laptop Please don't do that That's not safe We will help you get like a free Ubuntu machine on AWS if you want Uh there are smart things that people should be doing and dumb things that they should not be doing A progressive head of security is an important asset here where they let the innovation happen Y without making people do unsafe things Um we are custodians of data for some of the most valuable companies in the world and we take that part very very seriously and so it is that balance that uh that one needs But back to your point um about changing competitive dynamics very very very real

[译文] [Sridhar Ramaswamy]: 我会更多地将其区分为“进步型组织（progressive organizations）”。好的。我这么说的意思是，我们总是需要平衡。如果我发现有人在 Snowflake 的笔记本电脑上运行 OpenClaw，我会发飙的。请不要那样做，那不安全。如果你想要，我们会帮你搞一个 AWS 上的免费 Ubuntu 机器。有些是人们应该做的聪明事，有些是不该做的蠢事。一位进步的安全主管在这里是一项重要的资产，他们让创新发生，而不让人们做不安全的事情。我们是世界上一些最有价值公司的数据保管人，我们非常非常认真地对待这一部分。所以这就是我们需要的那种平衡。但回到你关于改变竞争动态的观点，那是极其、极其真实的。

章节 8：开源模型与地缘政治：DeepSeek 效应与全球生态

📝 本节摘要：

在访谈的最后部分，Alex 提到了 Sridhar 关于“科技巨头对 AI 模型掌控力将松动”的预测。Sridhar 指出，DeepSeek 等案例证明了构建高性能模型并非必须依赖昂贵的预算，这促使 Snowflake 放弃自研基础模型，转而在平台上支持多元化的开源模型。Sridhar 深入分析了地缘政治层面的隐忧：由于 OpenAI 和 Google 等美国公司将技术“围墙化（walled off）”，而中国公司积极公开发表研究，导致全球学术界正基于中国技术进行构建。但他对此持乐观态度，认为这种“存在性证明（knowing something is possible）”将刺激西方开源力量（如 Mistral、Meta 和 Reflection AI）加速反向工程与创新，这对整个生态系统最终是净利好。

[原文] [Alex Kantrowitz]: Uh I think we can end here You also have this interesting prediction about big tech big tech's grip on AI models loosening I'll just read a little bit of it For years conventional wisdom held that only a handful of tech giants could afford to build competitive AI models In 2026 that will change New approaches to training like those developed by Deep Seek have shown that building the biggest most expensive models isn't the only path to strong performance You know we're a year this is great timing We're a year after DeepSeek didn't fully change the AI industry in a way a lot of people anticipated And so it's interesting to see that that is the prediction you made especially if I'm if because if I'm right Snowflake did try to build some foundational models and then decided that was not the game you wanted to play

[译文] [Alex Kantrowitz]: 嗯，我想我们可以到此结束了。你还有一个有趣的预测，关于科技巨头对 AI 模型的掌控力正在松动。我读一点其中的内容：“多年来，传统智慧认为只有少数科技巨头负担得起构建具有竞争力的 AI 模型。在 2026 年，这将会改变。像 DeepSeek 开发的那些新训练方法表明，构建最大、最昂贵的模型并不是通往高性能的唯一路径。”你知道，我们现在的时间点很棒。距离 DeepSeek 并没有像很多人预期的那样完全改变 AI 行业已经过去一年了。所以很有趣看到你做出了这个预测，尤其是如果我没记错的话，Snowflake 确实尝试过构建一些基础模型，然后决定那不是你们想玩的游戏。

[原文] [Sridhar Ramaswamy]: I think foundation models became very expensive to build We now have four players that are creating models that are like widely acknowledged to be the state-of-the-art Um but a new Quen model came out yesterday that is shockingly close to the best sonnet model that there is from uh Anthropic There continues to be a lot of innovation in this space I think that's very very healthy for us and from a selfish perspective Snowflake as a data data platform prefers a world in which there are many people making great models especially open source models because we also have a really good infrastructure team We are very good at running them at uh at at scale But um this is a world where a lot of value is being created and a lot of change is happening And I think being nimble and ready for that future of agent KI that future of work while always having a laser focus on what makes a difference to your customer Those are the enduring qualities through the year Life will keep changing

[译文] [Sridhar Ramaswamy]: 我认为基础模型的构建变得非常昂贵。我们现在有四个玩家正在创造被广泛认为是最先进（state-of-the-art）的模型。但是，昨天发布了一个新的 Qwen（通义千问）模型，它令人震惊地接近 Anthropic 最好的 Sonnet 模型。这个领域持续存在大量的创新。我认为这对我们非常非常健康。从自私的角度来看，Snowflake 作为一个数据平台，更喜欢一个有很多人制造伟大模型——尤其是开源模型——的世界，因为我们也有一个非常优秀的各种基础设施团队。我们在大规模运行这些模型方面非常在行。但这还是一个正在创造大量价值、发生大量变化的世界。我认为保持敏捷，为代理 AI（Agent AI）的未来、工作的未来做好准备，同时始终聚焦于对客户有影响的事情，这些是贯穿全年的持久品质。生活将不断变化。

[原文] [Alex Kantrowitz]: You're comfortable with the Chinese open source models

[译文] [Alex Kantrowitz]: 你对使用中国的开源模型感到放心吗？

[原文] [Sridhar Ramaswamy]: So we test them we use them we try to um we we try to learn uh from them Uh we also partner with uh US companies that are trying to create open source models There's actually a company that's based in uh in in Brooklyn uh and San Francisco that is um uh that that that we work with Which one um this if I remember correct this is Reflection AI Okay And uh it's a remarkable company I think uh there is a lot that we are missing out in not having a robust open AI ecosystem We uh sometimes get caught up in this world of uh you know we have the best AI companies on the planet But we also should understand that much of their work has effectively become walled off from the rest of the world You and I simply do not know what techniques open AI and anthropic are adopting to produce the great models

