谢尔盖·布林（谷歌与Alphabet）- 谷歌起源与人工智能未来

是的，我的意思是，如果你一个月不看AI新闻，你就完全落伍了。

And it's yeah, I mean, like, you know, if you skip the news and AI for a month, you're, like, way behind.

就像，你知道的。

Like, you know.

那么它会发展到什么程度呢？

So where is it going to go?

我是说，我不知道...我想我们真的无从知晓。

I mean, I don't you know, I I think we just don't know.

智能是否存在上限？

Is there a ceiling to intelligence?

我想除了你提出的问题——比如它能否做到人类能做的任何事情之外，

I guess in addition to the question that you raised, like, can it do anything a person can do?

还存在另一个问题：它能做到哪些人类无法做到的事情？

There's the question, like, what things can it do that a person cannot do?

是啊。

Yeah.

这某种程度上是个超级智能的问题。

That's sort of a superintelligence question.

是啊。

Yeah.

而且我认为这还不得而知。

And I think that's just not known.

比如，一个事物能有多聪明？

Like, how smart can a thing be?

要知道，人类已经经历了数十万年的进化历程，还有数百万年的灵长类演化史。

You know, we've had however many hundreds of thousands of years of human evolution and, I don't know, whatever, millions of primate.

但相比AI的发展速度，那是个相当缓慢的过程。

But that's a pretty slow process, compared to what's going on with AI.

你认为我们准备好应对技术如此迅猛的发展速度了吗？

Do you think we're ready for the speed at which the technology is advancing?

我们真的准备好迎接技术发展的速度了吗？

Are we ready for the speed the technology is advancing?

我是说，看看目前的情况，我认为人们确实从技术中获益良多。

I mean, look, so far, I think people are getting definitely great use out of technology.

尽管各处都有悲观预测，但我觉得大家其实都获得了相当强的赋能。

I think even though there are doom and gloom forecasts here and there, like everybody's pretty well empowered.

说实话，AI时不时会犯些低级错误，所以你总得在旁边监督着它们。

And the AIs, truth be told, are periodically dumb enough that you're always, like, supervising them anyway.

但偶尔它们也会灵光一现，给你绝妙的主意。

But occasionally, they're brilliant and give you a great idea.

特别是对于非专业人士来说，偶尔...

And occasionally, especially as a non expert.

嗯，怎么说呢...

Well, like, whatever.

如果我想了解如何设计新的AI芯片，我猜可以咨询我们的专家设计师之类的。

If I wanna figure out how to create a new AI chip, I guess I could talk to our expert designers and stuff.

但最基础的情况是，我至少能掏出手机，和AI讨论这个问题。

But as a base case, I can at least I can whip out my phone, I can talk to an AI about it.

它可能会给我一个90%还算不错的概述，让我理解健康问题或其他什么。

It'll probably give me a 90 90% decent sort of overview and understand it or whatever, my health questions or whatnot.

我是说，我确实认为这让个人变得非常有能力，因为一般来说，你身边不可能随时都有各种领域的专家。

I mean, I I do think it makes individuals very empowered, because generally speaking, you don't have, like, experts in x y z all around you all the time.

我认为这种赋能能创造很多潜力，无论是在职业、企业、健康还是生活品质方面。

And I think that empowerment can create a lot of potential, whether it's career or enterprise or health or living well.

所以你看，我并不认为自己掌握了所有答案。

So look, I don't think I have all the answers.

我确实认为它在提升个人能力方面具有巨大潜力。

I just do think it has a huge potential to improve individual capability.

是的，这无疑是积极的愿景，它可能成为人类能力的强大增强器。

Yeah, that's certainly the positive vision, that it could be an incredible augmenter of human capability.

你能这样思考真的很棒。

It's great that you're thinking about it that way.

让我提个问题，这个问题在创业思想领袖课程中经常被问到，但在讨论AI时显得尤为贴切——因为斯坦福的每位学生，或许全国所有大学生都在思考：这项技术将如何影响他们的职业发展、就业机会以及未来的事业方向？

Let me ask a question, I think, which is always asked in the entrepreneurial thought leaders class, but is maybe particularly sailing with the discussion of AI because one of the things I think every student at Stanford and probably every college age student in the country is thinking about is, how will this technology affect their careers and their job opportunities and what they might go on to do?

我很好奇你对学生们有什么建议，关于他们未来面对就业市场时应该学习或思考些什么。

I'm curious if you have any advice for the students about what they ought to be studying or what they ought to be thinking about as they look at the job market in the future.

我是说，我认为要准确预测会发生什么非常困难。

I mean, I think it's super hard to predict exactly what will happen.

如果我们从互联网诞生到手机等技术的发展来看，这些已经深刻改变了我们的社会，无疑也改变了人们从事的工作、职业和学习方向。

I think if we look at from the advent of the web to cell phones and so forth, those have transformed our society profoundly, have transformed the kinds of jobs and careers and studies people do, for sure.

而人工智能100%会改变这一切。

And AI will 100% change that.

但我认为在快速变化的形势下，现在很难确切说明会改变什么。

But I think it's very hard right now, in a rapidly shifting landscape to say exactly what.

而且我们现在拥有的人工智能与五年前大不相同，与五年后也将完全不同。

And also the AI we have today is very different from the AI that we, well, had five years ago or the AI we are going to have in five years.

所以，是的，我也不知道。

So, yeah, I don't know.

我认为要真正预测确实很困难。

I think it's tough to really forecast.

我是说，我一定会利用AI来为你谋利。

I mean, I would for sure use AI to your benefit.

你能做的事情实在太多了。

There are just so many things that you can do.

对，就拿我个人来说，无论是为朋友或家人挑选礼物，还是为产品或艺术之类的事情头脑风暴新点子。

Yeah, I mean, just myself as an individual, whether it's choosing a gift for my friends or family or brainstorming new ideas for products or what have you or for art or something like that.

