📝 本节摘要：
在访谈的开篇，主持人Daniel介绍了Grady Booch作为软件工程奠基人之一的传奇背景。Grady随后分享了他正在筹备的纪录片项目——《计算与人类体验》（Computing the Human Experience）。该项目的灵感源于计算机历史博物馆（Computer History Museum）的一次对话，旨在打造一部计算领域的《宇宙》（Cosmos）。Grady希望能通过12集的篇幅，梳理计算与人类文明的交织历史，不仅作为他对这一行业的“情书”，更希望通过历史视角（如冯·诺依曼和图灵的早期思想）来缓解公众当下对生成式AI（Generative AI）的恐惧与迷茫，帮助人们在这个技术变革的时代做出更理性的决策。
[原文] [Daniel]: hey everyone this is Daniel coming up with another long form conversation and today my guest is Grady Buch he's a software engineer with more than 50 years of programming experience and he has been foundational to software engineering the modern version that we now enjoy he has worked on critical systems in every conceivable field and uh he's made significant contributions with programming languages and type systems and the UML design language and many many other things he has authored seven books and he's working on two more and he's also working on a documentary called Computing the Human Experience and we cover a lot in this conversation we cover good architecture good thinking principles AI and the role of the software engineer how to think software better communication styles education and a lot more it was a privilege to have him and I hope you'll enjoy the conversation at least as much as I did so please enjoy a fascinating conversation with Grady Buch grady thank you so much for making the time it's an honor to have you
[译文] [Daniel]: 大家好，我是Daniel，又给大家带来了一期长对话节目。今天的嘉宾是Grady Booch，他是一位拥有超过50年编程经验的软件工程师，他是我们今天所享有的现代软件工程的奠基人之一。他在你能想象到的各个领域的关键系统中都工作过，并在编程语言、类型系统、UML设计语言以及许多其他方面做出了重大贡献。他撰写了七本书，目前正在撰写另外两本，同时还在制作一部名为《计算与人类体验》（Computing the Human Experience）的纪录片。在这场对话中我们涵盖了很多内容，包括优秀的架构、优秀的思维原则、人工智能、软件工程师的角色、如何更好地思考软件、沟通风格、教育等等。能邀请到他是我的荣幸，希望大家能像我一样享受这次对话。那么请欣赏这场与Grady Booch的精彩对话。Grady，非常感谢你抽出时间，能邀请你来是我的荣幸。
[原文] [Grady]: my pleasure
[译文] [Grady]: 是我的荣幸。
[原文] [Daniel]: all right so you're known for many things uh and uh you have many significant contributions to the field of um uh software engineering and software architecture but the thing that I wanted to start with is your documentary project can you uh can you tell us what that's about and why you decided to make it
[译文] [Daniel]: 好的，您因许多成就而闻名，并且在软件工程和软件架构领域做出了许多重大贡献。但我想从您的纪录片项目开始聊起。您能告诉我们那是关于什么的吗？以及您为什么决定制作它？
[原文] [Grady]: i was on the board of trustees for the Computer History Museum for about a decade we had just hired our new CEO a gentleman by the name of John Holler john had come from Europe working with PBS and he'd landed a huge deal with the Gates Foundation which enabled the uh the foundation to actually proceed with building the physical displays that you see now so knowing his background I asked him hey John have you ever considered doing something like Carl Sean's Cosmos he paused and said Grady why don't you be our Carl and I paused and said I'm no Sean but that's an intriguing idea so I began the journey of exploring that very thing trying to do a documentary or explore producing a documentary that would examine the intersection of computing and what it meant to be human
