本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
我基本上实现了所有梦想，我们在多个领域都处于科学的最前沿，包括应用科学和机器学习。
I'm basically doing everything I ever dreamed of, and we're at the absolute frontier of science in so many ways, applied science as well as machine learning.
这种处于前沿并首次发现某事物的感觉令人振奋。
And that's exhilarating that feeling of being at the frontier and discovering something for the first time.
欢迎收听《谷歌DeepMind》播客，我是汉娜·弗莱教授。
Welcome to Google DeepMind, the podcast with me, professor Hannah Fry.
对人工智能而言，这是非凡的一年。
It has been an extraordinary year for AI.
我们见证了重心从大语言模型向代理型AI的转变。
We have seen the center of gravity shift from large language models to agentic AI.
我们见证了AI加速药物发现，以及多模态模型与机器人技术和无人驾驶汽车的融合。
We've seen AI accelerate drug discovery and multimodal models integrated into robotics and driverless cars.
这些都是我们在本播客中详细探讨过的话题。
Now these are all topics that we've explored in detail on this podcast.
但作为本年度的最后一期节目，我们希望采用更宏观的视角，超越头条新闻和产品发布，去思考一个更宏大的问题。
But for the final episode of this year, we wanted to take a broader view, something beyond the headlines and product launches, to consider a much bigger question.
这一切究竟将走向何方？
Where is all this heading really?
哪些科学和技术问题将定义下一个阶段？
What are the scientific and technological questions that will define the next phase?
而经常思考这些问题的人之一就是谷歌DeepMind的CEO兼联合创始人德米斯·哈萨比斯。
And someone who spends quite a lot of their time thinking about that is Demis Esabes, CEO and cofounder of Google DeepMind.
欢迎回到播客，德米斯。
Welcome back to the podcast, Demis.
很高兴再次见到你。
Lovely to see you again.
很高兴回来。
Great to be back.
我是说，确实
I mean, quite
过去一年发生了很多事。
a lot has happened in the last year.
确实如此。
It has.
你觉得最大的转变是什么？
What's sort of the biggest shift do you think?
哦，哇。
Oh, wow.
我是说，发生了太多事情。
I mean, it's just so much has happened.
就像你说的，感觉我们把十年的事压缩在了一年里。
As you said, it's just it feels like we packed in ten years in one year.
我认为发生了很多变化。
I think a lot's happened.
对我们来说，模型的进步——我们刚发布了Gemini 3，对其多模态能力非常满意，所有这些都进展得非常顺利。
I mean, certainly for us, the progress of the models, we've just released Gemini three, which we're really happy with multimodal capabilities, all of those things have just advanced really well.
而夏季最让我兴奋的可能是世界模型的突破性进展。
And then probably the thing, I guess, over the summer that I'm very excited about is world models being advanced.
我确信我们会讨论这个话题的。
I'm sure we're gonna talk about that.
是的，当然。
Yeah, absolutely.
我们稍后会更详细地讨论所有这些内容。
We will get onto all of that stuff a bit more detail in a moment.
记得我第一次在这个播客采访你时，你谈到了根节点问题，以及利用AI解锁下游效益的理念。
I remember the very first time I interviewed you for this podcast and you were talking about the root node problems, about this idea that you can use AI to kind of unlock these downstream benefits.
不得不说，你兑现得相当不错。
And you've made pretty good on your promise, have to say.
能给我们更新下这些项目的进展吗？
Do you wanna give us an update on where we are with those?
有哪些即将实现的目标？哪些是你们已经解决或接近解决的？
What are the things that are just around the corner and the things that you've sort of solved or near solved?
当然，最显著的成果证明就是AlphaFold——想到AlphaFold2问世即将迎来五周年，这感觉有点不可思议。
Yeah, well, of course, obviously the big proof point was AlphaFold and sort of crazy to think we're coming up to like five year sort of anniversary of AlphaFold being sort of announced to the world AlphaFold two at least.
所以我想这就是证明，解决这类根节点问题是可能的。
So that was the proof, I guess, that it was possible to do these root node type of problems.
我们现在正在探索所有其他方面。
And we're exploring all the other ones now.
想想材料科学，我特别想实现室温超导体和更好的电池这类东西。
Think material science, I love to do a room temperature superconductor and better batteries, these kinds of things.
我认为这些都在计划之中，各种更好的材料。
I think that's on the cards, better materials of all sorts.
我们还在研究核聚变。
We're also working on fusion.
这是新宣布的合作关系吗
Is this a new partnership that's been announced
关于核聚变的？
with Fusion?
是的。
Yeah.
我们刚刚宣布了一个深度合作伙伴关系。
We've just announced partnership with a deep one.
我们之前已经在与他们合作，但现在是与Commonwealth Fusion建立了更深入的合作，我认为他们可能是研究传统托卡马克反应堆中最优秀的初创公司。
We we already were collaborating with them, but it's a much deeper one now with Commonwealth Fusion, who, you know, I think are probably the best startup working on at least traditional tokamak reactors.
所以他们可能最接近拥有可行的方案，我们希望能帮助加速这一进程，协助他们控制等离子体和磁体，或许还包括一些材料设计。
So they're probably closest to to having something viable, and we wanna help accelerate that, helping them contain the plasma and the magnets and maybe even some material design there as well.
这非常令人兴奋。
So that's exciting.
此外，我们也在与量子团队的同事们合作，谷歌量子人工智能团队正在做惊人的工作。
And then we're collaborating also with our Quantum colleagues, which they're doing amazing work at the quantum AI team at Google.
我们正在用机器学习帮助他们改进纠错编码。
And we're helping them with error correction codes where we're using our machine learning to help them.
也许有一天他们也会帮助我们。
And then maybe one day they'll help us.
完美。
Perfect.
是的，正是如此。
Yes, exactly.
核聚变尤其关键，我的意思是，它能为世界带来的改变、所释放的潜力是巨大的。
The fusion one is particularly, I mean, the difference that would make to the world, that would be unlocked by that is gigantic.
没错。
Yeah.
我是说，核聚变一直是圣杯般的存在。
I mean, fusion's always been the holy grail.
当然，我认为太阳能也很有前景，对吧？
Of course, I think solar is very promising too, right?
本质上就是在利用云层和天空中的聚变反应堆。
Effectively using the fusion reactor in the clouds and the sky.
但如果我们能拥有模块化聚变反应堆，这种近乎无限的清洁可再生能源的前景显然将改变一切。
But I think if we could have modular fusion reactors, you know, this promise of almost unlimited renewable clean energy would be obviously transform everything.
这就是圣杯。
And that's the holy grail.
当然，这也是我们应对气候变化的一种方式。
And of course, that's one of the ways we could help with climate.
确实能让我们现有的许多问题迎刃而解。
Does make a lot of our existing problems sort of disappear.
毫无疑问。
Definitely.
我是说，它开辟了许多可能性，这就是为什么我们将其视为根源节点。
I mean, it opens up many this is why we think of it as a root node.
当然，它直接有助于解决能源和污染问题，并缓解气候危机。
Of course, it helps directly with energy and pollution and helps with the climate crisis.
而且如果能源真正可再生、清洁且极其廉价或近乎免费，那么许多其他事情也会变得可行，比如水资源获取——因为我们可以几乎在任何地方建立海水淡化厂，甚至制造火箭燃料。
But also if energy really was renewable and clean and super cheap or most free, then many other things would become viable, like, you know, water access because we could have desalination plants pretty much everywhere, even making rocket fuel.
要知道，海水中含有大量氢和氧。
You know, it's just there's lots of seawater that contains hydrogen and oxygen.
这基本上就是火箭燃料，只不过需要大量能量才能将其分解为氢和氧。
That's basically rocket fuel, but it just takes a lot of energy to split it out into hydrogen oxygen.
但如果能源既廉价又可再生还相当清洁，那为什么不做呢？
But if energy is cheap and renewable and sort of clean, then why not do that?
你知道，你可以让它24/7不间断生产。
You know, you could have that producing 20 fourseven.
你也看到人工智能在数学应用方面发生了很多变化，对吧？
You're also seeing a lot of change in the AI that is applying itself to mathematics, right?
比如在国际数学奥林匹克竞赛中获奖。
The, you know, winning medals in the International Maths Olympiad.
然而同时，这些模型在高中数学上却可能犯相当基础的错误。
And yet at the same time, these models can make quite basic mistakes in high school math.
是的。
Yes.
为什么会出现这种矛盾呢？
Why is there that paradox?
没错。
Yeah.
我觉得这其实很吸引人。
I think it's fascinating, actually.
这是最引人入胜的事情之一，或许也是我们尚未实现通用人工智能的关键问题之一，亟需解决。
One of the most fascinating things and probably that needs to be fixed as one of the key things why we're not at AGI yet.
正如你所说，我们在其他团队已取得诸多成功，比如在国际数学奥林匹克竞赛中获得金牌。
As you said, we've had a lot of success in other groups on getting, like, gold medals at the International Mass Olympiad.
你看看那些题目，都是超级难题，只有世界上最顶尖的学生才能解答。
You look at those questions, and they're super hard questions that only the top students in the world can do.
另一方面，如果你以某种方式提出问题，我们在日常生活中与聊天机器人互动时都见过，它会在逻辑问题上犯一些相当低级的错误。
And on the other hand, if you pose a question in a certain way, we've all seen that with experimenting with chatbots ourselves in our daily lives that it can make some fairly trivial mistakes on logic problems.
它们目前还下不好像样的国际象棋，这令人惊讶。
They can't really play decent games of chess yet, which is surprising.
所以这些系统在一致性方面仍然有所欠缺。
So there's something missing still from these systems in terms of their consistency.
我认为这正是你对通用人工智能（AGI系统）的期待之一，即它应该具有全局一致性。
And I think that's one of the things that's you would expect from a general intelligence, you know, an AGI system is that it would be consistent across the board.
所以有时候人们称之为锯齿状智能。
And so sometimes people call it jagged intelligences.
它们在特定领域表现非常出色，甚至能达到博士水平，但在其他方面可能连高中生水平都不到。
So they're really good at certain things, maybe even like PhD level, but then other things, they're like not even high school level.
这些系统的表现仍然非常不均衡。
So it's very uneven still, the performances of these systems.
在某些维度上它们令人印象深刻，但在其他方面仍相当基础。
They're very, very impressive in certain dimensions, but they're still pretty basic in others.
我们必须缩小这些差距。
And we've gotta close those gaps.
关于原因有各种理论，视情况而定，甚至可能与图像感知和标记化的方式有关。
And, you know, there were theories as to why, and it depending on the situation, it could even be the way that an image is perceived and tokenized.
有时候它实际上连单词中的所有字母都识别不全，比如在数单词中的字母时就会出错。
So sometimes, actually, it doesn't even get all the letters that you you know, so when you count letters in words, it sometimes gets that wrong.
它可能无法准确识别每一个单独的字母。
It may not be seeing that each individual letter.