就像，我现在总是求助于AI，虽然它不会完全代劳——因为我通常会要求'给我五个点子'之类的。

Like, I just turn to AI all the time now, and it it doesn't do it for me because I always you know, I typically will ask, give me five ideas, blah blah blah.

你知道，可能其中三个点子某种程度上都是垃圾，但我能分辨出来。

And, you know, probably three of them are gonna be junk in some way, but I'll just be able to tell.

但会有两个点子蕴含闪光点，或者能帮我打开思路，让我可以完善和深入思考。

But two will have some grain of brilliance and, you know, or possibly put it in perspective for me or something like that that I'll be able to refine and think through my ideas.

让我插一个非常具体的问题。

Let me jump in with a really concrete question.

我们现场大约有250名学生。

So we have about two fifty students out there.

他们中很多都是本科生。

A lot of them are undergraduates.

其中相当一部分人尚未选定专业，因为斯坦福大学为本科生提供了很大的灵活性。

A great number of them have not selected their major yet because we give them a lot of flexibility here at Stanford for the undergraduates as well.

几年前，我们还能预测会有大批学生选择计算机科学作为专业。

A few years ago, we could predict that a large number would choose computer science as their major.

你是否建议他们继续选择计算机科学作为专业？

Are you recommending they continue to pick computer science as their major?

他们正在认真聆听。

They're listening closely.

我选择计算机科学是因为对它充满热情。

I mean, I chose computer science because I had a passion for it.

所以这对我来说是个无需思考的决定。

So it was kind of a no brainer for me.

我想可以说我也很幸运，因为这个领域当时正处于变革时期。

I guess you could say I was also lucky because I was also such a transformative field.

我不会仅仅因为现在AI在编程方面表现不错，就放弃选择计算机科学。

I wouldn't, like, not choose computer science just because, you know, AI can be decent at coding nowadays.

AI在很多方面都表现得相当不错。

AI is pretty decent at a lot of things.

编程恰好具有很高的市场价值，这就是为什么很多人选择这个领域。

Coding just happens to have, like, a lot of market value, which is why a lot of people pursue it.

而且，更好的编程能催生更好的AI。

And furthermore, you know, better coding makes for better AI.

因此许多像我们这样的公司都非常重视这一点。

So a lot of the companies like our own that work on it care a lot about it.

比如，我们在自己的编程工作中大量使用它，甚至用于算法构思等方面。

Like, we use it a lot for our own coding, and even for our algorithmic ideas and so forth.

但这正是因为编程是如此重要。

But that's because it's such an important thing.

所以我想我不会因为觉得AI擅长编程，就跑去转修比较文学之类的专业。

So I guess I wouldn't, I wouldn't go off and, like, switch to comparative literature because you think the AI is good at coding.

说实话，AI在比较文学方面可能更出色。

The AI is probably even better at comparative literature, just to be perfectly honest.

总之，是的。

Anyway Yeah.

我并非不尊重比较文学专业的学生，但要知道，当AI编写代码时，老实说有时会出错，犯下相当严重的错误。

Like, I don't mean to disrespect comparative literature majors, but just like, you know, when you when the AI writes a code, and just to be honest, sometimes it doesn't work, like, it'll make a mistake that's pretty significant.

比如，在比较文学论文中写错一个句子，不会造成那么严重的后果。

Like, you know, getting a sentence wrong in your essay about comparative literature isn't gonna, like, really have that consequence.

所以坦白说，AI更容易完成某些创造性工作。

So it's honestly easier for AI to do some of the creative things.

我认为这是个关于技术非常有趣的观察，因为人们通常倾向于认为AI擅长解决技术问题，但未必能完成我们认为是人类特质的事情，比如在对话中展现同理心。

I think it's a very interesting observation about the technology, because I think one inclination to say about AI is it's going to be really good at solving these technical problems, but it won't necessarily do the things we associate with humans, like being empathetic in a conversation.

如果你让这些AI引擎模拟对话，它实际上非常擅长为你构建复杂对话的框架。

And if you ask one of these AI engines to, say, simulate a conversation, it's actually pretty good at doing a lot of giving you the structure for a complicated conversation.

我很欣赏你指出了这种不确定性。

I like that you're pointing to that uncertainty.

我想再提一个问题，向观众开放，给大家一个提问的机会。

One more question that I want to open up to the audience so that we give people a chance to ask questions.

今年是工程学院成立一百周年。

So this is the one hundredth anniversary of the School of Engineering.

如果你是珍妮弗，需要为学院的第二个世纪规划蓝图，你会为工程学院的第二个百年考虑些什么？

If you were Jennifer and had to launch the school's second century, what would you be thinking about for the second century of the School of Engineering?

哇，好吧，这个规划责任重大啊。

Wow, okay, that's a big responsibility to kind of plan.

确实责任重大。

It is a big responsibility.

我的意思是，我可能会重新思考大学存在的意义。

I mean, I I guess I just would rethink what it means to have a university.

我知道这话听起来有点烦人。

I mean, I know that sounds kind of annoying.

这种话要是拉里说的，我肯定会对他很恼火。

That's the kind of thing Larry would say, and I would be really annoyed with him.

但是，你知道，我们拥有这种地理上集中的实体，比如那些建筑和豪华的讲堂。

But, you know, I mean, we have this geographically concentrated thing and there's like the buildings and the fancy lecture halls.

那个特别烦人的闪烁灯光。

That really annoying blinking light.

抱歉。

Sorry.

我忍不住要说。

I can't help it.

你们真的需要解决这个问题，不过现实点说，现在信息传播得非常快。

You guys need to you guys need to fix your but, but but realistically now, you know, information spreads very quickly.

许多大学，包括斯坦福在内，显然都已经转向了线上教学。

And many of the universities have, you know, obviously, whatever, gone online, including Stanford.

但MIT很早就推出了OpenCourseWare，还有各种初创公司如Coursera、Udacity等等都走上了这条路。

But, you know, MIT with the OpenCourseWare early on and all these startups that have gone this with whatever Coursera, Udacity, you name it.

