专访戴密斯·哈萨比斯的传记作者

Demis has a Nobel Prize. 戴密斯·哈萨比斯有诺贝尔奖。 Sam didn't finish his first degree. 萨姆·奥特曼没读完本科。 Therefore, Demis doesn't take Sam very seriously. 所以戴密斯不太把萨姆放在眼里。 Do you think he feels like or or will have to become more of a public figure than he is today? 你觉得他是否感觉到，或者说将来会不得不，比现在更多地成为一个公众人物？ Demis in some ways is too sensible. 戴密斯在某些方面太理智了。 What was the moment I guess your view of him kind of shifted the most? 在我们的对话中，他的哪个时刻让你对他的看法转变最大？ He would sometimes erupt in these conversations I was having and he would start banging the table and saying maybe if we approach science the right way, we understand more about nature, we will be getting closer to something that we could perhaps call God. 他有时会在我们的对话中突然爆发，开始拍桌子，说也许如果我们以正确的方式探索科学，更深入地理解自然，我们就会越来越接近某种可以称之为上帝的东西。 Sebastian Malibby spent over 30 hours with DeepMind CEO and co-founder Deis Abus in preparation to write his book, The Infinity Machine. Sebastian Mallaby 花了超过30个小时与 DeepMind 首席执行官兼联合创始人戴密斯·哈萨比斯深谈，为写作《The Infinity Machine》这本书做准备。 I got to sit down with Sebastian to get all his reflections on this experience. 我有幸和 Sebastian 坐下来，听他讲述这段经历的全部感悟。 Uh it was fascinating to talk about some of the stories that Sebastian reported on in this book, including the fact that Reed Hoffman offered a billion dollars to Demis to spin DeepMind out of Google at some point. 能和他聊聊书中那些故事实在太棒了，包括里德·霍夫曼曾经承诺出资10亿美元、帮助戴密斯将 DeepMind 从谷歌分拆出去这件事。 We talked about Demis' relationship with Elon and how that's evolved over the years, as well as what he thinks of Sam, Daario, and the general relationship between all the AI leaders in the space today. 我们聊到了戴密斯与埃隆·马斯克的关系以及这些年来它的演变，还有他对萨姆、Dario 以及当今 AI 领域所有领导者之间关系的看法。 And we hit on Sebastian's reflections on Demis, what motivates him, what his blind spots are. 我们也听到了 Sebastian 对戴密斯的观察，是什么在驱动他，他的盲点又在哪里。 Just a fascinating conversation with someone who's spent unprecedented time with someone who's really uh at the cutting edge of AI today. 能和一个与当今 AI 前沿最核心的人物花了前所未有时间交流的人对谈，真的非常精彩。 I think folks will really enjoy Sebastian's candidates here. 我觉得大家会非常喜欢 Sebastian 在这里的坦诚分享。 Without further ado, here he is. 废话不多说，有请。 Well, Sebastian, thanks so much for coming on the podcast. Sebastian，非常感谢你来参加播客。 Really appreciate it. 真的很感激。 Thank you. 谢谢。 It's great to be with you. 很高兴能和你聊。 I feel like The Infinity Machine uh and your kind of story of Deep Mind and Demesis Aabis is uh has captivated a bunch of folks in the AI world. 我感觉《The Infinity Machine》，你关于 DeepMind 和戴密斯·哈萨比斯的故事，已经深深吸引了 AI 圈里的很多人。 I know that I inhald this book. 我知道自己是一口气把这本书读完的。 I think I finished it in like 24 hours. 我想大概24小时内就读完了。 It's just a a gripping story and you got this just incredible access to Demis. 这本书太引人入胜了，而且你获得了如此难得的、接触戴密斯的机会。 I think you know uh the sitting in this pub for for multi-ours at a time uh talking about all sorts of things uh just just a fascinating you know uh opportunity and window into definitely one of the most interesting people and and uh companies in this AI race today and so uh appreciate you writing the book and appreciate you coming on the podcast to talk about it. 你知道，在那家酒吧里一次坐好几个小时，谈天说地，这是了解当今 AI 竞赛中最有趣的人和公司之一的绝佳窗口，真的很难得。感谢你写了这本书，也感谢你来播客聊这些。 So our goal today is to you know use your reporting to understand the people decisions and dynamics that shape the AI race that we're living through. 