所以这些事情背后有各种不同的原因，每一个问题都可以被解决，然后你就能看到还剩下什么。
So there's sort of different reasons for some of these things, and each one of those can be fixed, and then you can see what's left.
但我认为一致性是另一个关键点，还有推理和思考能力。
But I think consistency, I think another thing is reasoning and thinking.
我们现在有了思考系统，在推理阶段它们会花更多时间思考，从而输出更好的答案。
So we have thinking systems now that at inference time, they spend more time thinking, and then they're better at outputting their answers.
但就目前而言，它还不能完全稳定地利用这段思考时间来有效复核，并使用工具来验证输出内容。
But it's not sort of super consistent yet in terms of, like, is it using that thinking time in a useful way to actually double check and use tools to double check what it's outputting?
我认为我们正在这条路上前进，但可能只完成了50%的进度。
I think we're we're on the way, but maybe we're only 50% of the way there to having that.
我也在想AlphaGo和AlphaZero的故事，当你们移除了所有人类经验后，发现模型反而有所提升。
I also wonder about that story of AlphaGo and then AlphaZero, where you sort of took away all of the human experience and found that the model actually improved.
是啊。
Yeah.
在你们创建的模型中，是否存在这种科学或大规模应用的版本？
There a scientific or a mass version of that in the models that you're creating?
我认为也许，就我们当前正在构建的系统而言，它更像是AlphaGo。
I think maybe, I think with what we're trying to build today, it's more like AlphaGo.
实际上这些大型语言模型、基础模型，它们始于全人类的知识——也就是我们上传到互联网的一切（如今几乎包罗万象），并将这些知识压缩成某种实用的产物，供其查询和泛化使用。
So effectively these large language models, these foundation models, they're starting with all of human knowledge, you know, what we put on the internet, which is pretty much everything these days, and compressing that into some useful artifact, which they can look up and generalize from.
但我确实认为，我们仍处于发展这种搜索或顶层思考能力的早期阶段，就像AlphaGo需要利用模型来引导有效的推理轨迹和规划思路，从而针对特定问题得出最佳解决方案。
But I do think we're still in the early days of having this search or thinking on top like AlphaGo had to kind of use that model to direct in useful reasoning traces, useful planning ideas, and then come up with the best, you know, solution to whatever the problem is at that point in time.
所以目前我并不觉得我们受到人类知识（如互联网）边界的限制。
So I I don't feel like we're constrained at the moment with the kind of limit of human knowledge, like the Internet.
我认为当前主要问题在于，我们还未能像运用AlphaGo那样完全可靠地使用这些系统。
I think the main issue at the moment is we don't know how to use those systems in a reliable way fully yet in the way we did with AlphaGo.
不过当然，AlphaGo更容易实现，因为它只是个游戏。
But, of course, that was a lot easier because it was just it was a game.
我认为一旦拥有AlphaGo这样的系统，就可以像Alpha系列那样回溯，开发出类似AlphaZero的版本——让系统自主探索知识。
I think once you have AlphaGo, there, you could go back just like we did with the alpha series and do an alpha zero, where it starts sort of discovering knowledge for itself.
我认为这将是下一步发展方向。
I think that would be the next step.
这显然更难。
That's obviously harder.
所以我认为，首先尝试创建一个类似AlphaGo的系统作为第一步是好的，然后我们可以考虑类似AlphaZero的系统。
And so I think it's good to try and create the first step first with some kind of AlphaGo like system, and then we can think about an AlphaZero like system.
但这也是当前系统缺失的一个能力，即在线学习和持续学习的能力。
But that is also one of the things missing from today's systems is the ability to online learn and continually learn.
你看，我们训练这些系统，对它们进行平衡和后训练，然后它们就投入使用了，但它们不会像我们一样在现实世界中持续学习。
So, you know, we train these systems, we balance them, we post train them, and then they're out in the world, but they don't continue to learn out in the world like we would.
我认为这是实现AGI之前这些系统所缺失的另一关键部分。
And I think that's another critical missing piece from these systems that will be needed before AGI.
关于所有这些缺失的部分，我知道目前大家都在竞相推出商业产品，但我也知道Google DeepMind的根基其实在于科学研究这一理念。
In terms of all of those missing pieces, I mean, I know that there's this big race at the moment to release commercial products, but I also know that Google DeepMind's roots really lie in in that idea of scientific research.
是的。
Yes.
我最近看到你的一句话：'如果按我的想法，我们会让AI在实验室里待更长时间，做更多像AlphaFold这样的项目，或许能攻克癌症之类的难题。'
And I I found a quote from you where you recently said, if I'd had my way, we would have left AI in the lab for longer and done more things like AlphaFold, maybe cured cancer or something like that.
确实
Do
你认为我们没有选择那条更慢的路线是否有所损失？
you think that we lost something by not taking that slower route?
我认为我们既有所失，亦有所得。
I think we lost and gained something.
我觉得那本应是更纯粹的科学探索路径。
So I feel like that would have been the more pure scientific approach.
至少那是我最初的计划——大约十五、二十年前，当几乎没人在研究人工智能时，我们刚准备创立DeepMind。
At least that was my original plan, say fifteen, twenty years ago, when almost no one was working on AI, we're just about to start DeepMind.
当时人们觉得研究这个简直是疯了。
People thought it was a crazy thing to work on.
但我们始终坚信不疑。
But we believed in it.
我认为当时的理念是：如果我们取得进展，就会持续渐进式地向AGI迈进，谨慎对待每个步骤及其安全影响，不断分析系统行为等等。
And I think that the idea was if we would make progress, we would continue to sort of incrementally build towards AGI, be very careful about what each step was and the safety aspects of it and so on, analyze what the system was doing and so on.
但与此同时，你不必等到AGI实现就能让它发挥作用。
But in the meantime, you wouldn't have to wait till AGI arrived before it was useful.
你可以将这项技术分支出来，以对社会真正有益的方式应用它，特别是推动科学和医学进步。
You could branch off that technology and use it in really beneficial ways to society, namely advancing science and medicine.
实际上这正是我们用AlphaFold所做的——它本身不是一个基础模型或通用模型，但采用了相同的技术（比如Transformer等），再融合该领域更专业的内容。
So exactly what we did with AlphaFold actually, which it's not a foundation model itself, general model, but it uses the same techniques, you know, transformers and other things, and then blends it with more specific things to that domain.
所以我设想过完成大量这类项目，它们会产生巨大影响。就像我们发布AlphaFold一样向世界开放，甚至实现治愈癌症等突破，同时我们在实验室继续推进AGI研究。
So I imagined a whole bunch of those things getting done, which would be hugely bet you know, you'd release to the world for just like we do with AlphaFold and indeed do things like cure cancer and so on, whilst we were working on the sort of more the AGI track in the lab.
现在事实证明大规模聊天机器人是可行的，人们觉得有用，它们已演变成能做远超聊天和文本的基础模型，比如Gemini。
Now it's turned out that chatbots were possible at scale and people find them useful, and then they've now morphed into these foundation models that can do more than chat and text, obviously, including Gemini.
它们能处理图像、视频等各种内容。
They can do images and video and all sorts of things.
作为产品在商业上也取得了巨大成功。
And that's also been very successful commercially in terms of a product.
我也非常喜欢这一点。
And I love that too.
比如，我一直梦想拥有一个终极助手，能在日常生活中帮助你，提高效率，甚至可能保护你的思维空间免受注意力分散，让你能专注并保持心流状态。
Like, I've always dreamed of having the ultimate assistant that would help you in everyday life, make it more productive, maybe even protect your brain space a bit as well from an attention so that you can focus and be in flow and so on.
因为你知道，如今社交媒体上全是噪音、噪音、噪音。
Because, you know, today with social media, it's just noise, noise, noise.
而我认为真正为你服务的人工智能可以帮助我们解决这个问题。
And I think AI actually that works for you could help us with that.
所以我觉得这是好事，但也造成了这种近乎疯狂的竞赛状态——许多商业机构甚至国家都在竞相改进并超越彼此。
So I think that's good, but it has created this pretty crazy race condition where there's many commercial organizations and even nation states rushing to improve and overtake each other.
这使得同时进行严谨的科学研究变得困难。
And that makes it hard to do sort of rigorous science at the same time.
我们试图兼顾两者，我认为我们正在找到平衡点，但这确实增加了难度。
We try to do both, and I think we're getting that balance right, but it makes it harder.
另一方面，当前发展模式也有许多优势，当然最明显的是该领域获得了更多资源投入。
On the other hand, there are lots of pros of the way it's happened, which is, of course, there's a lot more resources coming into the area.
这无疑加速了技术进步。
So that's definitely accelerated progress.
而且，有趣的是，我认为普通大众实际上只比技术前沿落后几个月就能用上最新成果。
And, also, I think the general public are actually, interestingly, only a couple of months behind the absolute frontier in terms of what they can use.
所以每个人都有机会亲身体验AI会是什么样子。
So everyone gets a chance to sort of feel for themselves what AI is gonna be like.
我认为这是件好事，而且各国政府也逐渐对此有了更深的理解。
And I think I think that's a good thing and then governments sort of understanding this better.
奇怪的是，去年这个时候，很多人都在讨论规模扩展最终会碰壁，担心我们会耗尽数据。
The thing that's strange is that, I mean, this time last year, I think there was a lot of talk about scaling eventually hitting a wall about us running out of data.
然而现在我们正在录制时，Gemini 3刚刚发布，它在一系列不同基准测试中都处于领先地位。
And yet we're recording now Gemini three has just been released and it's leading on this whole range of different benchmarks.
这是怎么做到的？
How has that been possible?
不是说规模扩展会遇到瓶颈吗？
Like wasn't there supposed to be a problem with scaling hitting a wall?
我想很多人都这么认为，特别是当其他公司的进展相对缓慢时。
I think a lot of people thought that, especially as other companies are sort of had slower progress, should we say.
但我认为我们从未真正遇到过所谓的瓶颈。
But I think we've never really seen any wall as such.
我想说的是，也许存在收益递减的情况。
Like what I would say is, maybe there's like diminishing returns.
当我说这个时，人们往往只想到，哦，那就是没有收益了。
And people, when I say that, people think, only think like, oh, so there's no returns.
就像非此即彼一样。
Like, it's zero or one.
要么是指数增长，要么是渐进停滞。
It's either exponential or it's asymptotic.
不。
No.
实际上，这两种状态之间有很大空间，我认为我们正处于中间地带。
Actually, there's a lot of room between those two regimes, and I think we're in between those.
并不是说每次发布新版本都能在所有基准测试上实现性能翻倍。
So it's not like you're gonna double the performance on all the benchmarks every time you release a new iteration.
也许那正是非常早期的阶段发生的情况，你知道的，大约三四年前。
Maybe that's what was happening in the early very early days, you know, three, four years ago.
但你正在获得显著的改进，就像我们在Gemini 3上看到的那样，这些改进完全值得投入和回报。
But you are getting significant improvements like we've seen with Gemini three that are well worth the investment and the return on that investment and doing.