所以教学资源正在扩散，现在任何人都可以上网学习这些内容。

So the teaching is sort of getting spread, and and anybody can go online now and learn about it.

你知道，你可以和AI对话，或者参加这些课程，观看YouTube视频。

You know, you can talk to an AI or take one of these classes and watch the YouTube videos.

所以我想是的。

So I guess I yeah.

拥有一所大学意味着什么？

What does it mean to have a university?

你是想最大化影响力吗？

Are you trying to maximize the impact?

那样的话，你知道，仅仅在地理上限制可能效果不会太好。

In that case, you know, probably just limiting it geographically is not gonna be so effective.

公平地说，我想湾区是个特别的地方。

To be fair, I guess, you know, in the Bay Area is kind of a special place.

不过，确实。

But, yeah.

我是说，我知道我有点东拉西扯在思考这个问题，但我觉得下个世纪工程学院和大学的概念可能不会和过去一样了。

I mean, I I know I'm kind of rambling here and thinking through, but, yeah, I just thought I don't know that for the coming century that, you know, the idea of school of engineering and the university is gonna mean the same thing as it as it used to.

我是说，人们四处流动，远程工作，跨地域协作。

I mean, people move around, work remotely, collaborate across.

这有点矛盾，因为我们正努力让人们实际回到办公室，我认为他们面对面工作时确实效率更高。

It's a little bit at odds because we're trying to get people actually into the office, and I think they do work better in person together.

但这只是在特定规模下成立。

But But that's at a certain particular scale.

比如，某种程度上，如果那边有100人聚在一起，其实也没什么问题。

Like, at some level, if you have 100 people together over there, it's kind of fine.

他们不必和其他100个人待在同一个地方。

They don't have to be at the same place as these other 100 people.

越来越多地看到，那些创造新事物的人往往与学历无关。

Increasingly, do see sort of individuals kind of who create new things sort of regardless of degree.

虽然我们聘用了很多学术明星，但也雇佣了大量没有学士学位之类的人，他们就是靠自己摸索，在某些奇怪的角落里把事情搞明白。

I mean, as much as we've hired a lot of academic stars, we've hired tons of people who don't have bachelor's degrees or anything like that, and they just figure things out, you know, on their own in some weird corner.

我不知道。

I don't know.

我觉得这是个非常难回答的问题。

I I think it's really hard question.

是啊。

Yeah.

我想我无法神奇地给你一个全新方案，但我确实认为当前这种模式不太可能延续到下个世纪。

I guess I don't feel like I'm gonna magically deliver you like the new recipe, but I I I just don't think this format is likely to be the one for the next hundred years.

你把这个话题引向了比我预期更深刻的层面，这很棒。

You took that in a deeper direction than I was that was great.

实际上确实更有深度

Actually, it was a bit deeper

听起来更像总统而非院长在发言。

Sounded more presidential than dean like.

我觉得他是在对你说话。

I think he's he's talking to you.

不过...我同意你的观点。

But you I I agree.

这适用于整个大学。

It applies to the whole university.

你实际上提出了关于大学最根本的问题，即大学的一部分关乎知识的创造与传承。

You actually surfaced the the most fundamental questions about the university, which is that that the you know, part of the university is about the the creation and transmission of knowledge.

这是其根本使命。

That's the fundamental mission.

随着技术进步，这些可以通过不同方式实现。

Those can be done in different ways as technology advances.

然后是关于将大量人才集中在一个地方相互碰撞的模式问题——这种模式当然催生了谷歌的诞生，并带来了许多伟大成果。

And then there's a question about the model of having kind of a density of talent all in one place bumping into each other, which, of course, was what led to you creating Google and has led to a lot of great things.

未来是否会出现替代大学校园这种生态系统的方案？

Will there be substitutes for that kind of ecosystem that gets created on a university campus?

或者说这种模式的根本性如何？它是否会持续存在？

Or how fundamental is that and will it continue to be?

我真的很感谢你在本次讨论中提出如此深刻的问题。

I actually thought that was I appreciate that you surfaced such a deep question in this session.

好的，我想确保我们给观众席上的其他人留些提问机会。

All right, I want to make sure we give some questions for other folks out in the audience.

那么詹妮弗，接下来请你来收集现场观众的一些问题。

So Jennifer, going turn to you to take some questions from the folks out here.

是的，创业思想领袖课程的学生们提前提交了问题，其中一部分已被选中。

Yes, so the students in the entrepreneurial thought leaders class submitted questions in advance and a number of those were selected.

所以在剩余的时间里，我们将回答几位学生的问题。

And so with the time we have left, we're going to have a few questions from our students.

我想第一个问题在这边。

I think the first one is over over here.

威道院长、莱文校长和谢尔盖·布林，感谢你们的时间。

Dean Widow, President Levin, and Sergei Brin, thank you for your time.

我是来自堪萨斯城的拉莎·巴夫，主修材料科学与工程及国际关系。

My name is Rasha Barve from Kansas City studying MS and E and IR.

我的第一个问题要问谢尔盖。

And my first question goes out to Sergei.

这实际上正好触及了我们刚才讨论的内容。

It actually just touches on what we were discussing.

谷歌很大程度上源自您关于PageRank的学术研究成果。

Google largely grew out of the academic work you authored on PageRank.

鉴于如今产业界驱动着如此多的创新，您是否仍认为学术界向产业界的输送管道至关重要？

And with industry now driving so much of today's innovation, do you still feel that the academia to industry pipeline is crucial?

如果是的话，您会如何加强这一管道？

And if so, how might you strengthen it?

嗯，这是个很好的问题。

Well, it's a great question.

学术界向产业界的输送管道是否至关重要？

Is the academia to industry pipeline crucial?

是的。

Yeah.

关于这个问题，我打算给你一个'我不知道'的答案。

I'm gonna give you an I don't know on that.

因为我想，你知道，当我还是研究生的时候，从一个新想法到它可能具有商业价值往往需要几十年时间。

Because I guess, you know, when I was a grad student, the sort of time from some new idea to it being maybe commercially valuable was many decades.