[译文] [Grady]: 我曾在计算机历史博物馆（Computer History Museum）的董事会任职大约十年。当时我们要刚聘请了新的CEO，一位名叫John Hollar的先生。John来自欧洲，曾与PBS（美国公共广播公司）合作，他刚与盖茨基金会达成了一项巨额协议，这使得基金会能够真正着手建立你们现在看到的实体展区。得知他的背景后，我问他：“嘿，John，你有没有考虑过做一些像卡尔·萨根（Carl Sagan）的《宇宙》（Cosmos）那样的东西？”他停顿了一下说：“Grady，你为什么不成为我们的卡尔呢？”我也停顿了一下说：“我不是萨根，但这是个有趣的想法。”于是我开始了探索这件事情的旅程，试图制作一部纪录片，或者探索制作一部纪录片，以此来审视计算与“生而为人的意义”之间的交汇点。
[原文] [Grady]: i through the museum gave seven or eight different hour-long lectures just to try out the ideas and eventually came to the conclusion that there are 12 episodes that need to be told here ranging from at the very beginning uh what do we mean by computational thinking versus scientific thinking to the last episodes which address in the presence of uh computing and the computability of the mind which is a topic unto itself self we can speak of uh what does it mean to be human and we'll look at the journey through all of that so the idea being we'll do some history do some science but most of all look at the intersection of computing and humanity so that's the journey I've been on
[译文] [Grady]: 我通过博物馆举办了七八场时长一小时的讲座，只是为了测试这些想法。最终我得出结论，这里有12集的内容需要讲述。从最开始的——哪怕是“计算思维”与“科学思维”有何区别——一直讲到最后几集，探讨在计算和心智可计算性（computability of the mind）存在的情况下——这本身就是一个可以单独探讨的话题——“生而为人”意味着什么。我们将审视这其中的整个旅程。所以这个想法是，我们会讲一些历史，讲一些科学，但最重要的是审视计算与人性的交汇点。这就是我一直在进行的旅程。
[原文] [Grady]: um we worked with KQED the PBS station in the Bay Area they sponsored me for a couple of years and we did a trailer and a teaser and a treatment which we pitched to corporate PBS their reaction was "This is precisely the kind of series to which we aspire go find a producer." So the last several years I've been trying to find a producer co slowed things down uh at the same time I'm working on a book so you know one of these days I hope to have 12 one-hour episodes and a book that's what I'm working on in a way you can think of it as my love lo my love letter to computing it has given so much to me in my life now is my time to step back and express what I have found from it
[译文] [Grady]: 嗯，我们与湾区的PBS电视台KQED合作过，他们赞助了我几年。我们制作了预告片、先导片和一个剧本大纲，并向PBS总部进行了提案。他们的反应是：“这正是我们渴望的那种系列片，去找个制片人吧。”所以过去这几年我一直在试图寻找制片人，新冠疫情让事情慢了下来。与此同时，我也在写一本书。所以你知道，希望有一天我能完成这12集每集一小时的片子和这本书。这就是我正在做的事情。在某种程度上，你可以把它看作是我写给计算领域的一封情书。它在我生命中给予了我太多，现在是我退后一步，表达我从中领悟到什么的时候了。
[原文] [Grady]: i'm in a curious place because I began my career pre- internet i got my first email address in 1979 of all things that was on the arponet and I met many of the foundational folks in computing grace Hopper I didn't meet Turing he was dead by then but I met some of the people that worked with Alan Turing uh lot of folks like that um so that's what I'm working on
[译文] [Grady]: 我处于一个奇特的位置，因为我的职业生涯开始于互联网之前。我在1979年有了我的第一个电子邮件地址，居然还是在阿帕网（ARPANET）上。我见过许多计算领域的奠基人，比如格蕾丝·霍珀（Grace Hopper）。我没见过图灵（Turing），那时他已经去世了，但我见过一些曾与艾伦·图灵共事的人，以及很多像那样的人。嗯，这就是我正在做的事情。
[原文] [Daniel]: uh why do you think is it worth uh preserving this information uh of all the history of computing in this condensed 12-hour episode 12 episode 12 1-hour episode uh form and why uh do you believe is it worth investing so much time and so much effort uh if if my research is correct you've been working on this for many years more than uh 10 decade now yeah Yes why is it worth investing so much to preserve this information in this particular form what do you think