所以我们还没有看到任何放缓的迹象。
So that we haven't seen any slowdown on.
确实存在一些问题，比如我们是否正在耗尽可用的数据？
There are issues like, are we running out of just available data?
但有办法解决这个问题，比如合成数据。
But there are ways to get around that, you know, synthetic data.
要知道，这些系统已经足够强大了。
You know, these systems are good enough.
它们可以开始生成自己的数据，特别是在编程和数学等可以验证答案的领域。
They can start generating their own data, especially in certain domains like coding and math, where you can verify the answer.
从某种意义上说，你可以产生无限的数据。
In some sense, you could produce unlimited data.
不过要知道，这些问题本质上都是研究课题。
So, you know, all of these things though are research questions.
我认为这是我们一直以来的优势——始终将研究置于首位。
And I think that's the advantage that we've always had is that we've always been sort of research first.
而且我相信我们拥有最广最深的研究人才储备，这一直是我们的传统优势。
And I think we have the broadest and deepest research bench, always have done.
回顾过去十年的突破，无论是Transformer架构、AlphaGo、AlphaZero，还是我们讨论过的所有成果，它们都源自谷歌或DeepMind。
And if you look back at the last decade of advances, whether that's transformers or AlphaGo, AlphaZero, any of the things we just discussed, they were all came out of Google or DeepMind.
所以我常说，如果需要更多科学创新，我坚信我们会像过去十五年那样，继续成为重大突破的发源地。
So I've always said, like, if more innovations are needed, scientific ones, then I would back us to be the place to do it just like we were in the previous sort of fifteen years for a lot of the big breakthroughs.
我认为这正是当前正在发生的事。
So I think that's just what's transpiring.
实际上我特别喜欢研究难度升级的阶段，因为这不仅需要世界级工程能力（这本身已极具挑战），还必须结合世界级科研实力——而这正是我们的专长所在。
And I actually really like it when the terrain gets harder because then it's not just world class engineering you need, which is already hard enough, but you have to ally that with world class research and science, which is what we specialize in.
除此之外，我们还拥有TPU等世界级基础设施的优势，这些都是我们长期重点投入的领域。
And on top of that, we also have the advantage of world class infrastructure with our TPUs and and other things that we've invested in a lot for a long time.
因此我认为这种组合使我们能够同时处于创新前沿和规模化领域。
And so that combination, I think, allows us to be at the frontier of the innovations as well as the scaling part.
实际上你可以理解为我们将50%的精力投入规模化，另外50%用于创新。
And we effectively you can think of as 50% of our efforts on scaling, 50% of it is on innovation.
我认为我的判断是——要实现AGI这两者缺一不可。
And I think you my betting is you're gonna need both to get to AGI.
比如在Gemini 3.0这个并不算顶尖的模型中，我们仍能看到幻觉问题存在。
I mean, one thing that we are still seeing even in Gemini three point o, which isn't an exceptional model, is this idea of hallucinations.
有项指标显示，它仍会在应该拒绝回答时给出答案。
So I think there was one metric that said, it can still give an answer when actually it should decline.
是的。
Yes.
能否构建一个让Gemini像AlphaFold那样输出置信度评分的系统？
I mean, could you build a system where Gemini gives a confidence score in the same way that AlphaFold does?
是的，我认为可以做到。
Yeah, I think so.
我认为我们实际上需要这个功能。
And I think we need that actually.
我觉得这某种程度上是目前缺失的一环。
And I think that's sort of one of the missing things.
我认为我们已经接近目标了。
I think we're getting close.
模型越优秀，它们就越了解自己的认知边界——如果这么说能理解的话。
I think the better the models get, the more they know about what they know, if that makes sense.
而且我认为模型越可靠，我们就越能依赖它们进行某种程度的自省或深度思考，让它们自己意识到对某些答案存在不确定性。
And I think the more reliable we could sort of rely on them to actually introspect in some way or do more thinking and actually realize for themselves that they're uncertain or there's uncertainty over this answer.
接下来我们需要研究如何通过训练，让它能够把这种不确定性作为合理答案输出。
And then we've got to sort of work out how to train it in a way that where it can output that as a reasonable answer.
我们在这方面已有进步，但有时它仍会强迫自己回答本不该回答的问题。
We're getting better at it, but it still sometimes forces itself to answer when it probably shouldn't.
这种情况就可能导致幻觉回答。
And then that can lead to a hallucination.
目前很多幻觉都属于这种类型。
A lot of the hallucinations are of that type currently.
所以这里还缺少一个需要解决的环节。
So there's a missing piece there that sort of has to be solved.
你说得对，我们确实在AlphaFold上解决了这个问题，但显然是以更局限的方式。
And that you're right, as we did solve it with AlphaFold, but in in obviously a much more limited way.
因为理论上在后台应该存在某种对下一个token概率的评估机制。
Because presumably behind the scenes, there is some sort of measure of probability of of whatever the next token might be.
是的。
Yes.
是针对下一个token的。
There is of the next token.
这就是它的工作原理。
That's how it works.
但这并不能告诉你整体架构的情况。
But that doesn't tell you the overall arching piece.
你对这个事实或整个陈述有多大把握？
How confident are you about this entire fact or this entire statement?
我认为这就是为什么我们需要通过思考步骤和规划步骤来重新审视你刚才输出的内容。
And I think that's why we'll need to use the thinking steps and the planning steps to go back over what you just output.
目前的情况有点像这些系统就像在和人交谈，而当对方状态不佳时，他们真的只是把脑海里浮现的第一个念头告诉你。
At the moment, it's a little bit like the systems are just it's like talking to a person and they just you know, when when they're in on a bad day, they're just literally telling you the first thing that comes to their mind.
大多数情况下，这样是没问题的。
Most most of the time, that will be okay.
但有时遇到非常棘手的事情时，你会想要停下来稍作思考，或许应该重新斟酌即将要说的话并加以调整。
But then sometimes when this very difficult thing, you'd want to like stop pause for a moment and maybe go over what you were about to say and adjust what you were about to say.
不过这种现象在当今世界可能越来越少见了，但这仍然是更理想的对话方式。
But perhaps that's happening less and less in the world these days, but that's still the better way of having a discourse.
所以，你可以这样理解。
So, you know, I think you can think of it like that.
这些模型需要在这方面做得更好。
These models need to do that better.
好的。
Okay.
是的。
Yeah.
我也真的很想和你聊聊模拟世界以及在其中部署智能体的事，因为我们今年早些时候确实和你们的Genie团队讨论过。
I also really wanna talk to you about the simulated worlds and putting agents in them, because we've to talk to your Genie team Yes, early this year.
告诉我你为什么关心这个问题：世界模型能做哪些语言模型做不到的事？
Tell me why you care about What can a world model do that a language model can't?
嗯，你看，这可能是我最持久的热情了——除了AI之外，还有世界模型和模拟。
Well, look, it's probably my longest standing passion is world models and simulations in addition to AI.
当然，这一切在我们最近的工作中融合在了一起，比如Genie项目。
And of course, it's all coming together in our most recent work like Genie.
我认为语言模型能够理解很多关于世界的事情。
And I think language models are able to understand a lot about the world.
实际上我认为它们理解得比我预期的还要多，因为语言可能比我们想象的更加丰富。
I think actually expected, more than I expected, because language is actually probably richer than we thought.
它蕴含的世界信息可能比我们——甚至语言学家——所想象的还要丰富。
It contains more about the world than we maybe even even linguists maybe imagined.
这一点现在已被这些新系统所证实。
And that's, you know, proven now with these new systems.
但关于世界的空间动态、空间感知、我们所处的物理环境及其机械运作原理，仍有大量内容难以用语言描述，且通常不会出现在语料库中。
But there's still a lot about the spatial dynamics of the world, spatial awareness, and the physical context we're in, and how that works mechanically, that it's hard to describe in words and isn't generally described in corpuses of words.
其中很多内容与从经验中学习——特别是实时经验——密切相关。
And a lot of this is allied to learning from experience, online experience.
有很多事物是你无法用语言准确描述的。
There's a lot of things which you can't really describe something.
你必须亲身体验才能理解。
You have to just experience it.
比如传感器这类事物就极难用语言表述——无论是运动角度、气味还是各类传感器，用任何语言都难以准确描述。
Maybe the sensors and so on are very hard to put into words, you know, whether that's motor angles and smell and these kinds of sensors, it's very difficult to describe that in any kind of language.
因此我认为这背后存在一整套需要探索的领域。
So I think there's a whole set of things around that.
我认为如果我们想让机器人技术发挥作用，或者想要一个可能在日常生活中陪伴你的通用助手——无论是通过眼镜、手机等设备帮助你日常生活而不仅限于电脑——我们就需要这种对世界的理解。
And I think if we want robotics to work or a universal assistant that maybe comes along with you in your daily life, maybe on glasses or, you know, on your phone and helps you in your everyday life, not just on your computer, you're gonna need this kind of world understanding.
而世界模型正是这一理解的核心。
And world models are at the core of that.
我们所说的世界模型，是指那种能够理解世界运行机制因果关系的模型，包括直觉物理、物体运动方式及行为规律。
What we mean by world model is this sort of model that understands the causative effect of the mechanics of the world, intuitive physics, but how things move, how things behave.
实际上，我们现在已经在视频模型中看到了大量这样的能力。
Now we're seeing a lot of that in our video models actually.
要测试是否具备这种理解能力，一个方法是：你能生成逼真的世界场景吗？
And one way to show how do you test you have that kind of understanding, well, can you generate realistic worlds?
因为如果你能生成它，那么在某种意义上，你必然已经理解了该系统必须封装的世界运行机制。
Because if you can generate it, then in a sense, you must have understood the system must have encapsulated a lot of the mechanics of the world.
这就是为什么我们的视频模型Genie和Vio，以及这类交互式世界模型如此令人印象深刻，同时也是展示我们拥有通用世界模型的重要里程碑。
So that's why Genie and Vio, our video models, and our sort of interactive world models are really impressive, but also important steps towards showing we have generalized world models.
希望未来某天，我们能把这项技术应用到机器人技术和通用辅助领域。
And then, hopefully, some point, we can apply it to robotics and universal assistance.
当然，我肯定要做的一件事就是在某个时候重新应用它回到游戏和游戏模拟中，创造终极游戏，这当然可能一直是我的潜意识计划。
And then, of course, one of my favorite things I'm definitely gonna have to do at some point is reapplying it back to games and and, you know, game simulations and create the ultimate games, which, of course, was maybe always my subconscious plan.
这一切。
All of this.
是的。
Yeah.
一直以来。
All of this time.
没错。
Exactly.
那科学领域呢？
What about science too, though?
你能在那个领域应用它吗？
Could you use it in that domain?
可以。
Yes.
确实可以。
You could.
因此要构建科学复杂领域的模型，无论是原子级别的材料与生物学，还是像天气这样的物理现象。
So building models of scientifically complex domains, whether that's materials on atomic level and biology, but also like some physical things as well, like weather.