所以我们今天的目标是借助你的报道，去理解那些塑造了我们正在经历的 AI 竞赛的人、决策和动态。 And so maybe to start, I feel like one of the interesting themes in your book is whether this race that we have today between the labs was inevitable or could have gone differently uh if different players or decisions had had been done. 也许先从这里开始，我觉得你书中一个很有意思的主题，是今天实验室之间这场竞赛是否是必然的，或者如果不同的玩家做了不同的决定，是否可能走向另一条路。 Um I guess I'm curious after doing this work, do you think there was another path or or was this kind of inevitable? 在做完这些研究之后，你觉得存在另一条路吗，还是说这一切其实是注定的？ You know, I think it was inevitable. 我觉得这是必然的。 I think when you have this sort of supremely strong technology, there's going to be multiple labs in multiple countries that are just desperate to try and build it. 当你拥有这种极度强大的技术，自然会有多个国家的多个实验室拼命想把它造出来。 And we know the China stack, you know, is pretty strong. 而且我们知道，中国的技术栈已经相当强了。 And so despite the lack of semiconductors um they were going to have a go at it and they are actually doing pretty well. 所以尽管缺少芯片，他们还是会尝试，而且实际上做得相当不错。 And then clearly in the US uh and then in a few other countries you've got Mistral in France 然后显然在美国，还有其他几个国家，比如法国的 Mistral， Coher in Canada you know there was bound to be many players just as it's the technology is too sweet for it to be only interesting to one team. 加拿大的 Cohere，注定会有很多参与者，因为这项技术太诱人了，不可能只对一个团队感兴趣。 What's strange about this whole debate though is that's not how people saw it Xanti that you know when Demis was starting Deep Mind he really hoped that he could avoid the race dynamic. 但这场争论真正奇怪的地方在于，当初人们并不是这么看的。你知道，当戴密斯创立 DeepMind 时，他真心希望能避免这种竞争态势。 It seemed naive in retrospect, but that's what he hoped. 回头看这似乎很天真，但那就是他当时的希望。 One really compelling anecdote you had was I believe Demis in his interviews would say uh you know he was interviewing candidate and he'd say look at some point we may be really close to AGI and we'll go uh fly to a bunker and and and figure this all out as kind of like one team. 你书中有一个很有说服力的细节，我记得是说，戴密斯在访谈候选人时会问，你知道，在某个时刻我们可能真的非常接近 AGI 了，到时候我们会飞到某个掩体里，作为一个团队把这一切搞定。 Would you be willing to, you know, get on that flight and and and do that? 你愿意登上那架飞机，一起去做这件事吗？ You know, I think certainly part of the the founding origin of Anthropic is also this belief like, hey, we need to be the team that like gets, you know, as we get to the precipice of AGI, we're the ones figuring this out. 我想 Anthropic 的创立初衷在某种程度上也有这样一种信念：我们必须成为那个在接近 AGI 临界点时把关的团队。 you know clearly a a key part of starting all these companies. 这显然是创立所有这些公司的核心驱动力之一。 Do you think these folks I mean particularly Demis like do they still believe this or or how has their kind of thinking evolved uh having seen the way things have played out these past years? 你觉得这些人，尤其是戴密斯，现在还相信这一点吗，还是说看到这几年的发展后，他们的想法有所转变？ No, I think Demis has swung from one extreme to the other. 不，我觉得戴密斯已经从一个极端转向了另一个极端。 You know he began by thinking there could be a singleton scenario just one lab and by that he really meant deep minded himself. 他最初认为可能存在一个单一主体的场景，就是一家实验室，而且他真正的意思是 DeepMind 自己。 um uh to the opposite extreme where he now sees that there's a very crowded field and therefore that it's almost pointless for one lab to pursue safety by itself because if one lab is safe and then the other ones aren't, it doesn't make the world safer. 