[译文] [Sridhar Ramaswamy]: 我们测试它们，使用它们，我们试图……我们试图从中学到东西。我们也与试图创建开源模型的美国公司合作。实际上有一家位于布鲁克林和旧金山的公司，我们正在与他们合作。哪一家？如果我没记错的话，是 Reflection AI。好的。这是一家非凡的公司。我认为，如果没有一个强大的开放 AI 生态系统，我们会失去很多东西。我们有时会沉浸在这个“我们拥有星球上最好的 AI 公司”的世界里。但我们也应该明白，他们的很多工作实际上已经与世界其他地方隔绝（walled off）了。你和我根本不知道 OpenAI 和 Anthropic 正在采用什么技术来生产这些伟大的模型。

[原文] [Sridhar Ramaswamy]: You can say how does it matter google search for example pretty much died as an academic uh as an academic area after Google became big Why they published nothing and they were ahead of everyone else by a million miles Area just died M and that was okay for us geopolitically because Google was an American company I think part of what you're reacting to is this fear now of open source is not here but much more in in a situation where there is no winner What is happening right now is that it's the Chinese companies that are publishing their work And what then happens is all the universities all the students and professors in our country are looking at their work and figuring out how to build on top of it And so academia is diverging from what's happening in the research labs That's part of the danger of this moment And that's the reason why we need to have a more robust ecosystem If it been if it had been a world in which there was one model maker that was a winner and there was an American company I think we'd have a slightly different attitude It's very clear now that that's not going to happen Hence the fear about about open models

[译文] [Sridhar Ramaswamy]: 你可能会说这有什么关系？以 Google 搜索为例，在 Google 变大之后，搜索作为一个学术领域基本上就死掉了。为什么？因为他们什么都不发表，而且他们比其他所有人都领先一百万英里。这个领域就这样死掉了。嗯。在地缘政治上这对我们来说没问题，因为 Google 是一家美国公司。我认为你现在反应的部分恐惧在于，开源并不在这里（美国），而更多是处于一种没有赢家的局面。现在正在发生的是，是中国公司在发表他们的工作。随之发生的是，所有的大学、我们国家所有的学生和教授都在看他们的工作，并弄清楚如何在他们的基础上进行构建。因此，学术界正在与（美国私有）研究实验室里发生的事情分道扬镳。这就是当下危险的一部分。这也是为什么我们需要一个更强大的生态系统。如果这是一个只有一个模型制造商赢家且是一家美国公司的世界，我想我们的态度会略有不同。现在很清楚这不会发生。因此产生了对开放模型的恐惧。

[原文] [Alex Kantrowitz]: And then if these you know I think there's been such so much conversation about the Chinese open models over the past couple weeks Um you know I think Demisab has said uh at the crack of the new year that the US or the west is four years ahead sorry four months ahead of them Recently there's been some uh discussion that it's kind of you know closer than that uh is what happens in the world where like those models become on par with the leading US foundational models for most of us

[译文] [Alex Kantrowitz]: 然后如果这些……你知道，过去几周关于中国开放模型的讨论太多了。我想 Demis Hassabis（DeepMind CEO）在新年伊始说过，美国或西方领先他们四年——抱歉，是四个月。最近有一些讨论认为差距其实比那更小。如果对于我们要大多数人来说，这些模型变得与美国领先的基础模型并驾齐驱，世界会发生什么？

[原文] [Sridhar Ramaswamy]: Yes it opens up lots of opportunity The as you know the very existence of something knowledge about the existence of something can spur innovation in other areas You don't even have to know exactly what someone did This history has shown this repeatedly Just knowing that something is possible makes people work feverishly on making the same thing happen You can bet that reflection is looking at it and going we can do better than this right so from a macro perspective I would say that that is actually a positive because Mistrol is going to figure out how to reverse engineer all of this stuff stuff and go one step forward which will be good for Europe and Reflection will figure out how to do this in the US This will also force Meta to be doing more things in the um in the US I think in a weird way that's actually a net positive for us as a whole I think the impact on the model companies that becomes a little bit more little bit more murky but welcome to welcome to this world Alex you know this change every month it's constant

[译文] [Sridhar Ramaswamy]: 是的，这开启了很多机会。如你所知，某事物的存在——关于某事物存在的知识本身——就能刺激其他领域的创新。你甚至不需要确切知道别人是怎么做的。历史已经反复证明了这一点。仅仅知道某事是可能的，就会让人们狂热地工作以实现同样的事情。你可以打赌 Reflection AI 正在看着它说：“我们可以做得比这更好。”对吧？所以从宏观角度来看，我想说这实际上是一个积极因素，因为 Mistral 会弄清楚如何反向工程所有这些东西并更进一步，这对欧洲有好处；Reflection 会弄清楚如何在美国做这件事。这也会迫使 Meta 在美国做更多的事情。我认为以一种奇怪的方式，这对我们要整体来说实际上是一个净利好（net positive）。我认为这对于模型公司的影响可能会变得有点模糊不清，但欢迎来到这个世界，Alex。你知道，这里每个月都在变，这是常态。

[原文] [Alex Kantrowitz]: the website is snowflake.com streetar so great to see you thank you for coming down thank you Alex always a great conversation definitely really is we hope hope we can do this again soon thank you all everybody thank you for listening and watching and we'll see you next time on Big Technology podcast

[译文] [Alex Kantrowitz]: 网址是 snowflake.com。Sridhar，很高兴见到你，谢谢你过来。谢谢，Alex，总是很棒的对话。确实是。希望我们很快能再聊。谢谢大家，感谢大家的收听和观看，我们下期 Big Technology Podcast 再见。