[译文] [Daniel]: 呃，为什么您认为值得用这种浓缩的12小时、12集、每集一小时的形式来保存这些信息——这些计算领域的历史？以及，为什么您认为值得投入这么多的时间和精力？如果我的研究没错的话，您已经为此工作了很多年，超过十年了吧？是的。为什么值得投入这么多去以这种特定的形式保存这些信息？您怎么看？
[原文] [Grady]: an educated population is better able to make decisions about its present and its future and so my intent is to provide that kind of information look today and I'm sure we'll get into it further in the discussion look today at what generative AI has done to the world and I I had a friend of mine she's a priest that that reached out to me shortly after the first chat GPT came and she said she was very worried like what does this mean for us is is super intelligence going to going to take us all over
[译文] [Grady]: 一个受过教育的群体能够更好地对其当下和未来做出决策，所以我的意图就是提供这类信息。看看今天——我相信我们会在之后的讨论中深入探讨——看看今天生成式AI（Generative AI）对世界做了什么。我有一位朋友，她是一名牧师，在ChatGPT刚出来不久后联系了我。她说她非常担心，比如这对我们意味着什么？超级智能是不是要接管我们所有人类了？
[原文] [Grady]: and so having had a perspective for the history of AI and the kinds of things that that John von Noman that Turring himself that others were saying in the early days of AI they are suspiciously important phrases that I think are relevant today as well so it's that kind of history that I think helps inform us and in a way if you look at it from the perspective of the history and science you will see the forces that both shape computing and how computing has shaped us as humanity so in many ways this documentary is an exploration of those forces that I hope will help the current populace understand what it means today
[译文] [Grady]: 因此，拥有AI历史的视角，了解约翰·冯·诺依曼（John von Neumann）、图灵本人以及其他人在AI早期所说过的那些事情——那些都是极为重要的论述，我认为它们在今天依然适用。所以正是这种历史，我认为能给我们带来启示。从某种意义上说，如果你从历史和科学的角度来看，你会看到那些既塑造了计算，又让计算塑造了我们人类的力量。所以在很多方面，这部纪录片是对这些力量的一种探索，我希望能以此帮助当下的民众理解这在今天意味着什么。
[原文] [Grady]: i'm trying to do something much like Sean did that doesn't just you know disappear after a year because oh my gosh this new thing was created but my my study has certainly revealed to me that there are some there's some common threads throughout the history of computing since really the beginnings of computing that still are with us today and therefore they can represent what the future might bring for us in the coming generations
[译文] [Grady]: 我试图做一些像萨根所做的那样的事情，它不会因为“天哪，又有新东西被发明出来了”而在一年后就消失。我的研究明确地向我揭示，自计算诞生之初起，整个计算历史中就存在着一些共同的脉络，它们至今仍伴随着我们，因此它们能够代表未来几代人可能面临的图景。
📝 本节摘要：
在本章中，Grady Booch 指出计算历史中存在着两条贯穿始终的脉络。首先是对“机器取代人类”的恐惧，他列举了20世纪40年代的书籍《巨型大脑》（Giant Brains）以及戏剧《R.U.R.》（“机器人”一词的起源）作为早期案例，表明这种恐慌并非今日特有。其次是计算作为“创造力扩音器”的角色。Grady 通过 FM 音频合成技术（催生了 Yamaha DX7 合成器）以及早期 AI 绘画系统（如 AARON）的历史证明，技术虽然引发了“艺术家失业”的担忧，但最终实际上为人类提供了前所未有的创作“颜料”和表达维度。
[原文] [Daniel]: can you give us any examples
[译文] [Daniel]: 您能给我们举一些例子吗？
[原文] [Grady]: sure um there's this great book published in the 1940s called Giant Brains and it came out around the time we started seeing mainframe computers this is prior to the IBM 370 and there was a lot of fear in the world that these things were going to take over uh humanity there of course was the great play RUR uh from which we get the term robots in which those themes were explored