如果这个周期缩短了，那么这样做就不再那么有意义了。

If that compresses, then that no longer makes as much sense to do that.

我的意思是，在学术界，你有自由去长时间思考问题。

I mean, in academia, you have freedom to think about it for a while.

你可以申请资助，做这做那，基本上能花上几十年时间思考，然后慢慢沉淀。

You, whatever, you apply for grants, you do this and that, and you can kind of spend a couple decades thinking about it, and then it percolates.

然后，你知道的，最终可能会有某家大公司或你的初创企业去追求这个方向。

And then, you know, eventually, maybe there's some big company or your startup kind of pursues it.

问题是，如果这个时间线大幅缩短，那还合理吗？

The question is, does that make sense if that timeline shrinks a lot?

我认为确实有些方向是绝对合理的，而且我肯定，即使是人工智能领域，你也知道，斯坦福和其他大学的研究成果会周期性地涌现。

I think there are think there are certain things that for sure make sense and I definitely, you know, even with an AI, you know, periodically you pop with the Stanford research and other universities.

偶尔我们会雇佣那些人并与他们合作，诸如此类。

Occasionally, we, whatever, hire those folks and collaborate with them and whatnot.

但我不确定他们是否真的需要那么长的尝试期，比如试验某种新的注意力机制，花上几年时间实验，然后以某种形式将其引入工业界。

But I guess I don't know that they needed to have that sort of period of time that they were trying, whatever, some new attention thing, let's say, and they spent a couple years experimenting and then, you know, they took it to industry in one form or another.

显然，工业界也在做所有这些事情。

I mean, obviously, industry is also doing all those things.

所以这个论点可能不太站得住脚。

So probably not a huge argument for that.

那些彻底的新架构之类的东西呢？

Radical sort of new architectures and things?

也许吧。

Maybe.

但工业界将其规模化的速度会快得多。

But it's sort of, you know, the time that industry will scale it will be much faster.

量子计算倒是让我想到了这个。

I guess, you know, quantum computing comes to mind.

这个概念最早被提出时——不知道具体时间，好像是费曼在八十年代左右提出的量子计算构想。

There was sort of first brainstormed I don't know when did Feynman, like, in the eighties or something kind of postulate this idea of quantum computing.

现在有一大批公司都参与其中。

And now there are a bunch of companies that are are included.

他们某种程度上正在实践。

They're sort of doing it.

还有一些大学实验室尝试新的实现方法。

There are also university labs that try, like, new ways to do it.

这种情况可能有点难以定论。

That's kind of, like, on the fence maybe.

我想说的是，如果你有全新的想法——比如不是像我们这样做超导量子比特，也不是像许多初创公司那样做离子阱，而是有全新方案的话。

I I guess I would say if you have some completely new idea, like, you're not doing superconducting qubits like we are or whatever the trapped ions, like, a bunch of startups are, but you have some new way.

或许需要让它在大学里酝酿几年，这些事情确实挺难的。

Maybe you need to let it marinate in university for some number of years, those things are kinda hard.

这可能是有道理的。

It could make sense.

但到了某个阶段，如果你确信这个方向确实有前景，很可能就会以某种方式推进商业化。

But then at some point, if you decide it's really compelling, you're probably gonna go ahead and take it commercial in one way or another.

是的。

Yeah.

我不知道。

I don't know.

我想给你一个明确的答案，因为现在顶尖公司确实在投资更多基础研究。

I wanna give you a clear cut answer because, the top companies now do invest in much more fundamental research.

我认为随着AI开始见效，这些投资正在获得回报。

And I think it's sort of with AI starting to pay off that those investments are paying off.

所以我猜这会改变你从事项目的比例分配。

So I guess I I guess it would shift the proportion of endeavors that you would do.

但我确实认为仍有些事情需要十年左右更纯粹的研究，公司可能会更不愿意追求，因为产品化周期实在太长了。

But I I do think there are still some things that do that do take, you know, like the decade of kind of more pure research, that maybe companies are gonna be more reluctant to pursue because that's just too long a time to market.

好的。

Alright.

下一个问题，我想是在这边。

Next question, I think, is over here.

大家好。

Hi, everyone.

我叫Arnav，是一名计算机科学与数学专业的大一学生。

My name is Arnav, and I'm a freshman studying computer science and math.

我的问题是提给谢尔盖·布林的。

My question is for Sergey Brin.

随着AI以前所未有的速度发展，像我这样年轻的创业者应该采取怎样的思维方式，才能避免重蹈覆辙？

As AI accelerates at this unprecedented rate, what mindset should young aspiring entrepreneurs like myself adopt to avoid repeating earlier mistakes?

应该采取怎样的思维方式才能避免重蹈覆辙？

What mindset should you adopt to avoid repeating earlier mistakes?

是的。

Yeah.

当你有了很酷的可穿戴设备创意时，一定要在搞出跳伞飞艇之类的噱头前彻底完善它。

When you have, like, your cool new wearable device idea, really fully bake it before you have a cool stunt involving skydiving in airships.

只是想逗逗你。

Just want to bug you.

不。

No.

其实我挺喜欢我们当年为Google Glass做的那种尝试。

I actually I like the sort of what we were doing back in the day for Google Glass.

这就像，你知道，一个之前犯错的例子。

It's like, you know, an example of prior mistakes.

但我觉得我当时试图过早将其商业化，在我们还没能把它做得更符合成本效益、从消费者角度看更完善之前。

But I think I tried to commercialize it too quickly before, you know, we could make it more, you know, as cost effectively as we need to and as polished as we need to from a consumer standpoint and so forth.

我有点，你知道，操之过急了。

I sort of, you know, jumped the gun.

我当时还想着，哦，我就是下一个史蒂夫·乔布斯。

And I thought, oh, I'm the next Steve Jobs.

我能搞定这个东西。

I can make this thing.

嗒哒。

Ta da.

这可能就是其中之一。

That's probably one.

我想是的，如果让我总结的话。

I guess, yeah, if I encapsulate it.