到了另一个极端，他现在看到这个领域已经非常拥挤，因此一家实验室单独追求安全几乎毫无意义，因为如果一家安全而其他家不安全，世界并不会因此更安全。 So he really has shifted to seeing this as a collective action problem that only a government can solve. 所以他真的转变了，把这视为只有政府才能解决的集体行动问题。 Because one thing I thought was really compelling from your book is obviously there was this early AI safety summit that Dennis and and the team put together and they're sharing you know the progress they've made and and why they're worried about it and you know some people in the room like Reed Hoffman, Elon Musk take that information and then they're like oh this technology really is starting to work. 你书中有一点我觉得很有力，就是那次早期的 AI 安全峰会，戴密斯和团队组织了这次峰会，分享了他们取得的进展以及他们为何感到担忧，而里德·霍夫曼、埃隆·马斯克这些在场的人拿到这些信息后，心想，这项技术真的开始有点起色了。 We should ourselves uh potentially do something around that. 我们或许也应该做点什么。 You know do you think uh obviously to to get safety to work going forward you need these different companies to to agree to share things and and collaborate. 要让安全真正发挥作用，你需要这些不同的公司同意分享信息、开展合作。 Uh I imagine that left a bad taste in in in Demis' mouth. 我想那次经历对戴密斯来说一定留下了不好的印象。 How do you think he and and other kind of key players in this space think about, you know, potentially sharing what they're doing uh with with given that, you know, past experience? 你觉得他和这个领域其他关键人物，在有过那段经历之后，如何看待潜在的信息共享这件事？ Yeah, I mean, you're totally right. 对，我是说，你说得完全对。 That was in 2015. 那是2015年的事。 Um, summer of 2015, they have this summit at SpaceX. 2015年夏天，他们在 SpaceX 举办了这次峰会。 Elon Musk is hosting it. 埃隆·马斯克主持。 And the idea is he's going to be brought into the tent by Deep Mind. 目的是把他拉进 DeepMind 的阵营。 He's going to be part of their efforts. 让他参与他们的工作。 He's going to be, you know, chairing this sort of safety oversight board. 让他主持一个安全监督委员会。 And therefore, he wouldn't set up a competitor. 这样他就不会再去创立竞争对手了。 And of course, at the end of 2015, he did set up a competitor. 当然，2015年底，他还是创立了竞争对手。 open AI and so that kind of really drove home the reality that we're going to have a race and I think if you ask now okay so what about future collaboration I think from's point of view or from any of the lab leaders OpenAI。这一切让我们清楚地看到，一场竞赛是不可避免的。我认为如果你现在问，好，那未来的合作呢，不管是戴密斯的角度，还是任何一位实验室领导者的角度， frankly um you can't trust the other guys uh and therefore the only way you get trust is if you have a government enforcer that comes along and say look here's the rules for everybody there's going to be a level playing field you're all going to have to you know abide by some sort of safety slow down you know pre-esting of models before you release them all that stuff and you all have to do it. 坦率地说，你没法信任对方。因此，获得信任的唯一方式，是由政府来执法，说好，这是对所有人的规则，要有公平的竞争环境，大家都必须遵守某种安全减速要求，在发布模型前进行预先测试，诸如此类，而且所有人都必须遵守。 Um and then the reaction of course will be yeah but what about the guys in China and and that's why ultimately I think this has to be a USChina collaboration you know remote though that prospect may seem to many listeners right now. 然后当然有人会说，那中国那边的人呢？这就是为什么我认为这最终必须是一个美中合作，尽管这个前景对很多听众来说现在看起来很遥远。 I mean do you think Demis and others 我是说，你觉得戴密斯和其他人 I think that's actually realistic. 我觉得这实际上是现实可行的。 