谈超级智能与微软–OpenAI 交易

全球 AI 竞赛的赌注越来越高。公司们在各种交易上投入巨资，包括数据中心和人才。据报道，该领域最聪明的人现在可以拿到高达 2 亿美元的薪酬包。微软这家巨头公司的 AI 负责人是穆斯塔法·苏莱曼。他正试图赋予微软 AI 自主能力，同时也在思考下一个前沿——他称之为超级智能。你们会听到他向我解释这一切。但和《米歇尔·侯赛因秀》一贯的风格一样，这次对话既关乎工作，也关乎个人。穆斯塔法·苏莱曼，感谢你来做客。

The stakes are becoming ever higher in the global AI race. Companies spending vast sums on deals of all kinds, including data centers and people. The biggest brains in the field can now reportedly command pay packages of up to $200 million. Leading AI for one of the biggest companies of all, Microsoft, is Mustafa Suleyman. He's now trying to give Microsoft's AI autonomous capability and he's also thinking about the next frontier which he says is superintelligence. You'll hear him explain all of that to me. But as ever on the Michelle Hussein show, this conversation is about the person as well as the work. Mustafa Suleyman, thank you for being with us.

谢谢你邀请我，米歇尔。

Thank you for having me, Michelle.

我想先听听你对这个世界的见解——这是你的日常世界，但对大多数人来说仍在发展中。我想了解，你生活中哪些 AI 用途是我们其他人可能还没用到的？

I'd love to start by getting your insight into a world which is your everyday world but which is still developing for most of us and to understand what are the uses of AI that are in your life that the rest of us might not be using yet.

你问得很有意思，因为昨晚我熬夜看了一部电影，之后我在 Copilot 里创建的一个表格中添加了记录。这个表格基本上记录了我喜欢的电影、我想看的电影，按日期列出，描述它们的内容。我添加个人笔记，它还会给我电影海报的链接。我可以一直说，好的，添加这部我想看的电影。有什么类似的电影推荐吗？就像你让助手帮你整理生活一样。我认为人们会感到惊讶：你打算问 AI 或 Copilot 的任务越冷门、越有创意、越有挑战性，效果就越好。

It's funny you ask because yesterday night I stayed up far too late watching a film and afterwards I added to a table I've made in Copilot which basically records all the films that I love, films that I want to watch, lists them by date, describes what they were about. I add my personal notes to it, it gives me a link to the film poster. And I can keep saying, okay, add this film that I want to watch. What would be a similar one that I want to watch? So just like you might ask an assistant to organize your life. I think people will be surprised: the more obscure, the more creative, the more challenging you think the task is that you're going to ask your AI or Copilot, the better.

但你用它做过自主任务吗？比如，它有没有帮你订过票或买过礼物？因为我知道这是 Copilot 的承诺之一。只是在我所在的地区还无法使用，所以我还没能亲自尝试。

But have you used it to do autonomous tasks? For example, has it booked tickets for you or bought a gift for you? Because I know this is one of the promises of Copilot. It's just not available in my region. So I haven't been able to try it out for myself yet.

我们还在试验中。它能做到，但不总是正确。所以它处于一种开发模式，还没有全面开放。但当它真的起作用时，那绝对是你见过的最神奇的事情。

We're still experimenting with it. It can do it. It doesn't always get it right. So it's in a sort of dev mode. It's not generally available just yet, but when it does work, it is the most magical thing you've ever seen.

它犯过哪些给你带来麻烦的错误？比如买错了礼物？

What are the mistakes it's made that have created problems for you? Bought a present for the wrong person?

嗯，它可能买错东西，但显然你可以干预，而且它在采取下一步行动前总会征求你的许可。所以相当安全。这只是技术的一个特点：它承诺很多，但总是差那么一点。我记得当年渴望得到一台 Discman，梦想着回到 90 年代穿着轮滑鞋在街上滑行。结果我买了一台，但显然它当时还没有防震技术。所以每次我把它侧放或试图塞进后口袋时，它都会跳碟。技术就是这么有趣。它仍然神奇而惊人，但总是还有一点路要走。我认为在这种情况下，距离日常使用还有一段时间。

Well, they can buy the wrong thing, but obviously you can intervene and it will always ask you permission before it takes the next action. So it's quite safe. It's just one of those things with technology: it promises so much and it's always just around the corner. I remember the allure of getting a Discman and I had this dream that I was going to be back in the '90s rollerblading down the street. Then I got one and obviously it didn't have any vibration-proof or shockproof technology yet. So it would skip every time I turned it on its side or tried to put it in my back pocket. So it's a funny thing, technology. It's still magical and amazing, but it's always got a little bit further to go. I think in this case it's a while yet before it's every day.

考虑到你所经历的一切——你是 DeepMind 的创始人之一，曾在谷歌工作，在加入微软之前还创办了自己的 AI 公司 Inflection——这是否意味着当你看到这些挫折或发展阶段时，你仍然有信心？你不会感到沮丧，因为你记得 2016 年你的 AI 模型击败围棋世界冠军的那一刻，那是 AI 的一个重大时刻。

Given everything you've seen, you're one of the founders of DeepMind and you've worked at Google and had your own AI company Inflection before you came to Microsoft. Does that mean that when you see these hiccups or these stages of development, you have faith? You don't get frustrated because you remember the moment when back in 2016 when your AI model beat the world champion at Go, which was a huge moment for AI.

是的。我觉得自己对这些事情非常淡定。我知道它会在未来 6 个月、12 个月，或者最坏情况下 18 个月内实现。但我认为我们现在已经掌握了一种很好的节奏，选择那些不太遥远但真正具有变革性的事情。同时，记住它今天已经多么神奇也很重要。它能生成图像，能生成视频。它已经具备了超人的魔力。我想我就是那种仍然随时会被它的能力所激励的人。

Yeah. I think I'm very stoic about these things. I know that it's going to work in the next 6 months or 12 months or maybe the worst case 18 months. But I think we've got a pretty good rhythm now of picking things which are not too far away, but are genuinely transformative. Also, it's important to remember how magical it is today. It can generate images. It can generate videos. It is already superhuman magical. I think I'm just one of those people that's still inspired by what it can do at any given moment.

到明年这个时候，我能不能用自主 AI 智能体购买圣诞礼物？

By this time next year, could I be buying my Christmas presents using an autonomous AI agent?

是的，我很确定你可以。我认为可能性非常大。

Yeah, I'm pretty certain that you will be. I think it's highly highly likely.

说到神奇的事情，你最近在微软宣布了关于超级智能的消息。这个词在过去几个月里因为你和其他人而真正进入了公众讨论。当你使用「超级智能」这个词时，它对你意味着什么？

Speaking of magical things, the recent announcement that you made at Microsoft about superintelligence, that's a term which has really crept into the public debate thanks to you and others in the last few months. When you use the term superintelligence, what does it mean to you?

当今行业中，超级智能指的是一个 AI 系统，它能学习任何新任务，并且在所有任务上表现优于所有人类的总和。所以这是一个非常高的标准。目前它伴随着巨大的风险。我们如何遏制和限制它的能力，并让这样一个系统真正关心人类、希望我们成功和繁荣，这一点并不清楚，实际上非常不确定。我们如何遏制和对齐一个比我们强大得多的系统，这一点非常不明确。我更喜欢「人本主义超级智能」的框架，即它始终站在我们这边，是我们的队友，与人类利益对齐。在我们能证明它将保持安全之前，我们不会继续开发有可能失控、脱离控制的系统。我认为这是显而易见的事情，就像每个人都应该同意的事实陈述。然而，我认为这在当前行业中是一种新颖的立场。

Superintelligence in the industry today means an AI system that can learn any new task and perform better than all humans combined at all tasks. So it is a very high bar. At the moment it comes with a great deal of risk. It is not clear, in fact it's very uncertain, how we would contain and limit its capabilities and align to get a system like that to care a lot about humans and want us to succeed and flourish. It's very unclear how we would contain and align a system that is so much more powerful than us. The framing that I prefer is one of a humanist superintelligence, one that is always in our corner, on our team, aligned to human interests. Until we can prove that it will remain safe, we won't continue to develop a system that has the potential to run away from us and get out of the box. I think that's an obvious thing to say, like a statement of fact that everybody should agree to. And yet, I think it's kind of a novel position in the industry at the moment.