理解这些系统的一种方法是从原始数据中学习这些系统的模拟。
One way to understand those systems is to learn simulations of those systems from the raw data.
假设你有一堆原始数据。
So you have a bunch of raw data.
比如说天气数据——显然我们现在正在进行一些很棒的气象项目。
Let's say it's about the weather, and obviously, we have some amazing weather projects going on.
然后让模型学习这些动态规律，并能比暴力计算更高效地重现这些动态。
And then you have a model that kinda learns those dynamics and can recreate those dynamics more efficiently than doing it by brute force.
所以我认为模拟系统和世界模型——特别是针对科学与数学某些方面的专用模型——潜力巨大。
So I think there's huge potential for simulations and kind of world models, maybe specialized ones for aspects of science and mathematics.
不过话说回来，你还可以在模拟世界里投放智能体呢。
But then also, I mean, you can drop an agent into that simulator world too.
对吧？
Right?
是的。
Yes.
你的Genie团队有句非常棒的名言。
Your Genie team goes this really amazing quote.
他们说，几乎所有重大发明的前提条件都不是为了那个发明而创造的。
They said, almost no prerequisite to any major invention was made with that invention in mind.
他们讨论的是将智能体放入这些模拟环境中，让它们以好奇心为主要驱动力进行探索。
And they were talking about dropping agents into these simulated and allowing them to explore with sort of curiosity being their main motivator.
没错。
Right.
所以这是世界模型的另一个激动人心的应用——我们还有个叫Sima的项目。
And so that's another really exciting use of these world models is you can we have another project called Sima.
我们刚发布了Sima二代，就是模拟智能体，你可以把一个虚拟形象或智能体放入虚拟世界中。
We just released Sima two, sim you know, simulated agents, where you have an avatar or an agent, and you put it down into a virtual world.
它可以是某种实际的商业游戏，比如《无人深空》这类非常复杂的开放世界太空游戏。
It can be a kind of actual commercial game or something like that, a very complex one, like no man's sky, a kind of open world space game.
然后你可以指挥它，因为它的底层搭载了Gemini。
And then you can instruct it because it's got Gemini under the hood.
你可以直接与智能体对话并分配任务。
You can just talk to the agent and give it tasks.
但我们当时想，如果把Genie接入Sima，让Sima智能体进入另一个实时生成世界的AI会怎样？
But then we thought, well, wouldn't it be fun if we plug Genie into Sima and sort of drop a Sima agent into another AI that was creating the world on the fly?
现在两个AI在彼此的思维中进行着某种互动。
So now the two AIs are kinda interacting in the minds of each other.
Sima智能体试图在这个世界导航，而对Genie来说那只是个玩家角色。
So the simmer agent's trying to navigate this world, and Genie is as far as Genie's concerned, that's just a player.
作为虚拟形象，它根本不在乎对方是另一个AI。
And an avatar doesn't care there's another AI.
所以它只是根据Sima的行为实时生成周围的世界。
So it's just generating the world around whatever simmer's trying to do.
所以看到它们这样互动真的很神奇。
So that it's kind of amazing to see them both interacting together.
我认为这可能是一个有趣训练循环的开始，你几乎拥有无限的训练样本，因为无论模拟智能体想学习什么，Genie都能即时生成那个世界。
And I think this could be the beginning of an interesting training loop where you almost have infinite training examples because whatever the simmer agent's trying to learn, Genie can basically create on the fly that world.
所以你可以想象整个任务设定与解决的世界，自动生成数百万个任务，而且难度会不断提升。
So I think that you could imagine a whole world of, like, setting and solving tasks, just millions of tasks automatically, and they're just getting increasingly more difficult.
我们可能会尝试建立这样的循环系统，当然这些模拟智能体也能成为很棒的游戏伙伴，它们学到的某些东西对机器人技术也有价值。
So we might try to set up a kind of loop like that, as well as, obviously, those simmer agents could be great as game companions or also some of the things that they learn could be useful also for robotics.
是啊。
Yeah.
基本上就是无聊NPC的终结。
The end of boring NPCs, basically.
没错。
Exactly.
这对游戏来说将会非常棒。
It's gonna be amazing for these games.
是啊。
Yeah.
不过你创造的那些世界，如何确保它们真的逼真呢？
Those worlds that you're creating, though, how do you make sure that they really are realistic?
我的意思是，怎么保证不会出现看起来合理但实际上错误的物理效果？
I mean, how do you ensure that you don't end up with physics that looks plausible but is actually wrong?
没错。
Yeah.
这是个好问题，也确实可能是个隐患。
That's a great question and can be an issue.
本质上这又回到了幻觉问题。
There's basically hallucinations again.
有些幻觉是好的，因为这也意味着你可能创造出有趣的新事物。
So some hallucinations are good because you it also means you might create something interesting and new.
事实上，当你尝试做创意性工作或让系统创造新事物时，适量的幻觉可能反而是有益的。
So in fact, sometimes if you're trying to do creative things or trying to get your system to create new things, novel things, a bit of hallucination might be good.
但你希望这是有意为之的。
But you want it to be intentional.
所以你现在某种程度上是开启了幻觉模式，或者说创意探索模式。
So you kinda switch on the hallucinations now, right, or the the creative exploration.
但当你试图训练一个模拟智能体时，你肯定不希望Genie产生错误的物理幻觉。
But, yes, when you're trying to train a simmer agent, you don't want Genie hallucinating kind of physics that are wrong.
实际上，我们现在正在做的是建立一个物理基准测试，利用游戏引擎精确的物理特性，创建大量类似物理实验课的简单场景。
So, actually, what we're doing now is we're almost creating a physics benchmark where we can use game engines, are very accurate with physics to create lots of, like, fairly simple, like, the sorts of things you would do in your physics a level lab lessons.
对吧？
Right?
比如让小球从不同轨道滚下，观察它们的速度变化。
Like, you know, rolling little balls down different tracks and seeing how fast they go.
从根本上剖析牛顿运动三定律是否被准确掌握——无论是Vio还是Genie模型，是否100%精确地理解了这些物理原理。
And so, like, really teasing apart on a very basic level, like Newton's three laws of motion, has it encapsulated it, whether that's Vio or Genie, have these models encapsulated the physics of that a 100% accurately.
而目前它们还没有做到。
And right now, they're not.
它们目前还只是近似模拟，乍看之下很逼真，但精度尚不足以应用于机器人等领域。
They're kind of approximations, and they look realistic when you just casually look at them, but they're not accurate enough yet to rely on, for say, robotics.
所以这是我们的下一步目标。
So that's the next step.
我认为现在我们已经拥有了这些非常有趣的模型。
So I think now we've got these really interesting models.
就像我们所有模型的研究方向一样，重点在于减少幻觉现象并增强其现实基础。
I think one of the things just like we're trying with all of our models is to reduce the hallucinations and make them even more grounded.
在物理模拟方面，可能需要生成海量基础真实的简单视频，比如钟摆运动。
And with physics, I think that's gonna probably involve generating loads and loads of ground truth, simple videos of pendulums.
比如两个相互环绕的钟摆会发生什么？
You know, what happens when two pendulums go around each other?
但很快你就会遇到三体问题这类本就无解的难题。
But then very quickly, you get to, like, three body problems, which are not solvable anyway.
所以我觉得这会是个很有意思的研究方向。
So I think it's gonna be interesting.
但已经令人惊叹的是，当你观察像Vio这样的视频模型处理反射和液体的方式时，它已经精确得难以置信，至少肉眼看来如此。
But what's amazing already is when you look at the video models like Vio and just the way it treats reflections and liquids, it's pretty unbelievably accurate already, at least to the naked eye.
所以下一步实际上是要超越人类业余爱好者能感知的范围。
So the next step is actually going beyond what a human amateur can perceive.
它能真正经受住正规物理实验的检验吗？
And would it really hold up to a proper physics grade experiment?
我知道你思考这些模拟世界已经很久了。
I know you've been thinking about these simulated worlds for a really long time.
我回顾了我们第一次访谈的文字记录。
And I went back to the transcript of our first interview.
在里面你说过，你非常认同意识是进化产物的理论。
And in it, said that you really liked the theory that consciousness was this consequence of evolution.
嗯。
Mhmm.
你知道，在我们进化史的某个阶段，理解他人的内在状态确实具有生存优势。是的。
That, you know, at some point in our evolutionary past, there was like an advantage to understanding the internal state of another and Yes.
然后我们某种程度上将其内化了。
Then we sort of turned it in ourselves.
是的。
Yes.
这会让你对在模拟环境中运行某种进化主体感到好奇吗？
Does that make you curious about running sort of an agent in evolution inside of a simulation?
我是说，我很想找个时间做这个实验。
I mean, I love to run that experiment at some point.
我们运行进化模拟，也几乎运行社会动力学模拟。
We run evolution, we run almost social dynamics as well.
就像圣塔菲研究所过去常在小型网格世界上运行很多酷炫实验。
Like, the Santa Fe used to run lots of cool experiments on little grid worlds.
我以前很喜欢其中一些实验，但主要都是经济学家在做。
I used to love some of these, but they're mostly economists.
他们试图运行小型人工社会，结果发现各种有趣的事物就这样被创造出来了。
And they were trying to, like, you know, run, like, little artificial societies, and they found that things all sorts of interesting things got invented like that.
如果你让智能体在合适的激励机制、市场、银行等各种疯狂环境中运行足够长时间
If you let agents run around for long enough with the right incentive structures, markets, and banks, and all sorts of crazy things.
所以我认为这将非常酷，而且能帮助我们理解生命起源和意识起源
So I think it would be really cool and also just to understand the origin of life and the origin of consciousness.
我认为这正是我最初从事人工智能工作的主要热情之一——你需要这类工具才能真正理解我们从何而来以及这些现象的本质
And I think that is one of the big passions I had for working on AI from the beginning was, I think you're gonna need these kinds of tools to really understand where we came from and what these phenomena are.
而我认为模拟是最强大的工具之一，因为你可以进行统计性研究
And I think simulations is one of the most powerful tools to do that because you can then do it statistically.
因为你可以在略微不同的初始条件下多次运行模拟，也许运行数百万次，然后以高度受控的实验方式理解这些微小差异——当然，对于我们想解答的那些真正有趣的问题，在现实世界中很难做到这一点。
Because you can run the simulation many times with slightly different initial starting conditions, maybe run it millions of times, and then understand what the slight differences are in a very controlled experiment sort of way, which, of course, is very difficult to do in the real world for any of the really interesting questions we wanna answer.
所以我认为精确的模拟将成为科学的巨大福音。
So I think accurate simulations will be an unbelievable boon to science.
考虑到我们已经发现的这些模型涌现出的特性——那些我们未曾预料到的概念性理解——在运行这类模拟时是否也需要格外谨慎？
Given, you know, what we've discovered about sort of emergent properties of these models, right, having sort of conceptual understanding that we weren't expecting, Do you also have to be quite careful about running this or simulation?
我认为确实需要谨慎，没错。
I think you would have to be, yes.
但这就是模拟的另一大优点。
But that's the other nice thing about simulations.