[译文] [Grady]: 当然。20世纪40年代有一本很棒的书叫《巨型大脑》（Giant Brains），它出版于我们开始看到大型机出现的时候，那是早在 IBM 370 之前的事了。当时世界上弥漫着一种恐惧，担心这些东西会接管人类。当然还有那部伟大的戏剧《R.U.R.》（罗素姆万能机器人），我们从中获得了“机器人”（robot）这个词，剧中也探讨了这些主题。
[原文] [Grady]: and so I think if you look at computing from an economic perspective from a from a legal perspective from uh the perspective of what the technology says that we know we can and can't do I think this helps allay the fears of oh my god you know AIs are going to kill us all i'm not a PD doomer i'm anything against that and so I at least for me that knowledge of what has happened in the past like what the fear-mongering that happened in the 40s it's going to happen again and again uh if it bleeds it leads and so it's certainly catching a lot of press but by and large I know that yes there are some things that this is touching in terms of our human spirit but I know that there are limits to what computing and can and can't do so that's one good example giant brains great book i forget the author it's sitting over here on my bookshelf
[译文] [Grady]: 所以我认为，如果你从经济角度、法律角度，以及从技术本身告诉我们“能做什么”和“不能做什么”的角度来看待计算，这有助于减轻那种“天哪，AI 要把我们全杀了”的恐惧。我不是那种悲观论者（P(doom)er），我完全反对那种观点。所以，至少对我来说，了解过去发生的事情——比如40年代发生的那些散播恐慌的行为——会让你明白这种事会一再发生。正如俗话所说，“流血的新闻才是头条”（if it bleeds, it leads），所以它肯定会占据很多媒体版面。但总的来说，我知道，是的，这确实触及了我们人类精神的一些东西，但我同时也知道计算能做和不能做的事情是有界限的。这就是一个很好的例子，《巨型大脑》，一本好书，我忘了作者是谁了，它就放在我这边的书架上。
[原文] [Daniel]: all right as you can see I'm surrounded by books yep uh and uh those books definitely give you enough context to be able to speak with enough authority on on the subject and uh uh if anything you are if not the only one of the very few people with enough context and with enough authority to be able to do this and uh I admire this sort of project
[译文] [Daniel]: 好的，正如大家所见，我也被书籍包围着。是的，而且这些书绝对给了您足够的背景知识，让您能够在这个话题上以足够的权威发言。而且，如果不夸张地说，您即便不是唯一一个，也是极少数拥有足够背景和权威来做这件事的人之一。我很钦佩这类项目。
[原文] [Grady]: um well you know I I I write in order to learn there is so much that I don't know um in my exploration of computing and art for example um it addresses the issue of what what is creativity what does that really mean and is it possible to have computational creativity it led me to understand some of the earliest roots of generative AI which was you know not just the last several years but there were artists exploring computation as a pallet for our art you know back in the 50s and 60s
[译文] [Grady]: 嗯，你知道，我写作是为了学习。我有太多不知道的东西了。例如，在我探索“计算与艺术”的过程中，它触及了什么是创造力、这究竟意味着什么，以及是否可能拥有“计算创造力”的问题。这引导我去了解生成式 AI（Generative AI）的一些最早根源——你知道，这不仅仅是过去几年的事，早在50年代和60年代，就有艺术家在探索将计算作为艺术创作的调色板。
[原文] [Grady]: and so it's that knowledge that has helped me understand there are some exciting things today there's no doubt and there will still come exciting things but there's also this steady beat this relentless rhythm that computing has offered us and that's what I'm putting my ear toward and trying to illuminate how did that look like in the early in the 50s and 60s with uh regards to creativity and uh also with regards to this fear that computers will simply take over our the work that seemingly defines us and only us