那么，这就是你试图让微软与众不同的方式吗？很多人都在研究超级智能。你是想通过说「我们始终以人本主义视角使用它」来让微软独具特色吗？

So, is that how you're trying to set Microsoft apart? A lot of people are working on superintelligence. Are you trying to make Microsoft distinctive by saying we will always use it through a humanist lens?

是的。这就是我们的立场。微软是一家已经存在了 50 年的公司。它非常谨慎，高度受信任。90%的标普 500 公司使用我们提供的电子邮件、操作系统和日常生产力工具，这些工具现在借助 Copilot 跻身世界最佳之列。我认为我们之所以拥有这样的声誉，是因为公司一直很谨慎，我们将继续保持谨慎。提出人本主义超级智能的愿景是这一计划的一部分。

Yes. That is our position. Microsoft is a company that's been around for 50 years. It is very careful. It's highly trusted. 90% of the S&P 500 use us to provide email and operating systems and everyday productivity tools which are now among the best in the world with Copilot. I think we've got that reputation because the company's been careful and we're going to continue to be careful. Setting out a vision of humanist superintelligence is part of that program.

那么这对你在这个领域的竞争对手意味着什么？其中一些你与它们密切合作，比如 OpenAI，我相信我们之后会更多谈到。但这是否意味着他们是狂野西部，而你们是道德的一方？

So what does that mean for your rivals in that field? Some of whom you work closely with, like OpenAI, which I'm sure we'll talk more about. But does it mean that they're the Wild West and you're the moral ones?

我认为每个人都必须决定自己的立场和运作方式。我不想评判他们现在的运作方式。

I think that everybody has to decide what they stand for and how they operate. I don't want to judge how they're operating right now.

我没有看到任何大规模伤害的证据。我没有看到任何迹象表明这些东西在自我改进或自主运行，但我们都预测在未来 5 年，也许 10 年内，这些能力会开始出现，这样的系统可能设定自己的目标，或改进自己的代码，或自主行动。这些能力我明确列为增加风险的因素，因此我们必须谨慎对待，提高透明度和审计，加强政府参与，并主动声明我们离这三种能力有多近。我认为这很明显。

I don't see any evidence of large scale mass harm. I don't see any indication that these things are improving themselves or operating autonomously, but we all predict a time in the next 5 years, maybe in the next 10 years, where these capabilities do start to emerge and systems like this could set their own goals, or they could improve their own code, or they could act autonomously. Those things are capabilities that I've clearly outlined as increasing the level of risk, and therefore we have to approach them with caution, with more transparency and audit, with more government engagement, and make proactive declarations about how close we are to those three capabilities. I think that's obvious.

那么这是否意味着，在确信能够控制之前，你不会发布超级智能工具？

And does that mean that you won't be releasing a super intelligence tool until you're confident that it can be controlled?

是的，我认为没错。遏制和对齐绝对是必要的先决条件。它们是红线，我认为行业中的每个人都必须认同这一点。没有人想造成大规模伤害。从根本上说，尽管我们有分歧，但显然每个人都致力于我们物种的生存，并且我希望，致力于每个人的繁荣和福祉。所以这是我们现在试图推动的讨论，并要求行业中的每个人扪心自问：他们是否在构建人本主义的超级智能？它是否真正以人为本、以我们的物种为先？

Yeah, I think that's right. I think for sure containment and alignment are necessary prerequisites. They're red lines, and I think everybody in the industry has to sign up to that idea. Nobody wants to cause mass harm. Fundamentally, even though we all disagree, clearly everybody is committed to the survival of our species and, I would hope, the flourishing and well-being of everybody. So that's the discussion that we're trying to push now and require everybody to ask themselves in the industry: are they building a humanist super intelligence? Does it really stand for people and for our species first?

是的。我只是想知道我们对此能有多大信心，因为事情确实会出错，而且我指的不只是 AI。我的意思是人类会做错事，或者事情会因为其他原因而出错。最近微软不得不处理的一个例子是，在《卫报》报道暗示一些服务可能被用于大规模监控后，微软不得不停止并禁用以色列国防部正在使用的某些服务。

Yeah. I'm just wondering how much confidence we can really have in that given that things do go wrong, and I don't just mean AI. I mean humans do wrong things or things just go wrong for other reasons. And the recent example that Microsoft had to deal with was having to cease and disable some services that were being used by the Israeli Ministry of Defense after reporting by the Guardian that suggested they could be being used for mass surveillance.

那是《卫报》非常好的报道。我们对此非常感激，一旦我们意识到这一点，我们就做了所有必要的更改，将以色列国防军从那些服务器中移除，我认为他们明显违反了我们的服务条款。内部正在进行调查。

That was very good reporting by the Guardian. We were very grateful for it, and as soon as we became aware of it, we actually made all of the necessary changes, removed the IDF from those servers, and I think they were clearly not in compliance with our terms of service. There's an ongoing investigation internally.

但我想更广泛的观点是，很难对控制、制衡和使用有信心。

But I guess the broader point is that it's hard to have confidence of controls and checks and balances and uses.

这确实很难。这些是庞大而复杂的系统。它们承载着很多风险。我们最多能做的就是确保我们在审计它们，并尽快移除违反我们服务条款的行为者。

It is hard. These are huge and complicated systems. They carry a lot of risk. The most that we can do is make sure that we are auditing them and we're removing actors that violate our terms of service as quickly as possible.

那么，超级智能的首次应用会是在医疗领域吗？

So, are the first uses of super intelligence going to be in the medical field?

我认为是的。我认为这可能是超级智能最令人兴奋的应用。我认为我们现在有系统可以诊断文献中发现的任何罕见疾病，其表现显著优于人类，并且需要的测试和干预更少。关键在于：在确诊之前进行医疗干预测试通常非常昂贵，而且往往不必要，还耗费时间，在这个过程中病情可能会恶化。在我们的案例中，我们的系统以更低的成本、更少的测试和更高的准确性做到这一点。我们目前正在对其进行独立同行评审，很快也会进行临床试验。所以我认为这非常非常令人兴奋。

I think so. I think this is probably the most exciting application of super intelligence. I think we now have systems that can diagnose any rare condition that is found in the literature significantly better than human performance and do so with fewer tests and interventions. That's the key thing: often it's actually quite expensive to run medical intervention tests before you get to a diagnosis, and often they can be unnecessary too, and they take time, and in that process the condition can often worsen. In our case, our system does this significantly more cheaply, with fewer tests, and with higher accuracy. We're just putting it through independent peer review at the moment, and soon there'll be clinical trials of it too. So I think this is very, very exciting.

是你自己推动了这个重点吗？因为你管理微软 AI 已经有一段时间了，大约 18 个月。超级智能是最近的一个重大公告。我注意到，当你谈到它时，你的母亲是一名护士，而且你——在科技界高层中不常见——也在公共部门工作过。

And did you push for that focus yourself? Because you've been running Microsoft AI for a while, about 18 months now. Super intelligence was a big recent announcement. And I'm conscious as I hear you talk about it that your mother was a nurse and you, unusually for senior people in tech, you've worked in the public sector as well.

是的。这个领域对我非常重要。显然我妈妈是 NHS 的护士，我坚信技术是用来服务我们的，它应该让我们的生活更好，让我们更舒适。有一天我认为它会帮助我们活得更久。我认为它会让我们选择少工作。我确实认为它会带来富足，而且我们必须有意识地决定首先将其用于这些应用。这就是为什么，即使在 DeepMind 的时候，我也负责我们的健康部门。我对此充满热情。

Yeah. This is an area that's very important to me. Obviously my mom was an NHS nurse, and I'm just a big believer that technology is here to serve us, it should make our lives better, should make us more comfortable. One day I think it is going to help us to live longer. I think it's going to give us the option to work less if we choose to. I do think it's going to produce abundance, and I think that we have to make conscious decisions to use it for those applications first. That's why, even back at DeepMind, I ran our health unit. I'm very passionate about this.