是的。

Yeah.

每个人都以为自己是下一个史蒂夫·乔布斯。

Everybody thinks they're the next Steve Jobs.

我确实犯过这个错误，但你知道，他是个相当独特的人。

I've definitely made that mistake, but, you know, he was a pretty unique kind of guy.

所以没错。

So yeah.

我想说的是，确保你的想法已经酝酿得足够久，开发得足够深入，达到足够成熟的阶段，然后再进入那个预期不断攀升、开支不断增加、必须在特定时间交付的轨道。

I I I guess I would say, you know, make sure you've baked your idea long enough and developed it sort of long enough to a far enough point before there's sort of a treadmill you get onto where you're sort of outside expectations increase, the expenses increase, and you're sort of then you kinda have to deliver by a certain time.

你可能无法在那段时间内完成所有需要做的事情。

And you might not have that you know, you might not be able to do everything you need to do in that amount of time.

你可能会遇到这种期望的雪球效应，我想，它不断累积，而你却没有给自己足够的时间去处理它们。

You kinda get this snowball, I guess, of expectations that happens and you don't give yourself all the time that you need to process them.

这就是我本想要避免的错误。

That's the mistake I would have tried to avoid.

所有

All

好的。

right.

我想我们要转到这边来。

I think we're gonna go over to this side.

你好，感谢你的演讲。

Hi, thank you for the talk.

这个问题是——我叫伊沙姆·巴尔卡塔克，是斯坦福大学的大一新生。

This question is So my name is Isham Barkatake, I'm an undergraduate freshman at Stanford University.

这个问题是给谢尔盖·布林和詹妮弗的。

This question is for Sergey Brin and Jennifer.

我们看到许多AI公司通过扩展数据和计算能力来改进大型语言模型。

So we see a lot of AI companies improving large language models via scaling data and scaling compute.

我的问题是，一旦我们耗尽数据和计算资源，您认为下一个发展方向会是什么？

My question is, once we do run out of data and once we do run out of compute, what do you think will be the next direction?

会是采用新架构吗？比如Transformer的替代方案？

Would it be in newer architecture, something an alternative to transformers?

还是更好的学习方法？

Or would it be a better learning method?

比我们现在用来训练这些大型语言模型的监督学习或强化学习更好的方法？

Something better than like supervised learning or RL that we use to train these large language models?

或者您之前考虑过的完全不同的方向？

Or is it a completely different direction that you have thought of before?

谢谢。

Thank you.

是的。

Yeah.

可以从我的角度来看这个问题。

Can take it from my point of view.

我是说，你列举的所有这些因素，我认为其实早已比扩大计算规模和增加数据量更为重要。

I mean, all of the things that you listed, I would say, have already been bigger factors than scaling computers, scaling data.

我想这正是人们注意到规模扩张的原因——因为你们在建设数据中心和购买芯片。

I think that's sort of why people notice the scaling because you're building data centers and buying chips.

确实，OpenAI和Anthropic都发表过关于不同类型规模定律的论文。

Well, there were all the publications from OpenAI and Anthropic about, like, different kinds of scaling laws.

所以我认为这吸引了大量关注。

So I think that attracts a lot of attention.

但如果你仔细梳理就会发现，实际上过去十年左右算法进步的速度甚至超过了规模扩张。

But I think if you carefully line things up, you'll see that actually the algorithmic progress has outpaced even the scaling over the last decade or something.

其实早在研究生时期，我就见过类似N体问题的分析图表。

At some point, actually, while ago, many while I was in grad school, I think I saw this kind of plot for, like, the n body problem.

你知道的，就像存在引力作用时物体四处飞散的情况。

You know, like, you have gravitation, they're all flying around.

实际上，自从上世纪五十年代人们开始担忧这个问题以来，直到我读到九十年代资料时，计算能力的摩尔定律增长已经非常巨大。

And it actually there's been huge Moore's law increase in compute over since people started worrying about that in the fifties to I don't know, by the time I read about nineties.

但实际上，解决N体问题的算法发展速度远超计算能力的提升。

But actually, the algorithms to do the n body problem far outpaced, that compute scale up.

所以我认为你会发现，像我们这样的公司永远不会拒绝站在计算技术的最前沿。

So I think you're gonna find that, you know, companies like ours are never gonna turn down being at the frontier, of compute.

不过确实如此。

But that's yeah.

这就像是主菜（算法工作）之后的甜点，计算能力只是锦上添花。

That's just sort of an that's the dessert after, you know, your main course in the veggies of actually having done your algorithmic work.

我想插句话，关于计算资源耗尽或数据耗尽的问题，特别是计算资源方面，我们这里已经非常熟悉这种情况了。

I I guess I'll jump in and say that in terms of running out of compute or running out of data or specifically running out of compute, we're very familiar with that here already.

事实上，大学很难拥有像企业那样的计算资源，这是个现实问题。

It's actually an issue that it's difficult for a university to have the type of compute that the companies have.

我们甚至远远达不到那个水平。

We don't even come close.

但这确实促使我们在计算资源有限的情况下进行大量创新工作，研究如何以少胜多。

But that does lead us to do quite a bit of innovative work in what happens when you have less compute and how to make more of less.

所以我们在这方面已经做了很多工作。

So we do a lot of that work here already.

下一个问题，我想还是这边提问。

Next question, I think also on this side.

大家好。

Hi, everyone.

我叫安迪·萨沃兹。

My name is Andy Savodzi.

我是化学工程专业的二年级研究生。

I am a second year graduate student in chemical engineering.

我的问题想问所有演讲嘉宾。

My question is to all the speakers.

你们认为哪种新兴技术的长期影响被严重低估了？

What which emerging technology do you think is seriously being underestimated in terms of its long term impact?

谢谢。

Thank you.

好的，哪种新兴技术正被严重低估？

Okay, what emergent technology is being seriously underestimated?

哇。

Wow.

好吧。

Okay.