[译文] [Grady]: 正是这些知识帮助我理解，毫无疑问，今天发生着一些令人兴奋的事情，未来也会有更多令人兴奋的事情发生，但也存在着计算赋予我们的这种稳定的节拍、这种无情的韵律。这就是我正在倾听的东西，并试图阐明：在50年代和60年代，关于创造力，以及关于“计算机将彻底接管那些看似定义了我们且仅定义了我们的工作”的恐惧，究竟是什么样子的。
[原文] [Daniel]: great questions so let's take an example from art um there is Chowers Chowing gosh I'm terrible with last names he's a gentleman from Stanford who I think it was in the 70s late '7s or thereabout that invented the notion of FM synthesis and from that was born the classic synthesizer of the era the Yamaha DX7 and the worry was oh my gosh we won't need you know artists anymore we won't need uh uh people to play any instruments of course that's not what happened
[译文] [Daniel] (注：此处应为Grady继续发言，Daniel可能是附和或并未打断，原文标注可能指代说话人转变，但内容显然是Grady的举例)
[译文] [Grady]: 好问题。那我们举一个艺术领域的例子。有一位叫约翰·乔宁（John Chowning）——天哪，我不擅长记姓氏——他是斯坦福大学的一位先生。我想是在70年代，大概是70年代末左右，他发明了 FM 合成（调频合成）的概念。由此诞生了那个时代的经典合成器——雅马哈 DX7（Yamaha DX7）。当时的担忧是：“天哪，我们不再需要艺术家了，我们不再需要人来演奏任何乐器了。”当然，这并没有发生。
[原文] [Grady]: um there was you know some legal issues in regards to that uh things like gosh if you're going to play on Broadway you still have to have a certain number of human uh performers and the like but what it did is that it opened up a creative pallet that brought music in a wider space than we could have ever imagined and that's I think the lesson again and again of what computing has done that yes we may fear it but it opens up opportunities in ways we may have never expected
[译文] [Grady]: 嗯，确实有一些相关的法律问题，比如“天哪，如果你要在百老汇演出，你仍然必须有一定数量的人类演奏者”之类的。但它真正做到的是，它开启了一个创造性的调色板，将音乐带入了一个我们要比以往任何时候能想象到的都更广阔的空间。我认为这就是计算所做之事的反复教训：是的，我们可能会害怕它，但它以我们从未预期的方式开启了新的机遇。
[原文] [Grady]: let's take a look at art again there's this delightful artist that I followed oh probably five six years now um Helen and again I'm blanking on the name um the way I look at her stuff she she built her own GANs and uh she trained her own stuff so you know this is good ethical use because she's not training it on copyrighted things it's all her own work
[译文] [Grady]: 让我们再看看美术。有一位令人愉快的艺术家，我关注她大概五六年了，嗯，海伦（Helen）——我又想不起名字了（注：指 Helena Sarin）。我看待她作品的方式是，她构建了自己的 GANs（生成式对抗网络），并且训练了自己的素材。所以你知道，这是良好的伦理使用，因为她没有在受版权保护的东西上训练，那全都是她自己的作品。
[原文] [Grady]: and the way I described it once I really got into her work was it's a lot like Michelangelo he was renowned in many ways but one of the things he did is he ground his own pigments that's exactly what she is doing she has this incredible artistic ability and her pigments are the GANs she has creating created that gives her this incredible pallet upon which she can do things that speak to us as humans so that's that's one of the examples that comes to mind
[译文] [Grady]: 当我真正深入了解她的作品时，我形容它的方式是：这很像米开朗基罗（Michelangelo）。他在很多方面都很著名，但他做的一件事是亲自研磨颜料。这正是她在做的事情。她拥有惊人的艺术天赋，而她的“颜料”就是她创造的那些 GANs。这给了她一个不可思议的调色板，在上面她可以做出能与我们人类产生共鸣的作品。这是我脑海中浮现的例子之一。
[原文] [Daniel]: nice gans the generative ad adversarial networks gen yeah generative adversarial network really this this is work that became before the diffusion stuff and before which is which fuels a lot of the current things like midjourney and and the like