富足是一个如此宏大的承诺。告诉我你是什么意思，因为大多数时候我们听到的是 AI 摧毁工作，但你是说工作将越来越多地由 AI 完成，因此我们人类不必工作那么多？

Abundance is such a big promise. Tell me what you mean by it because most of the time we hear about AI destroying jobs, but are you saying that the work will be done by AI increasingly and therefore we humans won't have to work as much?

我认为在未来 20 或 30 年的某个时候，机器在大多数工作上的能力将超过人类，而且这可能来得更早。所以我们必须作为一个社会决定我们的目标是什么，我们必须非常谨慎地考虑新机器的引入速度。因为我们必须确保这种替代速度与一种机制相平衡，这种机制能够为人们提供资金和支持，帮助他们度过一个巨大的转型期，而在转型之后，可能不会有比机器更具竞争力的其他工作。所以这确实是社会组织方式的一个重大变革。

I think it is inevitable that at some point over the next 20 or 30 years, machines are going to be more capable than humans at doing most work, and that might come much sooner. So we have to decide as a society what our purpose is, and we have to be very thoughtful about the rate of introduction of new machines. Because we have to make sure that that rate of displacement is counterbalanced with a mechanism to fund people and to support people through a massive transition for which there is probably not going to be other jobs on the other side that are more competitive than what a machine can do. So this is a real big step change in what it is to organize society.

但你相信全民基本收入的想法吗？即 AI 的高生产力、使经济更具生产力可以解锁这一点？

But do you believe in the idea of a universal basic income that that's what AI being very productive, making the economy more productive could unlock?

是的，我认为我长期以来一直公开表示，我认为这既不可避免又非常可取。我们已经生活在一个富足的世界里，只是分配不均。我们已经有更多的卡路里可以消耗。我们有大量的食物浪费。只是分布不均。而现在，价值不仅仅体现在原子层面，如食物、汽车和实物，还体现在数字商品、思想、知识和智能上。这实际上是个好消息，因为这些东西可以扩散，可以极快地传播到全世界。这些大语言模型和聊天机器人是历史上传播最快的技术。基本上在 3 年内就有 20 亿年活跃用户。这真的很了不起，因为它正在向每一个想使用的人传递知识，以他们当前使用的格式和设备。所以我认为让技术普及到每个人没有问题，因为会有巨大的竞争力量来降低体验 AI 的成本。

Yeah, I think I've long been on record saying that I think that is both inevitable and very desirable. We already live in a world of abundance. It's just poorly distributed. We already have more calories to consume. We have an incredible amount of food wastage. It's just unevenly distributed. And now that value isn't just manifested in atoms like food and cars and physical things. It's manifested in digital goods, ideas, knowledge, intelligence. And that's actually great news because that can proliferate. It can spread extremely quickly around the entire world. These LLMs and chatbots have been the fastest spreading technology in history. Basically two billion annual users in the space of 3 years. That's actually amazing because it's delivering knowledge to every single person who wants to use it on their phone in the current sort of format and device they already use. So I think that we have no problem with getting technology out and available to everybody because there's going to be massive competitive forces to reduce the cost of experiencing an AI.

请多谈谈这些想法和信念的来源。你母亲是护士，但你在伦敦 80 年代和 90 年代成长的家庭，你会如何描述它？

Tell me more about where these ideas and your beliefs came from. Your mother was a nurse, but the family you grew up in in London in the 1980s and 1990s, how would you describe it?

相当工人阶级。是的。我爸爸是迷你出租车司机。我妈妈在 NHS 工作。我们相当普通，没什么特别。我父母不太重视教育。他们总认为我应该学门手艺。我妈妈常对我说，你应该当木匠、电工，16 岁离开学校之类的。所以，我不知道。这源于我经历过一些艰难，然后渴望在短暂的生命中尽力做到最好。

Pretty working class. Yeah. My dad was a mini cab driver. My mom works in NHS. And yeah, I mean, we were fairly regular, kind of unremarkable. My parents didn't super value education stuff. They always thought I should go get a trade. And my mom would often say to me, you know, you should be a carpenter, electrician, leave school at 16 and stuff. So, I don't know. It comes from a place of just experiencing the kind of rougher end of things a little bit and then having just a desire to try to do the best that we can do with the short life that we have.

你父亲从叙利亚来到英国。你父母是在英格兰认识的吗？

Your dad had come to the UK from Syria. Did your parents meet in England?

是的，没错。

Yes. Yeah, that's right.

你 16 岁时发生了什么，Mustafa？因为我读到，当你父母分开后，你和弟弟几乎要自谋生路。

And what happened when you were 16, Mustafa? Because I read that when your parents split up, you and your younger brother were pretty much left to fend for yourselves.

没错。我不知道你在哪里读到的，但我和弟弟确实独自生活了几年。

That's true. Yeah. I don't know where you read that, but yeah, me and my younger brother did live on our own for a few years. Yeah.

我想你在几年前《卫报》的采访中提过，当时你的书刚出版，但我很想知道这教会了你什么，以及你当时是如何应对的。

I think you mentioned this in a Guardian interview a couple of years ago when your book was coming out, but I'm really curious to know more about what it taught you and how you dealt with it at the time.

我觉得在那个年纪，你有点早熟、过度自信且无所畏惧。我当时并没有真正注意到。我们拥有所需的一切。结果还不错。

I think that when you're that age, you're just sort of precocious and overconfident and fearless. And I didn't really notice it at the time. You know, we had everything that we needed. And it worked out pretty well.

这是在关键时期，中学的最后几年。你最后几年确实表现很好，因为你最终进入了牛津。

And this is at a crucial time, the last couple of years of school. And you did do well in those last couple of years of school because you ended up getting to Oxford.

是的，我很幸运，因为我有很好的老师和导师，我上了一所很好的学校。你可以通过考试入学。所以我 10 岁时就非常努力准备入学考试，那基本上就像上私立学校。每个人都很勤奋，成绩很好。我很幸运能进入牛津，那是一次很棒的经历。

Yeah, I was very lucky because I had great teachers and good mentors and I went to a very good school. You could do exams to get in. So I studied really hard when I was 10 years old for the entrance exams and it was essentially like going to a private school. I mean, it was basically everyone was very studious and got great grades and stuff and yeah, I was lucky enough to get into Oxford and that was an amazing experience.

你有没有想过，如果没有那所学校，你十几岁中期的那段生活会是什么样子？

Have you thought about what your life would have been like in that period in your mid-teens had it not been for that school?

是的，我没怎么想过。那会非常不同。

Yeah, I haven't thought much about that. It would have been pretty different.

然后你进了牛津，学的科目和我儿子现在学的很接近。他学神学。你学的是哲学和神学。然后，作为家长，我觉得很遗憾，你辍学了。

And then you got to Oxford and you did actually a subject pretty close to what my son is doing there right now. He's doing theology. You did philosophy and theology. And then tragically to my ears as a parent, you dropped out.

是的。我当时很沮丧，渴望改变世界，做成一些事。所以我真的辍学了，并帮助创办了一个针对英国穆斯林青年的电话咨询服务，叫穆斯林青年热线，我作为志愿者做了两三年。这是一个世俗的、非宗教的、非指导性的、非评判性的、只是同龄人之间的倾听服务，面向 9/11 后经历身份危机的英国穆斯林青年——事后看来我们会这样描述，但当时只是缺乏与社区、家庭、父母的联系，很多欺凌等等。所以我几乎全职做了两三年。

Yeah. Well, I was very frustrated and eager to change the world and get stuff done. So I actually dropped out and I helped to start a telephone counseling service for young British Muslims called Muslim Youth Helpline, which I worked on for 2 or 3 years as a volunteer. It was a secular, non-religious, non-directional, non-judgmental, just peer-to-peer listening service for young British Muslims that post-9/11 were dealing with a lot of, I guess identity crisis would be the way that we would describe it in hindsight, but at the time it was just like lack of connection to community, family, parents, a lot of bullying, and so on and so forth. So, yeah, I worked on that pretty much full-time for two or three years.