我显然不能说人工智能，因为这很难辩驳，但它可能被低估了。

I probably obviously can't say AI because it's hard to argue, but it could be underestimated.

它可能被低估了。

It could be underestimated.

它可能被低估了，但此刻或许已不算新兴技术了。

It could be underestimated, but probably not emergent at this point.

我们不能用那个。

We couldn't use that one.

我是说，很多人确实对量子计算及其带来的影响感到好奇。

I mean, a lot of people do wonder about quantum quantum computing, what it will bring.

这可能不是我用来回答这个问题的首选方向。

It's probably not what I would hang my hat on to answer that question.

尽管我绝对支持我们在量子计算等方面的努力，但仍有诸多未知数。

Although I'm a definitely support sort of our efforts in quantum computing and so forth, but there are many unknowns.

从技术角度讲，我们甚至无法确定P是否不等于NP。

I mean, technically speaking, we don't even know if p is not equal to NP.

在计算领域，有太多未解之谜，而量子算法仅针对特定、高度结构化的问题。

Like, on the computation front, there are just so many unanswered questions, and the quantum algorithms are specific for particular, very particular structured problems.

话虽如此，我仍是个坚定的支持者。

That said, I'm a big proponent.

但很难确切指出其具体价值。

But it's hard to put my finger on that.

或许AI和量子计算在材料科学中的应用值得关注——我们能利用性能全面提升的新型材料做些什么？

I mean, perhaps the applications of both AI and and for them at quantum computing to, material science, because what could we do with different kinds of materials that are, better in a whole host of ways?

我是说，可能性几乎是无限的。

I mean, kind of the sky's the limit.

其实我也在考虑材料领域，部分原因是许多低估现象相当有趣。

I was thinking of materials as well, actually, but partly because a lot of the the underestimate is sort of interesting.

现在有如此多的关注点集中在技术创新机会上。

There's so much attention right now on on, you know, like, are the opportunities for technological innovation.

像核聚变能源或量子计算等许多尚未成熟的技术，很难说人们现在在AI领域忽视了它们。

So many technologies that aren't there yet, like fusion energy or quantum, it'd be hard to say that people are missing them and not paying any attention to them right now in AI.

但我认为材料科学会是其中之一。

But I think materials, in my mind, would be one of them.

生物学和健康领域的某些机会，特别是分子科学中的众多突破，可能比AI获得的关注少，但分子科学领域也正在发生巨大变革。

Probably some of the opportunities in biology and health, of which there are many in molecular science, that's probably getting less attention than AI right now, but there's also a huge revolution in molecular science.

我正想说同样的话。

I was going to say exactly the same thing.

我看着聚光灯不断移动，现在AI确实占据着巨大焦点，但它曾照亮生物学领域，这种关注不应停止。

I kind of watch the spotlight move around, and the spotlight is very large on AI right now, but it's it's it was shining on biology, and it shouldn't stop shining on it.

合成生物学领域正在发生各种令人振奋的事情。

There's all kinds of things going on in synthetic biology, very exciting things.

所以我认为我们需要稍微拓宽一下关注范围。

So I think we need to broaden that spotlight a little bit.

好的。

Okay.

这边请。

Over here.

你好。

Hi.

我叫Jeremy，是一名来自新加坡的学生。

My name is Jeremy, and I'm a student coming from Singapore.

我今天的问题是想问Sergei，这更偏向个人方面。

My question today is for Sergei, and it's a bit more personal personal.

我们都在成长过程中有过一些限制性信念，我很好奇你在创建谷歌时曾有哪些限制性或根深蒂固的信念需要改变？这些改变又如何影响了你的决策过程？

So we all grew up having limiting beliefs, and I was curious of what limiting beliefs or deeply held beliefs you had while building Google that you had to change, and how did that affect your decision making?

谢谢。

Thank you.

限制性信念。

Limiting beliefs.

是的。

Yeah.

我想，我的生活经历了几次巨大的转折。

I I guess, like, I had a very like, my life expanded pretty dramatically at a bunch of stages.

我出生在苏联时期的莫斯科，那是个完全不同的世界。

Like, I I was born in Moscow in The Soviet Union, and it's very different.

你知道的，非常贫穷。

You know, very poor.

嗯，那时候所有人都很穷。

Well, everybody was very poor.

我和父母、祖母住在一个400平方英尺的小公寓里，每天要爬五层楼梯。

And I lived in a little 400 square foot apartment with my parents and my grandmother, and I had to walk up five flights of stairs.

我不知道。

I don't know.

我那时并没有真正思考过外面的世界。

I didn't really think about the world outside.

我想我很幸运，父亲对外界有所察觉。

I guess I was lucky that my father kind of got a hint of the world outside.

他参加了波兰的一个会议，了解到西方世界的样子后决定举家搬迁，这在当时家族里极具争议。

He, I guess, went to some conference in Poland, where they told him what the Western world was like, and he decided to move us, which was very controversial at the time in the family.

但最终他成功带我们来到美国，虽然依旧贫困，不得不白手起家。

But eventually he got to The US and we're still, you know, very poor and had to make our way out of having nothing.

那时候我还得学习一门新语言，放弃了很多。

And, you know, at the time I had to learn a new language, gave up.

你知道，我得重新结交所有朋友。

You know, I had to meet all make all new friends.

所以这段转型期充满挑战，却也令人觉醒。

So it was sort of challenging transition, but awakening.

我觉得当我进入斯坦福研究生院时，情况有些类似，教授们给予我充分的信任和自由，加州这个地方本身就因其传统而让人在思想上感到无比解放。

And I think when I came to grad school at Stanford, it was sort of a similar, like now I had sort of this, all this freedom in the way the professors entrusted me and just something about California that was very freeing and liberating in thought given the tradition of the state.

说实话，加州现在某种程度上正在背离这种传统，不过我也不想过多抱怨。

One that we're a little bit getting away from in California, if I'm being honest, but I'm not gonna complain about that.

但我想正是这段经历。

But I guess it's this experience.