I mean obviously there there's it it it you know is is such a a pressing problem but you know it doesn't seem like uh folks have the most confidence in in governments and certainly you know intergovernment uh collaboration across the world like it it intellectually makes sense that that is kind of the the clear way to solve this problem but do you think they think there's actually a probability this this does happen successfully? 显然这是一个非常紧迫的问题，但人们对政府的信心不太高，尤其是跨国的政府间合作，从智识层面来说，这显然是解决问题最清晰的路径，但你觉得他们真的认为这有实现的可能吗？ Well, you know, I think they ought to believe that it could happen. 嗯，我认为他们应该相信这是可以实现的。 And we do have the Food and Drug Administration and people complain that it's slow and all that, but but it does, you know, have expert reviewers who look at clinical trials and say, is this drug safe or not? 我们确实有美国食品药品监督管理局，人们抱怨它太慢之类的，但它确实有专业审查员去审查临床试验，判断这种药是否安全，是否可以上市。 And can it be released and if you can do that for pharmaceuticals and you should do it, AI models that can be more destructive on a grander scale than a drug certainly need to be reviewed Xanti. 如果能对药物做到这一点，就应该做，对于可能比药物造成更大范围破坏的 AI 模型，显然也需要接受审查。 And I think in Britain actually you have an example of a government AI safety institute that's pretty effective has some highass scientists. 我认为英国其实有一个相当有效的政府 AI 安全研究所的例子，里面有一些高水平的科学家。 I mean the chief scientist there is Jeffrey Irving who was at deep mind originally did a lot of the post training on the early models who's a really serious researcher was also at open AI by the way and a close collaborator of Darios. 那里的首席科学家是 Jeffrey Irving，他最初在 DeepMind，在早期模型的后训练上做了很多工作，是一位非常严肃的研究者，顺便说一句他也在 OpenAI 待过，是 Dario 的密切合作者。 Um so you've got serious people there and they have found vulnerabilities in systems in the past and quietly shared them with the private labs that have failed to find the same vulnerability vulnerabilities. 在那里，有真正优秀的人，他们确实在过去发现了系统中的漏洞，并悄悄地将这些漏洞分享给了未能自行发现同样漏洞的私营实验室。 So it shows that you can aggregate some technical expertise uh inside these public institutions because a certain number of folk feel motivated and public spirited enough to go work for the government. 这证明了你可以在这些公共机构内部积累一定的技术专业能力，因为有一定数量的人受到激励和公民责任感的驱动，愿意去为政府工作。 Um so I don't think we should give up on that. 所以我认为不应该对此放弃希望。 From the outside it's hard to tell. 从外部来看很难判断。 you know, it's almost like such a dispiriting story of, you know, Demis kind of being idealistic and then realizing this dynamic isn't going to play out. 这个故事有点令人沮丧，戴密斯带着理想主义出发，然后意识到这种动态不会按他期望的方式发展。所以在你书中的字里行间，很难看出他究竟是已经接受了这种现实，还是对政府间合作的实际可能性重新燃起了希望，毕竟你给的例子只是在单一国家内部，跨国协调的规模根本还没达到。 And so, it's hard to tell, you know, I guess reading between the lines in your book, whether he's kind of just resigned to this dynamic or is has found newfound hope in in like the the actual possibility of, you know, intergovernment collaboration because obviously even the examples you give are just within a single country, right? 而且当然，跨国协调的规模根本还没达到。 And certainly not at the scale of of of coordination across countries. 是的。 Yeah. 嗯。 I mean, what he what he says, I mean, the very last line of my book is, I'm optimistic still. 我是说，他说的，我书的最后一句话是：我仍然是乐观的。 And that's a quote from Demis trying to, you know, persuade me. 这是戴密斯说的，他是在试图说服我。 I've been I'm pushing him in in my interviews with him. 在采访中我一直在追问他。 Are you really optimistic? 你真的乐观吗？ Because it seems like we've got a race dynamic. 因为看起来我们已经陷入了竞争态势。 You know, in of course January 2025, you have Deep Seek come out and that really signals the advent of the Chinese labs into this space. 