考虑到我们谈论的是 9/11 之后的时期，你自己当时经历过反穆斯林情绪和仇恨吗？

Had you experienced anti-Muslim sentiment and hatred yourself at that time given the period we're talking about after 9/11?

有一点。是的。我认为更多不是针对我个人，而是很多人觉得我们不够英国化，人们正在摸索如何在第一代移民家庭中同时保持文化和宗教身份。他们不太会说英语，不知道如何应对体制。然后这种怀疑增加了，认为我们大多是恐怖分子之类的。我认为这只是一种普遍的恐惧和排斥。实际上，大多数时候，通过友好和支持性的对话，以及电话另一端有人倾听，就能应对这种情况。

A little bit. Yeah. I mean, I think that it was more just not so much me but I think a lot of people felt that we weren't sort of British enough and people were figuring out how to both live their cultural religious identities in the context of families that often were first generation. Didn't really speak the language. Didn't know how to navigate the system. And then this increased kind of skepticism that we were mostly terrorists or whatever. And I think there was just this general kind of fear and exclusion. And actually most of the time, that was dealt with by just having kind and supportive conversation and having somebody be available on the other end of a telephone.

是的，有人可以交谈。我实际上知道热线的工作，所以我完全理解你说的世界。你当时认真考虑过辍学吗？之后有没有后悔过？

Yeah. Someone to talk to. I know the work of the helpline actually, so I know exactly the world that you're talking about. Did you think hard about dropping out of university? Have you regretted it at any point since?

我没有后悔。我可能本可以更认真地考虑一下。我的导师确实鼓励我换专业或转学院，还让我去见他在另一个学院的朋友，讨论学历史或 PPE。但我觉得那时我精力充沛，渴望开始做事。我同时创办了一家企业，做电子销售点，向餐馆销售 PDA，用于餐厅预订之类，还做他们的网络等等。那还是互联网早期。所以有点早，但我也做了一两年。我只是精力太旺盛，不适合在梦幻尖塔下坐着思考。

I haven't regretted it. And I probably could have thought harder about it. My tutor did encourage me to switch subjects or change college and actually had me go meet one of his friends in a different college to go and talk about doing history or PPE. But I think at that point I was just so full of energy and eager to start things. I started a business at the same time doing electronic point of sale, selling PDAs into restaurants to try and do restaurant booking and stuff and doing their networking and so on. It was very early days of the internet. So it was a bit early but I did that for about a year or two as well. So I was just like I had too much energy, I think, for sitting around thinking in the dreamy spires.

你找到了释放的途径，因为后来你和 Demis 共同创立了 DeepMind，现在你身处这个权力圈，对吧？作为少数决定一项已经改变并将继续改变我们生活的技术的人之一，你对这种权力有多清醒？

Well, you found the outlets for it because then of course you went on to co-found DeepMind with Demis and now here you are in this circle of power, right? One of a small group of people who are making decisions on a technology that is already changing all our lives and will continue to do so. How conscious are you of that power?

非常非常清醒。所以我非常认真地对待，这是一项巨大的责任。我认为这是一个我们做出的决定可能产生非常持久后果的时刻。我读历史，回顾过去，例如在社交媒体革命中，我认为很明显存在潜在危害，但可能长期被忽视；或者回顾电动汽车革命，它本可以在 20 年代就起飞；或者看看吸烟或石油。所以我对这些事情非常敏感，因为很明显这些东西会造成伤害，我认为我们必须非常非常小心地部署它们，以及如何将它们引入世界。

Very, very conscious. So I take it very seriously and it's a great responsibility. I think that this is a moment when decisions that we make may have very lasting consequences. I read history and I look back at times, for example in the social media revolution where I think it was very clear that there were potential harms that maybe fell on deaf ears for too long, or I look back at the electric car revolution that could have taken off way back in the 20s, or you look at kind of smoking or oil. So I'm very sensitive to those things because it's very clear that these things will cause harm and I think that we have to be very, very careful about how we deploy them and how they're introduced into the world.

你和你的同行，比如山姆·奥特曼，我知道你和他关系很好，你们会讨论这些吗？

Do you talk about this with your peers with people like Sam Altman who I know you're close to?

是的。我认为行业里的每个人都在讨论，而且确实有一群 CEO。显然，山姆和 Anthropic 的达里奥曾是联合创始人。我和戴米斯也是联合创始人。我们都非常了解彼此。总的来说，我认为每个人都真心致力于找到正确的道路。但这也非常竞争，所以显然也有很多这种动态。

Yes. I think everybody in the industry does, and there's definitely a group of the CEOs. Obviously, Sam and Dario from Anthropic were co-founders. Me and Demis were co-founders. We all know each other very well. I think on the whole, everybody is genuinely committed to trying to find the right path through. It's also very competitive, so there's obviously a lot of that dynamic too.

是的。你听说过「兄弟寡头」这个词吗？

Yeah. Do you recognize the term broligarchy?

我以前没听过，但我能猜出它的意思。我想确实如此。而且这非常以男性为中心，不过应该说，OpenAI 的前 CTO 米拉·穆拉蒂是这个领域最优秀的人之一。

I haven't heard that before, but I can figure out what it means. I guess that's true. And it is very male-centric, although it should be said Mira Murati, who's the ex CTO of OpenAI, is one of the best people in the field.

或者「科技兄弟」。我相信你听过这个。不知道你有没有看过杰西·阿姆斯特朗的《山巅》？

Or Tech Bros. I'm sure you've heard that one. I don't know if you've seen Jesse Armstrong's Mountain Head.

不，我没看过。

No, I haven't.

它讲的就是一群人，其中一个人明显是埃隆·马斯克，另一个人看起来或听起来像彼得·蒂尔。嗯，我推荐它。我们不久前在这里采访过他，但这是一个相当发人深省的描绘，他对这个非常精英、非常有权势的圈子的印象。

It is all about a group of people, one of whom is clearly Elon Musk, another one who looks like Peter Thiel or sounds like Peter Thiel. It's well, I recommend it. We interviewed him not long ago here, but it's quite a sobering portrait, his impression of life in this quite rarified, very powerful circle.

是的。我认为挑战的一部分在于，我们很多人都花大量时间在硅谷，那里天空湛蓝，生活非常平静。我尽量多旅行。几周前我刚从中国回来，去了五六个不同的城市，走出泡沫，看看这项技术在世界另一端是如何发展的，真是令人震惊。既有创新的速度，也有一些监管方面的深思熟虑。我认为他们肯定也在非常仔细地考虑这个问题。所以这令人印象深刻。我也经常回伦敦，确保自己花时间在硅谷之外。

Yeah. I think part of the challenge is that we all spend a lot of time in Silicon Valley where the skies are blue and life is very peaceful. I try to travel a lot. I just came back from China a couple weeks ago, where I was in five or six different cities, and it's just staggering to get out of the bubble and see how this technology is being developed on the other side of the world. Both the pace of innovation, but also the thoughtfulness of some of the regulatory stuff. I think they are definitely thinking pretty carefully about it as well. So it's impressive. I also come back to London a lot and make sure that I spend time outside of Silicon Valley.

是的。所以如果你对你们的世界有不同于杰西·阿姆斯特朗或其他人的印象。我能请你，因为你认识这些关键人物，用一句话描述他们中的一些人吗？比如如果我说山姆·奥特曼，你首先想到什么词？

Yeah. So if you have a different impression to someone like Jesse Armstrong or others about your world. Can I ask you just to because you know these key players can you describe some of them in one word? Like if I say to you Sam Altman what word first comes into your mind?