我意识到我有点倒着回答你的问题了。

I guess I'm answering your question backwards.

这其实算不上什么限制性信念。

It's not like really a limiting belief.

鉴于我的个人经历，我曾经历过看似痛苦但最终带来回报的世界观拓展。

I guess I had had the experience of expanding my world in ways that seemed very painful at the times, but later paid off, just because of my personal history.

我想说的是，那些充满挑战的转变最终都可能带来回报。

And I guess, you know, those challenging transitions can pay off.

没错。

Right.

下一个问题。

Next question.

你好。

Hello.

感谢各位的到来。

Thank you to all of you for being here.

我叫卢芭芭。

My name is Lubaba.

我是管理科学与工程专业的二年级硕士生，来自摩洛哥卡萨布兰卡。

I'm a second year master's student in management science and engineering, originally from Casablanca, Morocco.

我的问题也是提给你的，谢尔盖。

My question is also for you, Sergei.

同样更偏向个人方面。

It's also more on the personal side.

你已取得了大多数人难以企及的成就。

So you've achieved success at the scale most people never experience.

审视你现在的生活，你对美好生活的定义是什么？

Looking at your life now, what is your definition of a good life?

除了这些成就之外，对你而言什么才是真正重要的？

What does it mean to you beyond all these accomplishments?

谢

Thank

谢。

you.

好的，谢谢。

Okay, thanks.

美好生活的定义是什么？

What is the definition of a good life?

嗯，我想那就是能够享受你的人生，无论你创造了什么。

Well, I guess that's, you know, being able to enjoy your life, you know, whatever you build.

我喜欢拥有家庭。

I like to have family.

我的一个孩子在这里。

I have one of my kiddos here.

我的女朋友也在这里。

My girlfriend is here.

你知道，我很感激能和他们共度美好时光。

You know, I feel grateful to be able to spend quality time with them.

我确实非常感激在这个阶段还能受到智力上的挑战。

I do feel quite grateful to be able to be intellectually challenged sort of at this stage.

其实我在新冠疫情爆发前一个月就退休了。

I actually retired like a month before COVID hit.

这简直是最糟糕的决定。

And it was like the worst decision.

我曾幻想坐在咖啡馆里研究物理，那是我当时的热情所在。

I had this vision that I was gonna sit in cafes and study physics, which was my passion at the time.

是啊，这没能实现，因为咖啡馆都关门了。

And yeah, that didn't work because there were no more cafes.

是啊，我也不知道。

And yeah, I don't know.

我当时就是有点闷闷不乐，感觉自己状态在螺旋式下滑，思维也不够敏锐。

I was just kind of stewing and kind of felt myself spiraling, kind of not being sharp.

然后我就想，哦，我得回办公室工作，但那时候办公室已经关闭了。

And then I was like, oh, I gotta get back to the office, which the time was closed.

不过几个月后，我们开始有部分同事返回办公室，我也偶尔去一下，后来就逐渐把越来越多时间投入到后来被称为Gemini的项目上，这真的非常令人兴奋。

But, you know, after a number of months, we started to have some folks going to the office, and I started to do that occasionally, and then started spending more and more time on what later became called Gemini, which is super exciting.

能够拥有这样一个技术创新的出口，我觉得这比当初如果坚持退休要有意义得多。

And to be able to have that technical creative outlet, I think that's very rewarding as opposed to if I'd, like, stayed retired.

我觉得那会是个重大错误。

I think that would have been a big mistake.

好的。

All right.

我想我们还有时间再讨论一两个问题，这边应该还有。

I think we have time for one or two more, I believe, over on this side.

大家好，非常感谢各位的到来。

Hello, thank you guys all so much for being here.

我叫刘斯坦利，是大一新生，计划攻读管理科学与工程专业。

My name is Stanley Liu, I'm a freshman planning on studying Management Science and Engineering.

我想向三位都提一个问题。

And a question for all three of you.

先说点背景，在来到这里之前我其实非常害怕——因为这里的每个人都才华横溢。

So for some context, like before arriving here I was absolutely terrified because everyone here is like super talented.

我当时就想：这是什么情况？

I'm like, what is going on?

完全不明白自己为何会在这里，感觉每个人都比我聪明得多。

I have no clue why I'm here, and everyone just seems way too smart for me.

但认识大家之后，我发现其实都很接地气，都是普通人。

But after getting to know people, I realized they're all just really relatable and normal people.

那么请问三位——你们被视为世界顶尖的，但如果有件能让听众感到亲切的真实小事，会是什么呢？

So for all three of you, you guys are viewed as like some of the best leaders, innovators in the world, But if there's one thing you'd like to share that is reassuringly relatable and human about yourself, what would that be?

好的。

Yeah.

你想先开始吗，抱歉，这位同学？

You wanna start, sorry, guy?

好的。

Okay.

我先分享，然后我会试着收回这句话。

I'm gonna share it, and then I'm going to try to undo it.

不过好吧。

But okay.

我意识到有时我会不好意思问自己不懂的事情，但我还是会继续。

I realize that sometimes I'm embarrassed to ask things I don't know, but I will go ahead.

等等。

Wait.

管理科学与工程是什么？

What is management science and engineering?

就像是《呆伯特》漫画里那种，我要当管理者的感觉吗？

Like, is it like a Dilbert kind of, like, I'm gonna manage?

这是怎么运作的？

How does that work?

这是一门课程。

It's a class.

这是一门课程。

It's a class.

这是一个专业。

It's a major.

等等。

Wait.

这是个系。

This It's a department.

管理科学这门课。

The class of management science.

我想我本该更仔细阅读细节的。

Is that the I guess I should have read the details more.

是主修专业。

It's major.

它叫管理科学，然后呢？

It's called management science and then?

像是系部那种。

Like the department.

这是一个系部。

It's a department.

这是一个系部。

It's a department.

是的。

Yes.

但你们具体学什么？

But what do you study?

比如，具体有哪些课程？

Like, what are the classes?

管理科学与工程运营。

So management science and operate and engineering.