你可以在相当安全的沙盒环境中运行它们，或许最终你会想要物理隔离它们。
You can run them in pretty safe sandboxes, maybe eventually you wanna air gap them.
当然你可以全天候监控模拟中的情况。
And you can of course monitor what's happening in the simulation twenty four seven.
而且你能获取所有数据。
And you have access to all the data.
所以我们可能需要AI工具来协助监控模拟，因为它们会变得极其复杂，内部会发生大量事件。
So we may need AI tools to help us monitor the simulations because they'll be so complex, and there'll be so much going on in them.
想象一下大量AI在模拟环境中运行，任何人类科学家都难以跟上节奏，但我们或许可以利用其他AI系统来自动分析和标记这些模拟中有趣或值得关注的内容。
If you imagine loads of AIs running around in a simulation, it will be hard for any human scientist to keep up with it, but we could probably use other AI systems to help us analyze and flag anything interesting or worrying in those simulations automatically.
我想，这应该算是中长期的规划吧
I mean, this I guess, we're still talking sort of medium to long term
是的。
Yeah.
展开剩余字幕（还有 312 条）
就这方面而言。
In terms of this stuff.
所以回到我们当前的轨迹上。
So just going back to the trajectory that we're on at the moment.
是的。
Yes.
我还想和你聊聊AI和AGI对更广泛社会的影响。
I also wanna talk to you about the impact that AI and AGI are gonna have on wider society.
上次我们交谈时，你说你认为AI在短期内被过度炒作。嗯。
And last time we spoke, you said that you thought AI was overhyped in the short term Mhmm.
但从长期来看却被低估了。
But underhyped in the long term.
我知道今年有很多关于AI泡沫的讨论。
And I know that this year, there's been a lot of chatter about an AI bubble.
是的。
Yes.
我是说，如果真有泡沫并且破裂了会怎样？
I mean, what what happens if there is a bubble and it bursts?
会发生什么？
What happens?
嗯，你看，我认为是的，我仍然认同短期内被高估而中长期潜力仍被低估的观点，你知道它将会带来多么变革性的改变。
Well, look, I think, yes, I still subscribe to it's overhyped in the short term and still underappreciated in the medium to long term, you know, how transformative transformative it's it's gonna gonna be.
的。
Be.
是的。
Yeah.
当然，现在关于AI泡沫的讨论很多。
There is a lot of talk, of course, right now about AI bubbles.
在我看来，这不是非黑即白的事情，我们到底有没有泡沫？
In my view, it's not one thing, binary thing, are we or aren't we?
我认为AI生态系统中有些部分可能确实存在泡沫。
I think there are parts of the AI ecosystem that are probably in bubbles.
一个例子就是那些甚至还没起步的初创公司，在种子轮融资时估值就已经高达数百亿美元。
One example would be just seed rounds for startups that basically haven't even got going yet, and they're raising at tens of billions of dollars valuations just out of the gate.
看到这种情况挺有意思的，这能持续下去吗？
It's sort of interesting to see, can that be sustainable?
你知道，我的猜测可能是不能，至少普遍来说不行。
You know, my guess is probably not, at least not in general.
所以这是其中一个领域。
So there's that area.
然后人们显然也在担心大型科技公司的估值和其他问题。
Then the people are worrying about, obviously, there's the big tech valuations and other things.
我认为这背后有很多真实的业务支撑。
I think there's a lot of real business underlying that.
不过还有待观察。
So but remains to be seen.
我的意思是，对于任何难以置信的变革性深远技术——当然AI可能是最深远的一个——某种程度上你都会看到这种过度修正。
I mean, I think maybe for any new unbelievably transformative and profound technology of which, of course, AI is probably the most profound, you're gonna get this overcorrection in a way.
我们创立DeepMind时，没人相信它能成功。
So when we started DeepMind, no one believed in it.
没人认为这是可能的。
No one thought it was possible.
人们甚至质疑AI到底有什么用。
People were wondering what's AI for anyway.
而现在短短十到十五年后，AI显然已成为商界唯一的热门话题。
And then now fast forward ten, fifteen years, and now, obviously, it seems to be the only thing people talk about in business.
但这种狂热某种程度上是对之前冷遇的过度补偿。
But you're sort of gonna get it's almost an overreaction to the underreaction.
我认为这种现象很自然。
So I think that's natural.
我们在互联网时代就见证过类似情况。
I think we saw that with the Internet.
移动互联网时代也是如此，而现在AI正经历或将再次经历同样的轮回。
I think we saw it with mobile, and I think we're seeing it or going to see again with AI.
我并不太担心我们是否处于泡沫中，因为从我的角度来看，领导Google DeepMind以及整个Alphabet旗下的谷歌，我们的职责就是确保无论情况如何，我们都能以非常强大的姿态走出来。
I don't worry too much about are we in a bubble or not because from my perspective, leading Google DeepMind and also, obviously, with Google as a as an Alphabet as a whole, our job and my job is to make sure either way, we come out of it very strong.
我认为我们处于非常有利的位置，无论怎样，我们都具备极其优越的竞争态势。
And I think and we're very well positioned, and I think we are tremendously well positioned either way.
如果按照现在的趋势继续发展下去，那就太棒了。
So if it continues going like it is now, fantastic.
我们将继续推进，你知道的，所有这些我们正在做的伟大工作、实验以及向通用人工智能迈进的进展。
We'll carry on, you know, all of these great things that we're doing and experiments and progress towards AGI.
如果有收缩调整，那也没关系。
If there's a retrenchment, fine.
而且我认为我们处于非常有利的位置，因为我们拥有自己的TPU技术栈。
Then, also, I think we're in a great position because we have our own stack with TPUs.
我们还拥有所有这些出色的谷歌产品及其创造的利润，可以将我们的AI技术融入其中。
We also have all these incredible Google products and the profits that all makes to plug in our AI into.
我们正在通过AI彻底革新搜索，AI模式下内置Gemini技术提供AI概览功能。
And we're doing that with search is totally revolutionized by AI overviews, AI mode with Gemini under the hood.
我们正在关注工作空间、电子邮件，还有YouTube。
We're looking at workspace, at email, at, you know, at YouTube.
Chrome浏览器中有所有这些令人惊叹的功能。
So there's all these amazing things in Chrome.
AI已经能看到许多唾手可得的应用场景，比如Gemini二代，当然还有表现同样出色的Gemini应用，以及通用助手的理念。
There's a lot of these amazing things that AI, can see already are low hanging fruit to apply Gemini two, as well, of course, as Gemini app, which is doing really well as well now and the, you know, idea of universal assistant.
这些都是新产品，我认为假以时日它们会极具价值，但我们不必依赖于此。
So there's new products, and I think they will, in the fullness of time, be super valuable, but we don't have to rely on that.
我们只需强化现有生态系统。
We can just power up our existing ecosystem.
我认为这就是过去一年发生的变化。
I think that's what's happened over the last year.
我们现在已经实现了高效运作。
We've got that really efficient now.
关于当前人们可接触的AI技术，您最近提到避免为最大化用户参与度而开发AI至关重要，这样我们就不会重蹈社交媒体的覆辙。
In terms of the AI that people have access to at the moment, I know you said recently how important it is not to build AI to maximize user engagement, just so we don't repeat the mistakes of social media.
但我也在想，我们是否已经在某种程度上看到了这种现象。
But I also wonder whether we are already seeing this in a way.
我是说，人们花大量时间与聊天机器人对话，结果却陷入自我激进的漩涡。
I mean, people spending so much time talking to their chatbots that they end up kind of spiraling into self radicalizing.
是啊。
Yeah.
你如何阻止这种情况发生？
How do you stop that?
如何构建以用户为中心的AI？这本质上是很多方面的初衷，但又要避免形成个人回音室？
How do you build AI that puts users at the center of their own universe, which is sort of the point of this in a lot of ways, but without creating echo chambers of one?
没错。
Yeah.
这是个需要非常谨慎把握的平衡点，我认为这是我们行业必须解决的最重要问题之一。
It's a very, you know, careful balance that, you know, I think is one of the most important things that we as an industry have gotta get right.
我们已经看到某些过度谄媚的系统会造成什么后果——那种回音室式的自我强化对个人非常有害。
So I think we've seen what happens with, you know, some systems that were overly sycophantic or, you know, then you get these sort of echo chamber reinforcements that are really bad for the person.
因此我认为，这其中的一部分实际上正是我们希望通过Gemini构建的。我对Gemini第三代人格特别满意，我们有一个很棒的团队在研发，我个人也参与其中——它就像一种科学家的性格特质，温暖、乐于助人、轻松愉快，但又能简明扼要地表达观点，并且会以友好的方式反驳那些不合理的主张，而不是一味附和用户。比如当有人说地球是平的时，它不会附和说'这想法真棒'。
So I think part of it is and actually, is what we want to build with Gemini, and I am really pleased with the Gemini three persona that we had a great team working on and I helped with too personally is just this sort of almost like a scientific personality that's warm, it's helpful, it's light, but it's succinct to the point, and it will push back on things in a friendly way that don't make sense rather than trying to reinforce you, you know, the idea that the earth's flat and you said it, and it's like wonderful idea.
我认为如果这种情况普遍发生，对社会整体而言并非好事。
I don't think that's good in general for society if that were to happen.
但你必须平衡用户的需求，因为人们希望这些系统能支持他们的想法，协助他们进行头脑风暴。
But you gotta balance it with what people want because people want these systems to be supportive, helpful with their ideas and their brainstorming.
所以必须把握好这个平衡点。
So you've gotta get that balance right.
我认为我们正在发展一套关于人格与角色设定的科学方法论，比如如何衡量它的表现，以及我们希望它在真实性、幽默感等方面的定位。
And I think we're sort of developing a science of personality and persona of, like, how to kinda measure what it's doing and where do we want it to be, like, on authenticity, on humor, you know, these sorts of things.
可以想象系统会内置一个基础人格模板。
And then you can imagine there's a kind of base personality that it ships with.
然后每个用户都可以根据自己的偏好进行定制。
And then everyone has their own preferences.
你明白我的意思吗？
You know?
你希望它更幽默、少些幽默、更简洁还是更详尽？
Do you want it to be more humorous, less humorous, or more succinct, or more verbose?
不同的人会有不同的偏好。
People would like different things.
所以你在核心基础人格之上还要添加这层个性化定制，但每个人最初获得的核心人格是相同的，对吧？
So you add that additional personalization layer on it as well, but there's still the core base personality that everyone gets, right?
这个核心始终遵循科学方法，这正是这些系统的根本目的。
Which is trying to adhere to the scientific method, which is the whole point of these.
我们希望人们将其应用于科学、医疗健康等领域。
And we want people to use these for science and for medicine and health issues and so on.
我认为这是完善大型语言模型科学的一部分。
And so I think it's part of the science of getting these large language models right.
我对我们当前的发展方向相当满意。
And I'm quite happy with the direction we're going in currently.