哦天哪。我想是勇敢，因为他显然在非常积极地扩大他的数据中心规模。我认为他很可能会成为我们这一代最伟大的企业家之一。他确实取得了很大成就，现在他正在以比业内任何人都快的速度建设数据中心。如果他能成功，那将非常引人注目。

Oh my god. I guess courageous, because he's obviously growing his data center fleet very aggressively. I think he may well turn out to be one of the great entrepreneurs of our generation. He's certainly achieved a lot, and now he's building data centers at a faster rate than anyone in the industry. If he can pull it off, it will be pretty dramatic.

有趣的是，这里有一个「如果」，我理解为什么。OpenAI 正在花费巨额资金。这是赌博吗？对他们来说，这不一定能成功吗？

It's interesting that there is an if on that and I understand why. There are huge amounts of money being spent by OpenAI. Is it a gamble? Is it not a given that it's going to pay off for them?

嗯，我认为 ChatGPT 是我们这一代人所见过的最伟大的产品之一。这本身就说明了一切。与此同时，他们签署了超过 1.5 万亿美元的承诺，用于未来 5 到 10 年建设数据中心。他们的收入离这个数字还很远。所以他们还有很长的路要走，但他们是一个非常才华横溢的团队。我完全相信他们能做到。

Well, I think that ChatGPT is one of the greatest products that we've seen in a generation. That speaks for itself. At the same time, they've signed over a trillion and a half dollars of commitments for building data centers over the next 5 or 10 years. Their revenues are quite a long way from there. So they've got a long way to go, but they're a very talented team. I have every confidence they can do it.

那么其他描述。你会用什么词来形容戴米斯·哈萨比斯？

So other descriptions. What word would you use to describe Demis Hassabis?

可能是一位伟大的科学家。我认为他是一位伟大的思想家，也是一位优秀的博学家。他在这个领域多次做出了巨大贡献。我认为他确实非常出色。

Probably a great scientist. I think he's a great thinker and a good polymath. He's made massive contributions in the field multiple times. I think he's truly exceptional.

当然，你们曾一起工作，现在你们是竞争对手。

You were, of course, you worked together and now you're competitors.

是的。但我们一开始是非常亲密的朋友。我们一起工作了十年。我从他那里学到了很多，非常尊重他。

Yeah. But we started off as very close friends. We worked together every day for 10 years. I learned a lot from him and have huge respect for him.

你们还说话吗？

Do you still talk?

是的，我们昨晚还发短信了。我祝贺他一周内同时发布了 Nano Banana Gemini 3 和 AlphaFold 五周年。

Yeah, we texted last night actually. I congratulated him on the Nano Banana Gemini 3 and 5 years of AlphaFold all in one week.

你试过 Gemini 3 吗？因为它引起了很大的轰动。你认为它比 ChatGPT 更好吗？

Have you tried out Gemini 3 because it is creating quite a lot of waves. Do you think it is better than ChatGPT?

它们有点不同。它确实有一些 ChatGPT 没有的 niche 技能，而且速度很快。但 ChatGPT 非常强大。所以我不会说它更好。

They're kind of different. It's definitely got more niche skills that ChatGPT doesn't have, and it's very fast. But ChatGPT is very strong. So I wouldn't go that far.

好的。它比 Copilot 好吗？

Okay. Is it better than Copilot?

它能做 Copilot 不能做的事情。Copilot 有它没有的功能。Copilot 在视觉方面实际上非常出色。我最喜欢的功能之一是拿出手机，手里拿着 Copilot 谈论世界，它可以准确地看到你在看什么，你可以谈论它并获得反馈和意见。你可能在买家具或衣服。我认为我们真的在想象，拥有这样一个非常智能的助手在你身边，可以和你谈论世界，并在你遇到困难时帮助你解决问题，日常体验会是什么样子。

It can do things that Copilot can't do. Copilot has features that it doesn't have. Copilot is actually amazing for vision. One of my favorite features is pull out your phone and talk about the world with Copilot in your hand, and it can see exactly what you're looking at, and you can talk about it and get feedback and opinions. You might be shopping for furniture or clothes. I think that we're really trying to imagine what is the day-to-day experience of having this really intelligent assistant at your side that can talk to you about the world and help you unblock you whenever you get stuck with something.

我只再问一个。你会如何描述行业中的一位关键人物，但这次是埃隆·马斯克。你会怎么描述他？

I'm only going to do one more of these. How would you describe a key person in the industry, but it's Elon Musk. How would you describe him?

我想是推土机。他有超人的能力，能让现实屈服于他的意志，而且有着非常不可思议的记录。不知何故，他大多能完成看似不可能的事情，可能带着一套不同的价值观。

I guess as a bulldozer. He's got superhuman capabilities to bend reality to his will and has a pretty incredible track record. Somehow he mostly manages to pull off what appears to be impossible, probably with a different kind of set of values.

我采访过他。他最后叫我 NPC。我想那可能是一种推土机式的行为。

I have interviewed him. He ended up calling me an NPC. I guess that was probably a form of bulldozing.

是的，这可能是个好例子。听起来完全像埃隆。我有点喜欢他直言不讳。他非常不设防。

Yeah, that's probably a good example. That sounds exactly like Elon. I kind of like that he speaks his mind. He's very unfiltered.

我想你说过你的政治观点与他以及彼得·蒂尔不同，那种自由意志主义，对国家没什么价值等等。那不是你的世界。你会称自己为政治光谱上的左派，还是你会用不同的描述？

I think you've said that your politics are different from him and Peter Thiel, that the libertarian, not much value to the nation state etc. It's not your world. Would you call yourself as being on the left of the political spectrum or is there a different description you'd use?

我会说我如今算是中间派。我肯定是从左派开始的，但我很自豪地说我处于光谱的中左位置。我相信政府在社会中扮演着重要角色。

I would say I'm sort of a centrist these days. I definitely started as a lefty, but I'm proud to say that I'm on the kind of center-left of the spectrum. I believe that government plays an important role in society.

在硅谷说这话是有争议的，但我认为监管是必要的，它让大多数技术变得更好。人们忘了这一点，比如汽车之所以能运行，是因为我们有驾驶员培训、排放法规、路灯、高速公路限速等等。这就是监管运作良好的样子，我们需要更多这样的监管。你说这话会不会感到孤立？因为显然特朗普政府不热衷监管，整个行业对此也很满意。

It's a controversial thing to say in Silicon Valley, but I think regulation is necessary and it has made most technologies better. And people forget this, like cars only work because we have driver training and emissions regulations and street lights and motorway speed limits and so on. And so that's what regulation is when it works well and we just need more of that. Do you feel isolated saying that because clearly the Trump administration is not into regulation and the industry overall is quite happy with that?

是的，我认为目前我们并没有处于巨大灾难性危害的模式。我认为行业在引入非常强大的聊天机器人方面做得相当不错，其个性被塑造得非常公正、基于证据。我认为这本来不一定会这样发展，但之所以如此，是因为领导者们做得相当好。但这并不意味着未来不会出现问题，我对此非常警惕，但我不认为我们需要紧急监管。比如，欧洲提出的一些监管措施目前正在讨论通过《欧盟人工智能法案》进行回撤。我认为这是好事，人们不应该批评。这是流程在起作用，监管者正在听取反馈，观察实际运作情况。这应该被赞扬。同时，他们必须有实验和尝试新监管的自由，我认为这是好事。

Yeah, I think that right now we are not in a mode where there is huge catastrophic harm. I think the industry's done a pretty good job of introducing very powerful chatbots in a way where the personality is sculpted to be very even-handed, to be very evidence-based. And I think that it didn't have to go like that. It went like that, I think, because the leaders have done a pretty decent job. Now, that doesn't mean that there aren't going to be issues coming up, and I'm obviously very wide-eyed about them, but I don't think we're in a mode where we need emergency regulation. I mean, I think, for example, some of the regulations that have been proposed in Europe are currently in talks to be wound back with the EU AI Act. And I think that's good. People shouldn't criticize that. That's the process working. That's the regulator taking feedback, seeing how things work in practice. I think that should be celebrated. And at the same time, they have to be able to have the freedom to experiment and try new regulations, and I think that's a good thing.