2025年1月，Deep Seek 横空出世，这真的标志着中国实验室进入这个领域。 And you know, I'm writing through 2025 into the end of 2025 and I'm saying, "Are you still optimistic?" 我一直写到2025年底，我问他，你还乐观吗？ And he says, "Yes, I'm optimistic still." 他说，是的，我仍然乐观。 And I think the reason for that insistence apart from like what what's he going to say is also that you know when something kind of dramatic happens the world does coales and take action you know if you look at covid the reaction to covid 我认为他坚持这个说法的原因，除了他能说什么，还在于，当发生某种剧烈变化时，世界确实会凝聚力量并采取行动。比如看看新冠，政府在应对新冠时 I mean governments on a national basis really did stuff that people didn't expect beforehand um you know so in a crisis people do react when you have something like you know trade where protectionism has a pretty slow and subtle effect on economic performance. 各国政府在国家层面确实做到了人们事先没有预料到的事情。所以在危机面前，人们确实会行动。相比之下，贸易保护主义对经济表现的影响相当缓慢而微妙， Um you don't really get a political reaction to try and fix that and that's what we've seen in the last few years. 没有真正引发试图修正的政治反应，这就是我们这几年看到的情况。 But when you when it if if you have an acute thing like mythos uh an Anthropic model comes out you know and people suddenly get worried you see the US government pretty much do a 180 from a less fair position to oh we better control this kind of position and I think that shows the shock effect is extremely important in determining whether you get government action or not. 但当你面对一件急性的事件，比如 Anthropic 的某个模型发布后人们突然开始担忧，你就会看到美国政府几乎来个180度转弯，从不管的立场变成我们得管这件事的立场。这说明冲击效应对于是否能激发政府行动极为重要。 Switching gears obviously a key part of you know in through line through your your whole uh work is just the story of of deep mind and it's ultimate relationship with Google and it's just an incredible uh story and so I guess as I recall you started this or you kind of got demis to agree to this like right before chatgpt and so obviously uh post alpha fold in some really interesting things but certainly the world uh evolved in a bunch of interesting ways from the start of your project people talk about demis and deep mind 换个话题，你整个工作贯穿始终的一条主线，就是 DeepMind 及其与谷歌最终关系的故事，这是一段非常精彩的历程。我记得你大概是在 ChatGPT 之前开始这个项目，在 AlphaFold 之后，当然后来世界发生了很多有意思的变化。现在人们谈论戴密斯和 DeepMind 的方式 all the time you know before you published this book but you kind of done the research 已经大不相同了，在你出书之前你就做了这些研究， what do you think 你觉得 the popular narrative got like most wrong about demis deep mind the way people would talk about, you know, the the two. 大众叙事对戴密斯和 DeepMind 最大的误解是什么？ Well, the extraordinary thing to me is how people basically discounted Demis. 让我最惊讶的是，人们基本上低估了戴密斯。 I mean, you know, I I've been interviewed a bunch of times since my book came out a few weeks ago, and, you know, often people can't pronounce his name. 我是说，我的书出版几周后接受了很多采访，很多人甚至不知道怎么读他的名字。 They say Deise, they say Hassabis instead of Habis. 有人叫他 Deise，有人把 Hassabis 读错成 Habis。 Uh, they kind of barely know who he is. 他们几乎不知道他是谁。 You know, my publisher, Penguin Press, was figuring out what to do the how to do the cover design, and they wanted to use this picture, but at the same time, they didn't think people would recognize him. 我的出版商企鹅出版社在考虑封面设计时，想用他的照片，但同时又觉得人们不会认出他来。 So you couldn't sell books based on some, you know, person nobody recognized 所以用一个没人认识的人的脸来卖书是行不通的 at this point. ，在这个时间点上。