[译文] [Daniel]: 不错，GANs，生成式对抗网络（Generative Adversarial Networks）。是的，这确实是在扩散模型（Diffusion）之前的工作，而扩散模型是现在像 Midjourney 之类工具背后的动力。
[原文] [Grady]: and even before that there was a gentleman in the 60s he had a a system called Aaron um of course the the 60s and 70s were a time in AI of u not large language models but knowledge systems which were rulebased systems uh they followed the symbolic systems that uh Newell and Simon and others developed during the 40s and 50s and people like Ed Fagenbalm pioneered the ideas of of knowledgebased systems which are if then else kind of things that's kind of like what this gentleman did for a system called Aaron
[译文] [Grady]: 甚至在那之前，在60年代有一位先生，他有一个叫 AARON 的系统。当然，60年代和70年代是 AI 的一个时期，那时不是大语言模型，而是知识系统（Knowledge Systems），即基于规则的系统。它们遵循了 Newell 和 Simon 等人在40年代和50年代开发的符号系统，以及像爱德华·费根鲍姆（Ed Feigenbaum）这样的人开创的知识库系统的理念——也就是那种“if-then-else”之类的东西。这有点像那位先生为 AARON 系统所做的事情。
[原文] [Grady]: and there's a great display at the computer history museum on it but it's really amazing art that can speak to you and and so you know we see these kinds of developers over and over again where new algorithms are created that create new opportunities for us to express our feelings in in new pallets
[译文] [Grady]: 在计算机历史博物馆有一个关于它的很棒的展览，那真的是能与你对话的惊人艺术。所以你知道，我们一次又一次地看到这类开发者，新的算法被创造出来，为我们用新的调色板表达情感创造了新的机会。
📝 本节摘要：
Grady 在本章中提出了纪录片的另一个核心主题：计算作为奴役工具的危险。他明确表示，他并不担心“超级智能”本身的崛起，而是担心那些贪婪且缺乏同理心的个人与公司利用计算技术攫取权力。Grady 引用了古希腊神话中的青铜巨人塔洛斯（Talos）和戏剧《R.U.R.》，指出人类自古以来就有“扮演上帝”的执念。他点名批评了 OpenAI 的 Sam Altman 等科技领袖，认为他们表现出了极度的傲慢与伪善——一方面声称担心 AI 的风险，另一方面却以“天选之人”自居，执意推进开发。尽管如此，Grady 对人类精神的韧性充满信心，认为我们能够经受住这次挑战。
[原文] [Daniel]: this is very interesting this is fascinating so I'm I'm already seeing two themes in your documentary one is the fear that computers might take over what makes us human and two the ability to express ourselves with the help of computers uh is there any other such theme that uh permeates the the thread of your documentary
[译文] [Daniel]: 这非常有意思，太迷人了。所以我已经在您的纪录片中看到了两个主题：一是对计算机可能会接管那些“生而为人”的特质的恐惧；二是借助计算机来表达自我的能力。除此之外，还有其他贯穿您纪录片脉络的主题吗？
[原文] [Grady]: there's another one and this relates to the danger of computing as a tool for subjugation i don't fear the rise of super intelligence i do fear the rise of the rapacious non-empathetic individuals and companies who seek to use computing as a means of power and we see this again and again this is the theme of rur the the the the play upon robotics
[译文] [Grady]: 还有一个主题，这与“计算作为一种奴役工具的危险”有关。我不担心超级智能（Super Intelligence）的崛起，但我确实担心那些贪婪、缺乏同理心的个人和公司的崛起，他们试图利用计算作为获取权力的手段。我们一次又一次地看到这种情况。这就是《R.U.R.》——那部关于机器人的戏剧——的主题。
[原文] [Grady]: we see it even if you go back to uh Greek mythology one of the earliest examples of AI was a uh a clay figure called Talos and I think this is in the Odyssey but Talos was a clay figure that walked around Cyprus and he would hurl stones at anybody that came too close he was powered by some magic elixir put into his heels the Greeks had a real obsession with heels but that's another topic unto itself
[译文] [Grady]: 我们甚至可以追溯到希腊神话。AI 最早的例子之一是一个名叫塔洛斯（Talos）的黏土人像。我想这出现在《奥德赛》里（注：实为《阿尔戈英雄纪》），塔洛斯是一个在塞浦路斯岛巡逻的黏土巨人，他会向任何靠得太近的人投掷石块。他是靠一种注入脚后跟的魔法药水驱动的。希腊人对脚后跟真是有种执念，但这又是另一个话题了。