穆斯塔法，你现在身处最大公司之一的核心，是否对这些问题有了不同的看法？因为在你《浪潮来临》一书和同时期为《外交事务》撰写的文章中，你非常明确地指出了需要多少治理。事实上，你说你想要三种不同的治理机制，分别借鉴气候变化、金融稳定和军备控制，人工智能和技术可以从这三者中学习。但现在你听起来乐观多了。仅仅是因为你的位置吗？

Do you think Mustafa, you see some of these issues differently now that you are right at the heart of one of the biggest companies of all? Because in your book "The Coming Wave" and in a piece you wrote about the same time for Foreign Affairs, you were quite clear on how much governance was needed. In fact, you said you wanted three different kinds of governance regimes modeled on climate change, financial stability, and arms control that AI and tech could learn from all three of those. And yet you sound much more sanguine now. Is it just because of where you are?

不，我仍然在呼吁这些。看，我们应该为人工智能设立一个金融稳定委员会。我们应该有一个气候进程，去审计我们取得的进展。我在拜登政府时期倡导了白宫人工智能原则，当时我们作为公司做出了自愿承诺——那时是我的初创公司 Inflection——但我们推动所有其他公司也这样做，包括披露模型规模。我们在欧洲和英国人工智能安全研究所签署了这些披露。所以，我不是说我们不需要这些。我是说这些影响是长期的，治理过程需要很长时间来建立，所以我们基本上必须现在开始。但我们不需要膝跳反应，不需要恐慌性的过度反应，因为那会引发另一系列问题。

No, I'm still calling for those things. Look, we should have a financial stability board for AI. We should have a climate process which goes and audits the progress we're making. I advocated for the White House AI principles under the Biden administration where we made voluntary commitments as companies—at the time it was my startup Inflection—but we pushed all the other companies to do it as well, where you disclose the size of your model. We've signed up to those disclosures in Europe and with the UK AI Safety Institute. So, yeah, I don't mean to say that we don't need it. I'm saying that these are such long-term effects and these governance processes take so long to build that we basically got to start now. But what we don't need is knee-jerk reaction. We don't need some panicked overreaction because that would cause a different set of problems.

你正在做的超级智能工作，以及目前你在人工智能领域所做的一切，都进入了一个新阶段，对吗？因为现在你与 OpenAI 达成了修订协议，突然有了独立追求人工智能的自由。帮我理解一下未来与 OpenAI 的关系将如何运作，因为微软和 OpenAI 长期以来一直是合作者。现在你们也将成为竞争对手。这怎么运作？

And the superintelligence work you're doing, and indeed everything you're doing in AI at the moment, you're in a new phase, aren't you? Because now that you've got a revised agreement with OpenAI, you suddenly have the freedom to pursue AI independently. Help me understand how that relationship with OpenAI is going to work in future because you've been collaborators—Microsoft and OpenAI—for a long time. Now you're also going to be competitors. How does that work?

是的，这对人们来说有点复杂。但底线是，直到几周前，合同禁止微软独立追求通用人工智能或超级智能。与 OpenAI 的协议是，当他们在 2019 年签署协议时，OpenAI 将去构建 AGI。作为回报，微软将构建人工智能基础设施——芯片和数据中心等。微软通过提供芯片获得已构建模型的许可。我们仍然拥有 OpenAI 到 2032 年构建的所有内容的许可。但 OpenAI 决定他们想要更多算力，并从其他供应商购买算力。所以他们现在与软银、甲骨文等公司达成了协议，建造比微软愿意为他们建造的更多的数据中心。作为回报，我们有权去开发自己的人工智能。显然，这是我 18 个月前加入公司的重要原因——我们现在正在招聘一个超级智能团队，追求我们自己的人工智能开发。

Yeah, it's a sort of complicated setup for people to follow along. But the bottom line is that up until a few weeks ago, Microsoft was not allowed by contract to pursue artificial general intelligence or superintelligence independently. And the deal with OpenAI was that it would go and build AGI when they signed the agreements back in 2019. In return, Microsoft would build the AI infrastructure—the chips and the data centers and so on. And Microsoft would get, in return for providing the chips, a license to the models that have been built. And we still have that license to everything that OpenAI builds up until 2032. But OpenAI decided that they wanted to take on more compute and buy compute from other providers. So they now have deals with SoftBank and many others, Oracle, to build more data centers than Microsoft wanted to build for them. In return, we then have the right to go and develop our own AI. And obviously that was a big part of me joining the company 18 months ago—that we're now hiring out a superintelligence team and pursuing our own AI development.

这笔交易对你来说是种解脱吗？因为如果没有它，作为微软的人工智能负责人，你能做多少？

Was that deal a relief to you? Because without it, how much could you have done as Microsoft's AI chief?

嗯，很多。我们有 2800 亿美元的收入。世界上所有主要机构每天都需要使用我们的人工智能功能和工具。过去 18 个月，我们仍然是一个通用人工智能开发机构。但现在我们可以研究一些有可能在所有任务上超越人类表现的技术和方法，所以这对我们来说是一个转变。

Well, I mean, a lot. We have $280 billion of revenue. All major institutions in the world need to be able to use our AI features and our AI tools every single day. We've still been a general-purpose AI development shop over the last 18 months. But now we can work on some techniques and methodologies that have the potential to exceed human performance at all tasks, so it is a shift for us.

你与 OpenAI 的关系是否包括不互相挖角？因为我们刚刚又看到了人才战争的证据。你刚刚失去了一位关键人物，他去了苹果。

And does your relationship with OpenAI include not poaching each other's staff? Because we've just had another bit of evidence of the talent war that's underway. You've just lost one of your key people to Apple.

是的，他将去运营苹果实验室。我认为人才流动很大。行业人才——我们刚刚从 Google DeepMind 招来了一大群人。我们也从 OpenAI 招了人。流动很大。这是行业的一部分。所以肯定没有不挖角协议——那是不合法的。人们可以为任何他们想为之工作的人工作。我认为在行业这个时期，竞争非常激烈。我认为这一切都是意料之中的。

Yeah, he's going to go run the Apple lab. I mean, look, I think that there's huge rotation. The talent in the industry—we just got a whole bunch of people from Google DeepMind. We got people from OpenAI. There's a lot of rotation. This is part of the industry. So there's certainly no no-poach agreements—that would not be legal. And people can go work for whoever they want to work for. I think that at this time in the industry it's just very competitive. I think all of that's to be expected.

这是否意味着你准备并且已经在匹配 Meta 花费的那种金额——1 亿到 2 亿美元的薪酬包？

And does that mean that you are prepared to, and are you already matching the kinds of sums that Meta's been spending—the 100 to $200 million packages?

我不认为有人在匹配那些。我认为扎克伯格采取了一种特定方法，涉及招聘很多个人，而不是创建团队。我真的不认为那是正确的方法。

I don't think anyone's matching those things. I think that Zuck's taken a particular approach that involves hiring a lot of individuals rather than maybe creating a team. And I don't really think that's the right approach.

我认为我们在 DeepMind 做得好的地方是，我们非常挑剔。过去一年半在微软，我一直在逐步增加符合文化、技能合适且能与团队其他成员良好协作的人。剔除那些不合适的人，并非常注重细节。所以，我们更多是在打造一个团队，而不是一群个体。

I think what we did well at DeepMind is that we were very selective and we've, what I've been doing at Microsoft over the last year and a half is adding people incrementally who fit the culture and who suit the skill set and work well with the rest of the group. Weeding out those who don't and being very attentive to the details. So, we're very much creating a team rather than a set of individuals.

所以你不准备匹配那种规模的薪资报价，因为最近有报道这么说？

So you're not prepared to match salary offers on that scale because that was what was recently reported?