我简单说一下，他们刚庆祝了二十五周年纪念，但这个系是由三个部门合并而成的：工业工程、运筹学和工程经济系统。

I'm going to just say they just had their twenty fifth anniversary, but they were the merger of three departments, industrial engineering, operations research, and engineering economic systems.

所以我想这能让你对他们研究的领域有个大致了解。

So I think that of gives you a little triangle there of what they do.

有些大学会单独设立工业工程或运筹学专业。

So some universities will have an industrial engineering or operations research.

我们这里把这些领域都整合在管理科学与工程系，这个系还赞助了创业思想领袖研讨会，也就是我们现在

We have this all bundled together here in management science and engineering, which is the department that sponsors the entrepreneurial thought leaders seminar, which is what we are

好的。

Okay.

正在进行的这个活动。

Conducting right now.

好的。

Alright.

嗯，我之前确实不太了解这个，这真是我的尴尬真相。

Well, I guess I didn't really know that, so that's my embarrassing truth.

不过我很高兴我问了这个问题。

But I'm glad I asked.

让我显得平易近人的是，我能向谢尔盖·布林解释清楚事情。

What makes me relatable is that I can explain things to Sergey Brin.

请关注他们。

Pay attention to them.

约翰，我就不为难你了，我们进入最后一个问题。

I'll let you off the hook, John, and we'll go to our last question.

我们还有最后一个问题吗？

Do we have one more question?

我想我们还有。

I think we do.

是的。

Yeah.

我们可以

We can

再问一个问题。

ask one more.

嗨，我叫齐娜。

Hi, my name is Zina.

实际上我是这门课程的助教。

I'm actually the course assistant for the class.

感谢你们的到来，能上这最后一堂课真是太好了。

So thank you for being here, and we it's a great thing that we can have this last class.

我想问一个我们经常问嘉宾的问题：您会给学生们什么建议，告诉他们该如何安排时间以保持领先优势？

I'm going to ask you something that we ask a lot of our speakers usually, is to give a recommendation to the students as to what do you do with your time to stay on top of things.

您刚才提到您很喜欢保持敏锐，紧跟人工智能等领域的最新动态。

And you just said you really like staying sharp and being on top of what's happening in AI and whatnot.

那你平时读什么书呢？

So what books do you read?

你在车里会听哪些播客节目？

What podcasts do you listen to in your car?

好的。

Okay.

我尽量在不做广告的情况下回答这个问题。

I'm gonna try to do this without advertising or so.

嗯，好吧。

Well okay.

我喜欢这样做，但你们现在不应该尝试，因为我们即将推出更好的版本。

So the thing I like to do, but you shouldn't do it now because we have, like, way better version coming.

不过我经常在车里和Gemini Live对话，向它提问。

But I do talk to Gemini live in the car often, and I ask.

但目前公开版本还不是我们的最佳版本，所以今天你们最好不要尝试。

But the publicly available version right now is not our good version, so, like, you shouldn't do it today.

但给我几周时间正式推出我能接触到的版本，因为目前公开版本背后运行的还是个老古董模型。

But give me a few weeks to actually ship what I have access to, because we have like an ancient model behind it in the publicly released version right now.

有点尴尬。

It's little embarrassing.

但我确实会随便问它些问题，你知道的，各种事情。

But I do like ask it like, you know, whatever.

我想开发个数据中心，需要什么就问什么。

I wanna develop a data center, you know, whatever I need.

这种电力需要几百兆瓦？那种电力需要多少？

How many hundreds of megawatts of this kind of power, that kind of power?

大概要花多少钱？

How much it's gonna cost?

我开车时就随便和它聊这些。

And I just talk to it about stuff on my drive.

好的。

Okay.

这确实有点像Gemini的自我宣传。

That does seem kind of self advertising with Gemini.

我是说，我确实会定期听一大堆播客。

I mean, I do periodically listen to a whole bunch of podcasts.

其实我最喜欢的是All In那帮家伙。

I like the, the All In Guys are actually one of my favorites.

他们都是很棒的主持人。

And they're great hosts.

我们刚在佛罗里达拜访了另一位播客主本·夏皮罗。

We just visited Ben Shapiro, another podcaster down in we were in Florida.

参观了他的录音室。

Got to see his studio.

说实话，很多播客主本人见面时都挺有意思的。

I mean, a bunch of these podcasters are actually pretty fun to meet in person.

不过，好吧，我猜这不是你了解它的方式。

But, yeah, I guess, okay, that's not how you're gonna learn about it.

你会去听，但我确实只是听听看他们有什么新鲜事。

You're gonna, but but I do just listen to them and see what's up.

不过我更喜欢在开车时进行互动讨论，所以这就是为什么我会和AI聊天，尽管听起来有点尴尬。

But but I do prefer to have an interactive discussion on my drive, so that's why I talk to the AI, as embarrassing as that sounds.

好的。

Okay.

我觉得这实际上是对未来的一瞥。

Sort of a glimpse of the future, I think, actually.

这是个不错的结束方式。

That's a good way to end.

我们可能都会这么做

We'll probably all be doing it

那么，谢谢你，约翰。

So, thank you, John.

谢谢你，谢尔盖。

Thank you, Sergei.

我还要感谢Emily Ma。

I also wanted to thank Emily Ma.

Emily是斯坦福大学的兼职讲师。

Emily is a Stanford adjunct lecturer.

Emily是这门课程的联合讲师。

Emily is a co instructor of the course.

她同时也是谷歌员工，她预见到了这次活动的潜力并与我们合作。

She's also a Google employee, and she saw the potential for this event and partnered with us.

非常感谢大家。

So thank you very much.

感谢各位莅临，共同庆祝工程学院成立一百周年。

Thank you all for being here, for celebrating the School of Engineering's one hundredth year.

这是为我们第一个百年画上句号的完美方式，让我们拭目以待未来。

This was a perfect way to close out our first century, and let's see what happens next.

谢谢。

Thank you.

谢谢。

Thank you.

谢谢。

Thank you.