[原文] [Grady]: but it it points out that there's something within us as humans in which we seek to become gods and that's I think where there we're resonating with current AI because there are some uh and Sam at OpenAI and others who I think are are really believe that we are on the path of creating our gods
[译文] [Grady]: 但这指出了我们人类内心的一种特质：我们试图成为神。我认为这也是我们与当前 AI 产生共鸣的地方，因为有一些人——比如 OpenAI 的 Sam（Sam Altman）以及其他人——我认为他们真的相信我们正走在创造我们自己的“神”的道路上。
[原文] [Grady]: of course the problem he has and this is the theme I deal with here you have a few individuals who think they are the chosen ones they know the answer and trust us in this process of course I don't trust them in this process and they don't have a solution to it and it's it's incredible hubris and and hypo hypocrisy on Sam's part where he says gosh you know we worry about all these things but you know we're going to build it anyway
[译文] [Grady]: 当然，他（Sam）的问题——这也是我要在这里探讨的主题——在于你有少数几个人认为他们是“天选之人”，他们知道答案，并让我们在这个过程中“相信他们”。当然，我在这个过程中并不信任他们，而且他们对此并没有解决方案。Sam 表现出了极度的傲慢（hubris）与伪善（hypocrisy），他说：“天哪，你知道我们担心所有这些事情，但你知道吗，无论如何我们都要把它造出来。”
[原文] [Grady]: and you see this you've seen this again and again in the history of computing and that's why I'm not freaking out over it because I have seen it again and I am I have great confidence in the resilience of the human spirit to endure
[译文] [Grady]: 你会看到这种情况，在计算的历史中你已经一次又一次地看到这种情况了。这就是为什么我没有对此感到惊慌失措，因为我已经见过这种事了，而且我对人类精神承受这一切的韧性（resilience）充满信心。
[原文] [Daniel]: well uh I share your your uh optimism of course with not nearly enough context and uh uh understanding and knowledge as as you have but uh it's really great to have you share some of that here
[译文] [Daniel]: 嗯，我分享您的这种乐观态度。当然，我不具备您那样丰富的背景、理解和知识，但真的很高兴能听您在这里分享这些。
[原文] [Grady]: uh I just well keep in mind sorry to interrupt but keep in mind that I have a perspective there are computing is a huge thing and the the the tagline I have is the story of computing is a story of humanity and there are many many stories I'm telling one story and it comes from the experience I've had inside the industry but also as somebody who you know cares about what it means to be human so it's a story you've got a story too and you should tell it
[译文] [Grady]: 呃，我只是——抱歉打断一下——请记住，我这只是“一种”视角。计算是一个巨大的事物，我的标语是：“计算的故事就是人类的故事。”这其中有许许多多的故事。我正在讲述其中一个故事，它源于我在行业内部的经验，也源于作为一个关心“生而为人的意义”的人的思考。所以这是一个故事。你也有一个故事，你也应该把它讲出来。
📝 本节摘要：
主持人 Daniel 高度赞扬了 Grady 将复杂技术概念讲得精准且极具细微差别的能力。Grady 透露，他在沟通上深受三位未曾谋面的导师影响：卡尔·萨根（Carl Sagan）教会他将科学与哲学交织；理查德·费曼（Richard Feynman）教会他“如果不能向五岁孩子解释清楚，就说明你自己没懂”；约瑟夫·坎贝尔（Joseph Campbell）则让他学会欣赏人类体验的壮丽与仪式感。此外，Grady 回忆了自己在德克萨斯州阿马里洛（Amarillo）的成长经历，特别是高中时期的辩论课和“即兴演讲”训练，那是他沟通能力的真正基石。他还打趣道，自己在做 TED 演讲时因为拒绝死记硬背而“激怒”了主办方，因为他更喜欢根据观众的反应即兴调整。
[原文] [Daniel]: one thing that I admire a lot about you uh as a uh as a key figure in uh in this industry is your ability to express yourself and communicate to such a precise and nuanced degree that I hope to be 10% of the communicator you are even though I'm uh I am a teacher and I want this is one of my goals to be able to express complex ideas with enough clarity preciseness and nuance um and I wanted to ask about this particular skill did you happen to grow it in any way uh with intention or did that just occur along the way how did that how did that skill grow