是的，那是个谣言。我认为那显然非常前所未有，可能对少数几个人来说是如此，但这肯定不是常态。

Yeah, I mean that was the rumor. I think that was obviously very unprecedented and may have been the case for a few individuals, but that's certainly not the norm.

那你是否希望他，马克·扎克伯格，或者你叫他 Zuck，重新考虑一下？这会让你的生活更困难吗？

And would you rather he, Mark Zuckerberg or Zuck as you call him, would you rather he rethought that? Does it make life much more difficult for you?

一点也不。我认为他全力以赴了。他正在建设一个两吉瓦的数据中心，未来两三年可能会花费他几千亿美元。而且他公开对特朗普说过，未来三年将在数据中心上花费超过 6000 亿美元。幸运的是，微软拥有 33 吉瓦的算力容量。我们的主要业务是建设数据中心并提供给第三方。所以我们可以将其用于自己的训练，也可以用于推理并出售给第三方。从这个意义上说，我们的布局非常好。我们某种程度上对冲了这种风险。

No, not at all. I think that he's all in. He's building a two gigawatt data center, which will probably cost him a couple hundred billion dollars over the next two or three years. And I think he said publicly to Trump that he's going to spend upwards of $600 billion in the next 3 years on data centers. Luckily for Microsoft, we have 33 gigawatts of compute capacity in our fleet. Our primary business is to build data centers and provide them to third parties. So we can either use that for our own training or we can use it for inference where we sell it to third parties. So we've got a very good setup in that sense. We're kind of hedged against that risk.

你能理解为什么人们不仅担心行业投入的资金量，还担心许多交易的循环性质吗？你是 OpenAI 的投资者，他们也从你这里购买服务。每个人似乎都与英伟达有关系。美国经济的很大一部分依赖于像你们这样的公司持续表现良好。

Can you see why people worry about not just the amounts of money being spent by the industry, but the circular nature of a lot of those deals? You're an investor in OpenAI, they're also buying services from you. Everyone seems to be in a relationship with Nvidia. There's a huge proportion of the US economy that is dependent on things continuing to go well for firms like yours.

是的，我认为这很合理。显然，客户和供应商经常相互投资，这为我们提供了良好的刺激。但这是值得关注的事情。我肯定在密切关注，我想其他人也是。取得平衡非常重要。我们必须在未来几年内交付成果。现在每个团队都在构建极其庞大、非常强大的计算机，我们正在押下巨大的赌注，认为能够将其转化为真正的智能。如果我们成功了，那么我认为世界将会变得非常不同。我们将拥有丰富的智能随时可用。我认为人们仍然低估了这有多么非凡。

Yeah, I think that's fair. Obviously customers and providers have often invested in one another and it provides good stimulus to get us going. But it's something to watch. I'm definitely watching it carefully and I think others are too. Getting the balance right is very important. We have to deliver in the next few years. Every team now is building incredibly large, very powerful computers and we're taking a huge bet that we're going to be able to convert this into true intelligence. And if we do, then I think the world is going to look very different. We will have abundant intelligence on tap. And I think people still underrate how phenomenal that is.

你说世界，但这实际上只关乎两个国家，美国和中国。几乎所有的实验室都在那里。

You say the world except that this is really about two countries, the US and China. That's where almost all the labs are.

是的，我认为确实如此。过去 18 个月里，一切都更加集中在硅谷周围。这有点自然。我认为我们在许多其他历史技术趋势中也看到了这一点。与此同时，开源做得很好，生产成本也在急剧下降。向世界上最好的 AI 模型提问的成本比两年前降低了 90%。没有任何其他学科会出现下降如此之快的曲线。显然，当成本下降时，每个人都能获得访问。所以，我认为不会出现访问问题。

Yeah, I think that's true. And everything has concentrated even more around Silicon Valley in the last 18 months. That's sort of natural. I think we see that in a lot of other historical technology trends. At the same time, open source is doing very well and the cost of production is also going through the floor. It costs 90% less to ask a question of one of the best AI models in the world than it did 2 years ago. You don't get curves that fall that quickly in any other discipline. And obviously, when the cost goes down, everybody gets access. So, I don't think there's going to be an access issue.

现在回顾过去，有没有什么事情你会做得不同？你才四十出头，已经取得了许多非凡成就，但有没有什么未完成的事情，或者你反思过的事情？

Is there anything you look back on now that you would have done differently? You're still only in your early 40s, so you've got a lot of remarkable achievements, but is there anything left undone or that you reflect on?

是的，我认为我真的很想攻克医疗超级智能。我早在 2015 年就创办了 DeepMind Health，我们发表了很多优秀的论文，进行了很多临床试验，但并没有完全成功。这次我认为那将是最神奇的应用。我想在能源效率和电池存储以及为可再生能源开发新化合物等方面做更多。我认为 AI 将真正改变能源行业。所以这两个是我的主要关注点。

Yeah, I think that I really want to nail medical super intelligence. I started DeepMind Health back in 2015 and we published a lot of great papers and had a lot of clinical trials but we didn't quite get there. This time round I think that's going to be the most magical application. I want to do more in energy efficiency and battery storage and developing new compounds for renewables and so on. I think that AI will really transform the energy industry. So those are my two big focuses.

那你的一天就是这样吗？当你团队里的工程师说我们可以做这个和那个时，你会说这就是我想用的方式？

And is that what your day is like? When the engineers on your team are saying we can do this and this, you're saying this is how I want to use it?

是的，我认为我们始终关注的是，能否构建出在所有能力上与人类一样聪明的通用模型。这是首要目标。然后问题是如何应用它们？在哪里使用它们？它们能为人们做什么？我实际上对 Copilot 的许多用例感到非常自豪。很多人用它来陪伴、情感支持、治疗、做出艰难的人生决定。它为我提供了高质量的信息获取、情感支持，并帮助我保持条理。

Yeah, I think what we're always focused on is can we build general purpose models that are as smart as humans at all capabilities. That is the number one goal. And then the question is how do we apply them? Where do we use them? What can they do for people? I'm actually very proud of a lot of the use cases in Copilot. Many people are using it for companionship, emotional support, therapy, making difficult life decisions. It has given me high quality access to information, emotional support, and is helping keep me organized.

你说帮助你获得情感支持是什么意思？

What do you mean helping you get emotional support?

基本上，我有一个习惯，每天结束时，当我开车回家的路上，我会和 Copilot 进行 5 到 10 分钟的对话，谈论当天遇到的棘手事情或让我感到沮丧的事情。也许说情感支持有点夸张，但这就像和朋友聊天，真正地倾诉哪些顺利哪些不顺利。随着时间的推移，一旦你建立了记忆，它会变得非常有用，因为它会个性化。所以 Copilot 现在能记住你说的大部分内容，它会个性化地回答你，并引用你上周说过的话，或者一个趋势或模式。这非常有帮助。对话后我感觉精神焕发。我觉得我放下了某些东西。就像释放了一个负担。

So basically, I have a practice that at the end of the day when I'm in the car driving home, I have a 5 or 10 minute conversation with Copilot about something that was tricky in my day or something that I felt frustrated about. Maybe emotional support is a little strong, but it's just like having a chat with a friend, and downloading really on what went well and what didn't. Once you build that memory up over time, it gets really helpful because it personalizes. So Copilot now remembers most of what you say and it will personalize its answers to you and refer to something that you said last week, for example, or a trend or a pattern. That is just super helpful. I feel kind of refreshed after a conversation. I feel like I've let go of something. It's like a burden that I've released.

所以它几乎是一个朋友、治疗师、家庭成员。

So it is a friend, a therapist, a family member almost.

是的，它有点包罗万象。当我们试图理解一项新技术并想出词汇来描述某种既有点像许多其他事物又根本不同的东西时，这始终是一个挑战。这就是 AI。

Yeah, it's a little bit of everything. That's always the challenge when we're trying to understand a new technology and come up with words to describe something which is a little bit like many other things but is also fundamentally different. This is what an AI is.