我们得专门和Shane Legg聊聊通用人工智能（AGI）。
We got to talk to Shane Legg about AGI in particular.
在当前AI领域发生的所有事情中，语言模型、世界模型等等，哪项最接近你对通用人工智能（AGI）的愿景？
Across everything that's happening in AI at the moment, the language models, the world models, you know, and so on, what's closest to your vision of AGI?
我认为实际上是Gemini 3（显然它非常强大）与我们上周发布的Nano Banana Pro系统的结合，这是我们图像创作工具的高级版本。
I think actually the combination of, obviously there's Gemini three, which I think is very capable, but the Nano Banana Pro system we also launched last week, which is an advanced version of our image creation tool.
最令人惊叹的是，它的底层也采用了Gemini技术。
What's really amazing about that, it has also Gemini under the hood.
所以它不仅能理解图像。
So it can understand not just images.
某种程度上它能理解这些图像中的语义内容。
It sort of understands what's going on semantically in those images.
人们才使用了一周，但我已经在社交媒体上看到很多用它创作的酷炫作品。
And people have been only playing with it for a week now, but I've seen so much cool stuff on social media about what people are using it for.
比如你可以给它一张复杂飞机图片，它能标注出飞机所有部件的示意图，甚至能将所有部件分解展示出来。
So for example, you know, you can give it a picture of a complex plane or something like that, and it can label all the diagrams of, you know, all the different parts of the plane and even visualize it with all the different parts sort of exposed.
这说明它对机械结构、物体组成部件和材料有深层次理解，而且现在文本渲染也极其精准。
So it has some kind of deep understanding of mechanics and what, you know, makes up parts of objects, what's materials, and it can, you know, render text really, really accurately now.
所以我认为这正在接近某种图像领域的通用人工智能。
So I think that's getting towards a kinda AGI for imaging.
我认为这是一种通用系统，可以处理任何图像相关任务。
I think it's a kinda general purpose system that can do anything across images.
所以我觉得这非常令人兴奋。
So I think that's very exciting.
还有世界模型方面的进展，比如Genie和Simmer项目以及我们的相关工作。
And then the advances in in world models, you know, Genie and Simmer and what we're doing there.
最终我们需要把这些技术都融合起来。
And then eventually, we gotta kinda converge all of those.
目前它们还是不同的项目，虽然相互关联，但我们需要将它们整合成一个大型模型。
They're kinda different projects at the moment, and they're intertwined, but we need to, you know, converge them all into one big model.
那样可能就会开始成为原始通用人工智能的候选方案。
And then that might start becoming, you know, candidate for proto AGI.
我知道你最近读了很多关于工业革命的资料。
I know you've been reading quite a lot about the industrial revolution recently.
我们能否从历史中汲取经验，以缓解预期将出现的某些颠覆性冲击？
Are there things that we can learn from what happened there to try and mitigate against the sort of some of the disruption that we can expect
作为AGI，我认为我们有很多可以学习的地方。
I as AGI think there's a lot we can learn.
这在学校里学过，至少在英国是这样，但学得很肤浅。
It's something you sort of study in school, at least in Britain, but on a very superficial level.
对我来说，深入研究整个工业革命如何发生、始于什么、背后的经济动因（比如纺织业），以及最早的计算机其实是缝纫机这些历史，真的很有趣。
Like, it was really interesting for me to look into how it all happened, what it started with, the economic reasons behind that, which is like the textile industry, and then the first computers were really the sewing machines.
对吧？
Right?
后来它们演变成了早期Fortran计算机的穿孔卡片，也就是大型机。
And then they became punch cards for the early Fortran computers, mainframes.
有段时间英国因此非常成功，成为了纺织业的中心，因为自动化系统能以极低成本生产出精良产品。
And for a while, it was very successful in Britain, became, like, the center of the textile world because they could make these amazingly high quality things for very cheap because of the automated systems.
再后来，蒸汽机之类的东西当然就出现了。
And then, obviously, the steam engines and all of those things came in.
我认为工业革命带来了许多惊人的进步。
I think there's a lot of incredible advances that came out of industrial revolution.
所以儿童死亡率下降了，现代医学和卫生条件都出现了，甚至工作与生活的分离以及所有这些是如何运作的，都是在工业革命期间解决的。
So child mortality went down and all of modern medicine and sanitary conditions, even the kind of work life split and how that all worked was kind of worked out during the industrial revolution.
但这也带来了许多挑战。
But it also came with a lot of challenges.
比如，它花费了相当长的时间，大约一个世纪，而且劳动力不同部分在不同时期都受到了冲击。
Like in it took quite a long time, roughly a century, and different parts of the labor force were dislocated at certain times.
于是为了重新平衡这种局面，不得不创建工会等新型组织。
And then new organizations like unions and other things had to be created in order to rebalance that.
所以，看着整个社会如何逐渐适应，最终形成现代世界的样子，这非常引人入胜。
So, like, it was fascinating to see the whole of society sort of had to over time adapt, and then you've got the modern world now.
显然工业革命利弊并存，但若纵观其整体成果——比如西方世界食物丰足、现代医学、现代交通等——这些都源于工业革命，没人愿意回到前工业革命时代。
So I think there were lots of, obviously, pros and cons of the industrial revolution why it was happening, but no one would want if you think about what it's done in total, like abundance of food and thing in in the Western world and modern medicine and all these things, modern transport, that was all because of the industrial revolution.
因此我们虽不愿退回前工业时代，但或许能通过借鉴历史，提前预判社会断层，并更早或更有效地缓解这次转型的阵痛。
So we wouldn't wanna go back to pre industrial revolution, but maybe we can figure out ahead of time by learning from it what those dislocations were and maybe mitigate those earlier or more effectively this time.
我们可能不得不这样做，因为这次的不同之处在于，它可能比工业革命还要大十倍，而且发生的速度可能快十倍。
And we're probably gonna have to because the difference this time is that it's probably gonna be 10 times bigger than industrial revolution, it'll and probably happen 10 times faster.
所以更像是十年内完成，而非一个世纪那么漫长。
So more like a decade than unfold over a decade than a century.
Shane告诉我们的一件事是，当前这种你通过劳动换取资源的经济体系，在后AGI社会中将无法以同样方式运作。
One of the things that Shane told us was that the kind of current economic system where, you know, you exchange your labor for resources effectively, It just won't function the same way in a post AGI society.
你对社会应该如何重构或可能以何种方式重构才能有效运作有设想吗？
Do you have a vision of how society should be reconfigured or might be reconfigured in a way that works?
是的，我现在花更多时间思考这个问题，Shane实际上正在领导一个相关项目，研究后AGI世界可能的面貌以及我们需要做的准备。
Yeah, I'm spending more time thinking about this now and Shane's actually leading an effort here on that to sort of think about what a post AGI world might look like and what we need to prepare for.
但我认为整个社会——经济学家、社会科学家和政府——都需要花更多时间思考这个问题。
But I think society in general needs to spend more time thinking about that, economists and social scientists and governments.
就像工业革命时期那样，整个工作世界和工作周制度都从农业主导的前工业革命时代发生了改变。
As with the industrial revolution, the whole working world and working week and everything got changed from from pre industrial revolution more to agriculture.
我认为至少那种程度的变革将再次发生。
And I think that's gonna at least that level of change is gonna happen again.
因此，如果需要新的经济体系和模式来协助这一转型，并确保利益得到广泛分配，我一点也不会感到惊讶。
So I would not be surprised if we needed new economic systems, new economic models to help with that transformation and make sure, for example, the benefits are widely distributed.
也许像全民基本收入这样的措施会是解决方案的一部分，但我认为这并不完整，这只是我们目前能设想到的。
And maybe things like universal basic income and things like that are part of the solution, but I don't think that's the complete I think that's just what we can model out now.
对吧？
Right?
因为这几乎只是对现有体系的补充。
Because that would be a almost an add on to what we have today.
但我认为可能存在更好的系统，比如更直接的民主制度，你可以用某种积分来投票决定想看到的变化。
But I think there might be something way better systems, more like direct democracy type systems where you can, you know, vote with a certain amount of credits or something for what you want to see.
这实际上已经在社区层面实施了。
It happens actually on local community level.
比如，这里有笔资金。
You know, here's a bunch of money.
你们是想要个游乐场、网球场，还是给学校加建一间教室？
Do you want a playground or tennis court or an extra classroom on the school?
然后你让社区进行投票决定。
And then you let the community sort of vote for it.
之后你或许还能衡量这些决策的成效。
And then maybe you could even measure the outcomes.
那些总能投票选出更受欢迎方案的人，在下次投票中将获得相应更大的影响力。
And then the people that could sort of consistently vote for things that end up being more well received, they have proportionally more influence for the next vote.
我听到很多有趣的想法，比如我那些经济学家朋友正在头脑风暴这个议题。
So there's there's a lot of interesting things I hear, you know, economist friends of mine who are brainstorming this.
我认为如果能有更多这方面的研究就太好了。
And I think that will be great if we had a lot more work on that.
然后这还涉及到哲学层面的思考，比如：
And then there's the philosophical side of it of like, okay.
工作岗位会发生变化等等，但也许那时我们已经攻克了核聚变技术。
So jobs will change and other things like that, but maybe we'll have fusion will have been solved.
这样我们就能拥有近乎无限的免费能源。
And so we have this sort of abundant free energy.
那么我们进入了后稀缺时代。
So we're post scarcity.
钱会变成什么样呢？
So what happens to money?
也许每个人的生活都变得更好了，但人生的意义又会如何变化？
Maybe everyone's better off, but then what happens to purpose?
对吧？
Right?
因为许多人从工作中获得人生目标，并通过工作养家糊口，这是非常崇高的追求。
Because a lot of people get their purpose from their jobs and then providing for their families, which is a very noble purpose.
我认为这些问题有些已从经济问题逐渐演变成了近乎哲学的问题。
I think some of these questions blend from economic questions into almost philosophical questions.
你是否担心人们似乎没有给予足够关注，行动速度也不如你期望的那样快？
Do you worry that people don't seem to be paying attention, sort of moving as quickly as you'd like to see?
确实。
Yeah.
我是说，要怎样才能让人们认识到这类议题需要国际合作呢？
Mean, what would it take for people to sort of recognise that we need international collaboration on this kind of topic?
我确实对此感到担忧。
I am worried about that.
在理想情况下，本应已有更多国际合作——特别是跨国层面的——以及更多关于这些议题的研究、探索和讨论正在进行。
And again, in a sort of ideal world, there would have been a lot more collaboration already and international specifically and a lot more research and sort of exploration and discussion going on about these topics.
实际上我挺惊讶这类合作竟然如此之少。
I'm actually pretty surprised there isn't more of that.
即便按我们的时间线——虽然外界有些预测非常紧迫——但我们的预估也是五到十年，这对建立应对机制的相关机构来说并不算长。
Even our timelines, which were there were some very short timelines out there, but even ours are five to ten years, which is not long for institutions or things like that to be built to handle this.