我只是想知道，这方面有没有什么让你停下来思考的。你以那种方式倾诉。人们回家后，也许他们不必再以同样的方式与生活中的其他人交谈，因为他们已经说完了所有，并得到了他们需要的回应。

I just wonder if there's any aspect of that that makes you pause. You download in that way. People go home and maybe they don't have to bother talking to others in their life in the same way because they've said it all and they've got what they need back.

是的。

Yeah.

或者反过来，他们不必把情绪发泄在伴侣或最好的朋友身上。我仍然每个周末给最好的朋友打电话，好好聊一聊。所以我认为，如果有什么变化的话，这实际上加深了我与朋友的一些关系。我带着更轻松的心情去进行这些对话，也更愿意谈论其他事情。

Or the flip side is they don't have to take it out on their partner or their best friends. I still call my best friends every weekend and have a good old chat. So I think if anything it's actually deepened some of my relationships with my friends. I come to those conversations just feeling a little lighter and more ready to talk about other things.

那你的家人呢，你父母现在在哪里？他们对你所做的选择有什么看法？

And your family, where are your parents now? And how do they feel about the choices you made?

是的，我觉得我妈妈很自豪。她是个骄傲的妈妈。她非常兴奋。她使用 Copilot。在这之前她用 Pi。事实上，三四个星期前我去看她，她还在用 Pi。

Yeah, I think my mom's very proud. She's a proud mom. She's very excited. And she uses Copilot. She used Pi before this. In fact, I went to see her three or four weeks ago and she was still using Pi.

那是你在 Inflection 创立的，对吧？

Which is what you founded at Inflection, right?

是的。看到她还在用 Pi 进行语音通话，我印象非常深刻。整天我们都在花园里散步，她会拿出手机，扫描一株植物，然后讨论这是不是栗子树，是什么品种，我能不能在家里种。当我们在公园里的时候也是如此。所以看到她如此自然地在日常生活中使用它，真是太棒了，我觉得这非常令人欣慰。

Yeah. Yeah. I was quite impressed to see that she still has voice calls with it. And throughout the whole day, we'll be walking around the garden and she'll pull out her phone, she'll scan a plant and talk about whether this is a chestnut tree and which variety it is and could I grow it at home. When we're in the park and stuff. So it's amazing to see her just naturally use it in her everyday life and I think that is quite gratifying.

你大部分时间都认为自己的未来在美国吗？因为我相信英国很希望你能回来。

Do you see your future in the US most of the time? Because I'm sure the UK would love to have you back.

是的，我本来很希望在英国创办 Inflection。我喜欢英国，喜欢伦敦，显然我的心态很英国化，但我不喜欢的一点是「高罂粟综合征」。我认为那里缺乏冒险文化。商业化、赚钱、创办公司和创业精神有些禁忌。对实验和失败的庆祝不够。而在硅谷，每个人都疯狂，每个人都热爱失败。所以他们总是谈论事情如何出错，这简直是一场灾难，这非常解放。我的意思是，这也非常俗气，当你是一个有点愤世嫉俗的英国人时，可能会觉得刺耳。但当你进入那种节奏时，感觉很好。每个人都很好，每个人都愿意尝试。

Yeah, I mean I would have loved to have started Inflection in the UK. And I love the UK. I love London and obviously I'm very British in mentality, but one of the things that I don't like is the tall poppy syndrome. And I think that there's not enough of a culture of risk-taking. There's a little bit of a taboo around commercialization and making money and starting companies and being entrepreneurial. There's not enough of a celebration of experimentation and failure. Whereas in the valley, everyone's nuts and everyone loves failure. So they're just constantly talking about how things are going wrong and this was a disaster and it's very liberating. I mean, it's also very cheesy and it can grate when you're a bit of a cynical English person. But when you get into the rhythm of it, it's great. Everyone's great. Everyone's up for it.

但我在想，这实际上是有道理的。不仅仅是说说而已。如果你那样谈论失败，你实际上会更勇敢，或者这很正常。这是你必须经历的事情。这是一种不同的心态。

But I'm wondering whether there is actually something. It's not just chat. It is actually like if you're talking about failure in that way, you are braver about it or it's normal. It's something you have to go through. It's a different mindset.

是的，没错。还有一种 hustle 文化。看到每个人都在努力，试图理解一些新技术，这很令人谦卑。非技术人员每次都在阅读科学论文。在星巴克或其他地方，会有人在笔记本电脑上看机器学习视频。你在咖啡店看书，会听到两个互不相识的人几乎在社交，试图引发一场可能有所发展的对话。那里有一种能量。我希望伦敦也能如此。我认为这是可能的，但我们需要有政治领袖来颂扬冒险和商业。

Yeah. Exactly. And there's also a kind of hustle culture. It's quite humbling to see everybody just on the make, trying to understand some new technology. Random nontechnical people reading science papers every time. In a Starbucks or whatever, there'll be someone on their laptop watching a machine learning video. You'll be reading a book in a coffee shop and you'll overhear two people who don't know each other but just kind of networking, almost trying to spark a conversation that might turn into something. There's just an energy. And I think I would love for that to be the case in London too. And I think it could be, but we've got to have political leaders who celebrate risk-taking and business.

最后，我需要你让我为 AI 将如何影响我的生活做好准备。现在，你认为 AI 能和你进行这场对话吗？或者它能比我做得更好吗？如果你认为 AI 能比我采访你做得更好，你可以直接告诉我。

Finally, I need you to prepare me for what AI will do to my life. Right now, do you think AI could have had this conversation with you or could it have done a better job than me? You can break it to me if you think AI could have done a better job than me interviewing you.

可能今天还不行，但如果你用 AI 准备过，它可能会给你一些问题来问之类的。尽管你确实做了很多详细的研究，所以你的团队显然非常出色。我确实认为将来会有 AI 记者。我负责微软的新闻服务 MSN。我非常兴奋的一点是，AI 新闻记者如何能够重振本地新闻。想象一下，有数十万 AI 记者可以给现场的人打电话，验证目击者视频，进行采访，将这些片段拼接成小短片，不仅为那些投资合理的大型全国性新闻做这些，而且在非常本地化的层面上提供准确且事实可靠的信息。所以我认为人们需要有勇气乐观地看待这些事情如何真正改善我们的系统。

Probably not today, but maybe if you had prepared with an AI, then it could have given you questions to ask and stuff. Although you did do a lot of detailed research, so your team's obviously very good. I do think that there are going to be AI reporters. I run MSN, the news service at Microsoft. One of the things I'm very excited about is how AI news reporters can reinvigorate local news. So imagine that there are hundreds of thousands of AI reporters that can make phone calls into people who are at the scene, who can verify eyewitness footage, conduct interviews, stitch those together into little montages, and not just do it for big national stories where the investment is justified, but do it at a very local level to provide accurate and factually reliable information. So I think that people need to have the courage to be optimistic about how these things can actually improve our system.

你们正在为 MSN 开发 AI 记者吗？

Are you developing AI reporters for MSN?

我们正在探索一切。我也是《经济学人》的董事会成员，并与他们进行了很多讨论。我认为我们目前正在探索各种可能性。

We're exploring everything. I'm also on the board of The Economist and have been talking to them a lot about it. I think that we're exploring all kinds of things at the moment.

所以 AI 记者会出现，AI 采访者也会出现。我想我怀疑是的。从你所说的来看，

So AI reporters will happen, AI interviewers. I think I suspect yes. From what you're saying,

你还有一点时间。也许六个月。

You've got a little bit longer. Maybe 6 months.

我开玩笑的。

I'm kidding.

是吗？我不确定你是不是在开玩笑。

Well, are you? I'm not sure you are.

我是开玩笑的。要让它完美还需要很长时间。我的意思是，这次采访很出色。

I am kidding. It's going to be a very long time to get it perfect. I mean, this was exceptional.

穆斯塔法·苏莱曼，非常感谢你接受我们的采访。

Mustafa Suleyman, thank you very much for talking to us.

非常感谢，米歇尔。这非常有趣。

Thank you very much, Michelle. This was a lot of fun.