我担心的一个问题是，现有机构似乎非常分散，且影响力远未达到所需水平。
And one of the worries I have is that the institutions that do exist seem to be very fragmented and not very influential to the level that you would need.
所以现状可能是：目前还没有合适的机构能处理这个问题。
So it may be that there aren't the right institutions to deal with this currently.
当然，如果再叠加当前全球地缘政治紧张局势，协作似乎变得比以往任何时候都更困难。
And then, of course, if you add in the geopolitical tensions that are going on at the moment around the world, it seems like collaboration, cooperation is harder than ever.
就拿气候变化来说，要达成任何相关协议都困难重重。
Just look at climate change and how hard it is to get any agreement on anything to do with that.
我们拭目以待吧。
So we'll see.
我认为随着风险越来越高，这些系统越来越强大——也许将它们融入产品的一个好处就是，那些不从事这项技术工作的普通人也能感受到它们能力和威力的增长。
I think as the stakes get higher and as these systems get more powerful, and maybe this is one of the benefits of them being in products is that, you know, everyday person that's not working on this technology will get to feel the increase in the power of these things and the capability.
这样就会引起政府重视，或许随着我们接近通用人工智能，他们会醒悟过来。
And so that will then reach government and then maybe they'll see sense as we get closer to AGI.
你认为是否需要某个时刻、某个事件才能让大家真正重视起来？
Do you think it will take a moment, an incident for everyone to sort of sit up and pay attention?
我不知道。
I don't know.
我是说，希望不会。
I mean, I hope not.
大多数主要实验室都相当负责任。
Most of the main labs are pretty responsible.
我们尽可能做到负责任。
We try to be as responsible as possible.
你知道，这一直是我们工作的核心，如果你多年来一直关注我们就会明白。
You know, that's always something we've as you know, if you followed us over the years, that's been at the heart of what everything we do.
这并不意味着我们能把每件事都做对，但我们会尽可能保持审慎和科学的态度。
Doesn't mean we'll get everything right, but we try to be as thoughtful and as scientific in our approach as possible.
我认为大多数主要实验室都在努力做到负责任。
I think most of the major labs are are trying to be responsible.
此外，实际上存在良好的商业压力促使企业保持负责任的态度。
Also, there's good commercial pressure actually to be responsible.
试想一下，如果你在向另一家公司出租智能体来完成某项任务，对方公司必然会想知道这些智能体的限制、边界和安全防护措施，也就是它们能做什么、不会搞乱数据等等。
If you think about agents and you're renting an agent to another company, let's say, to do something, the other company is going to want to know what the limits are and the boundaries are and the guardrails are on those agents, you know, in terms of what they might do and not just mess up the data and all of this stuff.
所以我认为这是好事，因为那些行事鲁莽的运营商将无法获得业务，企业不会选择他们。
So I think that's good because the more cowboy operations, they won't get the business because enterprises won't choose them.
因此我认为这种资本主义体系实际上将有助于强化负责任的行为，这是件好事。
So I think the kind of capitalist system will actually be useful here to reinforce responsible behavior, which is good.
但总会有一些流氓行为者，可能是流氓国家，可能是流氓组织，也可能是基于开源项目进行开发的人。
But then there'll be rogue actors, maybe rogue nations, maybe rogue organizations, maybe people building on top of open source.
我不知道。
I don't know.
显然，要阻止这种情况非常困难。
Like, obviously, it's very difficult to stop that.
然后可能会出问题，希望只是中等规模的，那将成为对全人类的一记警钟。
Then something may go wrong, and hopefully, it's just sort of medium sized, and then that will be a kind of warning shot to humanity across the bow.
那可能正是推动国际标准或国际合作的最佳时机，至少在高层基础标准上达成共识——我们想要并同意哪些基本标准。
And then that might be the moment to kind of advocate for international standards or international cooperation or collaboration, at least on some high level basic or, you know, kind of like, what's the basic standards we would want and agree to.
我希望这是有可能实现的。
I'm hopeful that that will be possible.
从长远来看，超越通用人工智能迈向超级人工智能的阶段，你认为是否存在一些人类能做而机器永远无法做到的事情？
In the long term, so beyond AGI and and towards ASI, right, artificial superintelligence, do you think that there are some things that humans can do that machines will never be able to manage?
嗯，我认为这是个重大问题。
Well, I think it's the big question.
我觉得这与我最喜欢的话题之一——图灵机有关。
And I feel like this is related to, as you know, one of my favorite topics is Turing machines.
我一直认为，如果我们构建出AGI，再回到之前的模拟讨论，用它作为心智的模拟，然后与真实心智进行对比，我们就能看出差异所在，以及人类心智可能保留的特殊之处。
And I've always felt this that if we build AGI and then almost talking back about our simulation discussion and then use that as a simulation of the mind and then compare that to the real mind, we will then see what the differences are and potentially what special and remaining about the human mind.
也许是创造力。
Maybe that's creativity.
也许是情感。
Maybe it's emotions.
也许是梦境、意识。
Maybe it's dreaming, consciousness.
关于哪些东西可计算或不可计算，目前存在很多假设。
There's a lot of hypotheses out there about what may or may not be computable.
这就引出了图灵机的核心问题：图灵机的极限是什么？
And this comes out to the Turing machine question of, like, what is the limit of a Turing machine?
我认为这确实是我生命中的核心问题，自从我了解图灵和图灵机以来就一直如此。
And I think that's the central question of my life, really, ever since I found out about Turing and Turing machines.
我对此深深着迷。
I fell in love with that.
这是我的核心热情所在。
That's my core passion.
我认为我们所做的一切都在某种程度上将图灵机的概念推向极限，包括蛋白质折叠。
And I think everything we've been doing is being sort of pushing the notion of what a Turing machine can do to the limit, including, you know, folding proteins.
结果发现我并不确定这个极限在哪里。
And so it turns out I'm not sure what the limit is.
也许根本不存在极限。
Maybe there isn't one.
当然，我研究量子计算的朋友们会说存在极限，认为你需要量子计算机来处理量子系统，但我真的不太确定。
And, of course, my quantum computing friends would say there are limits, and you need quantum computers to do quantum systems, but I'm really not so sure.
我实际上和一些量子领域的专家讨论过这个问题。
And I've actually discussed that with some of the quantum folks.
可能需要从这些量子系统中获取数据，才能创建出经典的模拟系统。
And it may be that we need data from these quantum systems in order to create a classical simulation.
这又回到了心灵的问题：这一切都是经典计算，还是有其他因素在起作用？
And then that comes back to the mind, which is, is it all classical computation, or is there something else going on?
就像罗杰·彭罗斯所相信的那样，大脑中存在量子效应。
You know, like Roger Penrose believes, you know, there's quantum effects in the brain.
如果确实如此，并且这与意识有关，那么机器永远无法拥有这种特性，至少经典计算机不行。
If there are, and that's what consciousness has to do with, then machines will never have that, at least the classical machines.
我们只能等待量子计算机的到来。
We'll have to wait for quantum computers.
但如果没有限制，那么可能就不存在任何界限。
But if there isn't, then there may not be any limit.
也许在宇宙中，只要以正确的方式看待，一切在计算上都是可处理的，因此图灵机或许能够模拟宇宙中的一切。
Maybe in the universe, everything is computationally tractable if you look at it in the right way, and therefore, Turing machines might be able to model everything in the universe.
目前如果要我猜测的话，我会这么认为。
I'm currently if you were to make me guess, I would guess that.
我正基于这一假设工作，直到物理学证明我错了为止。
And I'm working on that basis until physics shows me otherwise.
所以在这类计算范畴内没有什么事情是无法完成的
So there's nothing that cannot be done within these sort of computational
嗯，没人这么说过
Well, no one's it this way.
迄今为止，还没有人在宇宙中发现任何不可计算的事物
Nobody's found anything in the universe that's non computable so far.
目前为止
So far.
对吧？
Right?
我认为我们已经证明，你可以远远超越传统复杂性理论家关于P等于NP的观点，即当今经典计算机能做什么，比如蛋白质折叠、围棋等等。
And I think we've already shown you can go way beyond the usual complexity theorist p equals NP view of, like, what a classical computer could do today, things like protein folding and Go and so on.
所以我认为没人知道那个极限是什么。
So I don't think anyone knows what that limit is.
这实际上可以归结为我们在DeepMind和Google所做的事情，而我正在努力做的就是找到那个极限。
And that's really if you were boiled down to what were we doing at DeepMind and Google, and what I'm trying to do is find that limit.
但按照这个想法的极限来看，我们坐在这里，脸上能感受到灯光的温度。
But then in the limit of that though, right, is the in the limit of that idea is that we're sitting here sort of there's like the warmth of the lights on our face.
我们隐约能听到背景中机器的嗡嗡声。
We kind of hear the whir of the machine in the background.
手底下能感受到桌面的触感。
There's like the feel of the desk under our hands.
是的。
Yes.
所有这些都可以被经典计算机复现。
All of that could be replicable by a classical composer.
是的。
Yes.
嗯，我认为归根结底，这也是我热爱康德的原因——我最喜欢的两种哲学是康德和斯宾诺莎，但原因各不相同。
Well, I think in the end, my view on this is why I love Kant as well is all of my two favorite philosophies Kant and Spinoza for different reasons.
但康德认为，现实是心灵的建构物。
But Kant, the reality is a construct of the mind.
我认为这是正确的。
I think that's true.
所以，你提到的所有这些事物，它们进入我们的感官系统，感觉各不相同。
And so, all of those things you mentioned, they're coming into our sensory apparatus, and they feel different.
对吧？
Right?
光线、光的温暖、桌子的触感。
The light, the warmth of the light, the touch of the table.
但归根结底，它们都是信息。
But in the end, they're all it's all information.
我们是信息处理系统，我认为这就是生物学的本质。
And we're information processing systems, and I think that's what biology is.
这就是我们试图通过同构实现的目标。
It's what we're trying to do with isomorphic.
我认为我们最终将治愈所有疾病的方式，就是把生物学视为一个信息处理系统。
That's how I think we'll end up curing all diseases is by thinking about biology as an information processing system.
我认为最终，信息将成为宇宙最基本的单位——不是能量，也不是物质，而是信息。我利用业余时间研究这类物理学理论，比如信息本质论。
And I think in the end, that's gonna be and I'm I'm working on my spare time, my two minutes of spare time, you know, physics theories about things like information being the most fundamental unit, should we say, of the universe, not energy, not matter, but information.
所以最终这些可能都是可以相互转换的，只是我们通过不同方式感知它们。
So it may be that these are all interchangeable in the end, but we just sense it.
我们以不同的方式感受它。
We feel it in a different way.
但就目前所知，我们拥有的这些精妙传感器依然存在。
But, you know, as far as we know, this is still all these amazing sensors that we have.
它们仍然能被图灵机计算。
They're still computable by a Turing machine.
这就是为什么你模拟的世界如此重要。
But this is why your simulated world is so important.
是的。
Yes.
完全正确。
Exactly.
因为这将是实现它的途径之一。
Because that would be one of the ways to get to it.
我们能够模拟的极限是什么？
What's the limits of what we can simulate?
因为如果你能模拟它，那么在某种意义上，你就已经理解了它。
Because if you can simulate it, then in some sense, you've understood it.
我想以一些个人反思来结束，谈谈站在这一领域前沿的感受。
I wanted to finish with some personal reflections of what it's like to be at the forefront of this.
我的意思是，这种情感负担
I mean, does the emotional weight
这个是的。
of this Yes.
是否曾让你感到沉重？
Ever sort of weigh you down?
是否曾让你感到相当孤独？
Does it ever feel quite isolating?
是的。
Yes.
听着，我睡眠很少，部分是因为工作太多，但也有睡眠障碍。
Look, I don't sleep very much, partly because it's too much work, but also I have trouble sleeping.
要处理这些情绪非常复杂，因为它令人难以置信地兴奋。
It's very complex emotions to deal with because it's unbelievably exciting.
你知道，我基本上在做我梦想中的一切事情。
You know, I'm basically doing everything I ever dreamed of.
而且我们在许多方面都处于科学的绝对前沿，包括应用科学和机器学习。
And we're at the absolute frontier of science in so many ways, applied science as well as machine learning.
这令人振奋，所有科学家都知道那种身处前沿并首次发现某事物的感觉。
And that's exhilarating as all scientists know that feeling of being at the frontier and discovering something for the first time.
对我们来说，这几乎每月都在发生，这很神奇。
And that's happening almost on a monthly basis for us, which is amazing.
但当然，我们——肖恩、我以及其他长期从事这项工作的人——比任何人都更清楚即将到来的巨大变革。
But then, of course, we Shane and I and others who've been doing this for a long time, we understand it better than anybody, the enormity of what's coming.
而这件事实际上仍未被充分认识。
And this thing about is still under actually appreciated.
事实上，在未来十年左右的时间里将会发生什么，包括哲学层面的问题——何以为人、什么才是重要的，所有这些议题都将浮现。
In fact, what's gonna happen in more of a ten year time scale, including to things like the philosophical, what it means to be human, what's important about that, all of these questions are gonna come up.
因此这是个重大责任，但我们有支优秀的团队在思考这些问题。
And so it's a big responsibility, but we have an amazing team thinking about these things.
但某种程度上，我想至少对我个人而言，这是我毕生都在准备的事。
But also, it's something, I guess, at least for myself, I've trained for my whole life.
从我早年间下国际象棋，后来研究计算机、游戏、模拟和神经科学开始，一切都在为这样的时刻做准备，而现实与我的预期大致吻合。
So ever since my early days playing chess and then working on computers and games and simulations and neuroscience, It's all been for this kind of moment, and it's roughly what I imagined it was gonna be.
这部分也是我应对压力的方式。
So that's partly how I cope with it.
这就是训练的结果。
It's just training.
不过，有没有某些方面对你的冲击超出了预期？
Are there parts of it that have hit you harder than you expected, though?
是的。
Yes.
确实如此。
For sure.
我是说，甚至在AlphaGo比赛期间，看着我们如何攻克围棋这个美丽的谜题，并改变了它。
On on the way I mean, even the AlphaGo match, right, just seeing how we managed to crack Go, but Go was this beautiful mystery, and it changed it.
所以那既有趣又带着些苦乐参半。
And so that was interesting and kinda bittersweet.
我认为更近期的进展，比如语言和图像处理，对创造力意味着什么——你知道我对创意艺术怀有极大的敬意和热情，我自己也做过游戏设计。
I think even the more recent things of, like, language and then imaging and what does it mean for creativity, I'm you know, have huge respect and passion for the creative arts and having done game design myself.
而且，你知道，我和电影导演们聊过，这对他们来说也是个有趣的转折时刻。
And, you know, I talked to film directors, and it's it's an interesting jewel moment for them too.
首先，他们拥有了这些神奇的工具，能将创意原型开发速度提升十倍。
There's, first, on one hand, they've got these amazing tools that speed up prototyping ideas by 10 x.
但另一方面，这是否会取代某些创造性技能呢？
But on the other hand, is it replacing certain creative skills?
因此我认为这种权衡无处不在，对于像AI这样强大且具有变革性的技术来说，这不可避免——就像过去的电力和互联网一样。
So I think there's sort of these trade offs going on all over the place, which I think is inevitable with a technology as powerful and as transformative as as AI is as in the past electricity was and Internet.
我们已见证这正是人类的故事：我们是制造工具的动物，而这正是我们热爱之事。
And we've seen that that is the story of humanity is we are tool making animals, and that's what we love to do.
出于某种原因，我们的大脑还能理解科学并从事科研，这很神奇，但同时也充满永不满足的好奇心。
And for some reason, we also have brains that can understand science and do science, which is amazing, but also sort of insatiably curious.
我认为这正是人类本质的核心。
I think that's the heart of what it means to be human.
我想我从一开始就感染了这种‘好奇心病毒’。
And I think I've just had that bug from the beginning.
而我试图回应这份好奇的方式，就是构建人工智能。
And my expression of trying to answer that is to build AI.
当你和其他AI领袖共处一室时，你们之间是否存在某种团结感？
When you and the other AI leaders are in a room together, is there sort of sense of solidarity between you?
是否觉得这群人都深知利害关系，真正理解彼此的事业，还是竞争会让你们彼此疏远？
That this is a group of people who all know the stakes, who all really understand their things, or does the competition kind of keep you apart from one another?
嗯，我们都互相认识。
Well, we all know each other.
我和他们中的绝大多数人都相处得不错。
I get on with pretty much all of them.
有些人彼此之间相处不来。
Some of the others don't get on with each other.
而且这确实很难，因为我们可能正处于有史以来最激烈的资本主义竞争中。
And there is it's hard because that we're also in the most ferocious capitalist sort of competition there's ever been, probably.
你知道，我那些经历过互联网泡沫时期的投资人和风投朋友都说，现在的竞争比那时激烈十倍不止。
You know, investor friends of mine and VC friends of mine who were around in the .com era say this is, like, 10 x more ferocious and intense than that was.
从很多方面来说，我很喜欢这种状态。
In many ways, I love that.
我是说，我天生就为竞争而生。
I mean, I I live for competition.
从我下国际象棋的时代起，就一直热爱这种竞争。
I've always loved that since my chess days.
但退一步说，我理解也希望大家都明白，这关乎的远不止公司成功与否这类事情，还有更重大的利害关系。
But stepping back, I understand and I hope everyone understands that there's a much bigger thing at stake than just company successes and, you know, that type of thing.
展望未来十年，当你思考这个问题时，是否有某些即将到来的重大时刻让你个人感到最为忧虑？
When it comes to the next decade, when you think about it, are there big moments coming up that you're personally most apprehensive about?
我认为目前的系统还属于被动型系统。
I think right now, the systems are, you know, I call them passive systems.
作为用户，你需要输入能量——也就是提出问题或任务，然后这些系统会为你提供某种摘要或答案。
You you put the energy in as the user, you know, the question or the what's the task, and then these systems kind of provide you with some summary or some answer.
因此很大程度上，这是由人类主导、输入人类能量和想法的过程。
So very much, it's it's human directed and human energy going in and human ideas going in.
下一阶段是基于代理的系统，我认为现在已初现端倪，虽然还相当原始。
The next stage is agent based systems, which I think we're gonna start seeing we're seeing now, but they're pretty primitive.
比如在未来几年内，我们就会开始看到一些真正令人印象深刻且可靠的系统。
Like, in the next couple of years, I think we'll start seeing some really impressive reliable ones.
我认为这些系统如果被视为助手之类的角色，将会极其有用且强大，但同时它们也会更具自主性。
And I think those will be incredibly useful and capable if you think about them as an assistant or something like that, but, also, they'll be more autonomous.
因此我认为这类系统的风险也会随之增加。
So I think the risks go up as well with those types of systems.
所以我非常担忧这类系统在两三年后可能具备的能力。
So I'm quite worried about what those sorts of systems will be able to do maybe in two, three years' time.
你明白吗？
You know?
因此我们正在开展网络安全防御工作，为可能出现数百万智能体在互联网上漫游的世界做准备。
So we're working on cyber defense in preparation for a world like that where maybe there's millions of agents roaming around on the Internet.
那么你最期待的是什么呢？
And what about what you're most looking forward to?
我是说，是否会有那么一天，你可以功成身退？
I mean, is there a day when you'll be able to retire knowing that your work is done?
还是说这项工作需要耗费不止一生的时间？
Or is there more than a lifetime's worth of work left to do?
是啊。
Yeah.
我一直以来，嗯，确实很需要休假，而且我可能会经常利用这段时间。
I always, well, I could definitely do with sabbatical, and I would spend it doing often.
是啊。
Yeah.
所以能休息一周，哪怕一天也好。
So a week off or even a even a day would be good.
但你看，我的使命始终是帮助人类安全地引导通用人工智能跨越那道门槛。
But look, I think my mission has always been to kind of help the world steward AGI safely over the line for all of humanity.
所以我认为当我们达到那个阶段后，当然还会有超级智能、后AGI时代，以及我们讨论过的所有经济和社会问题。
So I think when we get to that point, of course, then there's superintelligence and there's post AGI and there's all the economic stuff we were discussing and societal stuff.
或许我还能在某些方面提供帮助，但如果说这是我的核心人生使命——其实这只是个小目标——那就算是完成了。
And maybe I can help in some way there, but I think that will be my core part of my life mission will be done if this is more I mean, it's only a small job.
你明白吗？
You know?
就是协助世界跨过那道门槛。
Just get that over the line or help the world get that over the line.
明白吗？
You know?
我认为这需要像我们之前讨论过的那种协作。
I think it's gonna require collaboration like we talked to earlier.
我是个很喜欢合作的人，所以希望以我现在的身份能为这个目标贡献力量。
And I'm quite a collaborative person, so I hope I can help with that from the position that I have.
然后你就能去度假了。
And then you get to have a holiday.
然后我就会...嗯。
And then I'll have the yeah.
没错。
Exactly.
应得的休假。
Well earned sabbatical.
是啊。
Yeah.
德米斯，非常感谢你。
Demis, thank you so much.
谢谢，帕特里克。
Thanks, Patrick.
一如既往。
As always.
好了，这就是本季《咕咕DeepMind》播客的全部内容，我是汉娜·弗莱教授。
Well, that is it for this season of Goo Goo DeepMind the podcast with me, professor Hannah Fry.
记得订阅我们的节目，这样你就能在2026年我们回归时第一时间收到通知。
But be sure to subscribe so you will be among the first to hear about our return in 2026.
在此期间，不妨重温我们丰富的往期节目库吧？
And in the meantime, why not revisit our vast episode library?
因为我们今年涵盖了众多话题，从无人驾驶汽车到机器人技术，从世界模型到药物研发，足够让你忙活一阵子了。
Because we have covered so much this year from driverless cars to robotics, world models to drug discovery, plenty to keep you occupied.
回头见。
See you soon.
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。