[译文] [Daniel]: 作为这个行业的一位关键人物，我非常钦佩您的一点是您的表达能力，您能以如此精准且充满细微差别的方式进行沟通。虽然我是一名教师，但我希望哪怕能达到您沟通能力的 10% 就好了。这也是我的目标之一：能够以足够的清晰度、精确度和细微差别来表达复杂的想法。我想问问关于这项技能，您是有意去培养它的吗？还是说这一路走来自然就形成了？这项技能是如何成长的？
[原文] [Grady]: that's a great question and and thank you i'm I'm deeply flattered by your observations i I try to communicate well there are three people whom I've never met that have been a big influence upon me in that particular topic and not necessarily in any order of importance there's of course Carl Sean who is just a tremendous educator and his ability to weave science with deep philosophical ideas i aspire to his degree of fluency
[译文] [Grady]: 这是个好问题，谢谢你。你的观察让我深受感动（受宠若惊）。我确实努力去做好沟通。有三位我从未见过面的人在这一特定话题上对我有很大影响，排名不分先后。当然有卡尔·萨根（Carl Sagan），他是一位了不起的教育家，他将科学与深刻的哲学思想编织在一起的能力令人惊叹，我渴望能达到他那种流畅度。
[原文] [Grady]: uh the second is Richard Feineman and I especially resonate with Dr feman's notion that if you can't explain a topic to like a 5-year-old or you know some normal person off the street then you probably don't understand it yourself and indeed this is something I've struggled with i'm so deep in the technology um one of the things I've had to develop over the years is to make sure that I do express these things that may have some very deep complicated elements to them but in a way they're understandable and if I don't do that then I have utterly failed
[译文] [Grady]: 第二位是理查德·费曼（Richard Feynman）。我特别认同费曼博士的观点：如果你不能把一个话题解释给一个五岁的孩子，或者是街上的普通人听，那你可能自己也没弄懂。这确实是我一直在努力克服的事情，因为我陷在技术里太深了。多年来我必须培养的一件事就是，确保在表达这些可能包含非常深奥复杂元素的事物时，能以一种人们可理解的方式呈现。如果我不这样做，那我就彻底失败了。
[原文] [Grady]: and the third person who's who's very much influenced my thinking is is Joseph Campbell um in his work with uh you know the hero of a thousand faces uh the transformation of myths through time these folks and I think especially what I learned from him is just stepping back and opening my aperture to the exquisite beauty and wonder of what it means to be human and and and the rituals we heal humans have produced the history of gone through these are all things that just you know thrill me and I'm hoping from what I learned from him to express that same kind of joy uh to the people to whom I speak
[译文] [Grady]: 第三位对我的思维影响很大的人是约瑟夫·坎贝尔（Joseph Campbell），比如他在《千面英雄》（The Hero with a Thousand Faces）以及神话随时间演变方面的著作。对于这些人，尤其是从坎贝尔身上，我学到的是退后一步，打开我的光圈，去通过它欣赏“生而为人”意味着什么的绝美与奇迹，以及我们人类创造的仪式和经历的历史。这些都是让我感到兴奋的事情。我希望通过从他那里学到的东西，将同样的快乐传达给我的听众。
[原文] [Grady]: and for the nuance I mean my gosh this is complicated stuff we're dealing with here and so I try to be careful in what I say because there are very few absolutes and this This bothers me by again what Sam and Elon and many others say because they they view the world in a black and white for me no the world is not that it is full of nuance and that's part of the wonder of being human
[译文] [Grady]: 至于细微差别（Nuance），天哪，我们处理的东西太复杂了。所以我尽量在言辞上保持谨慎，因为绝对的东西很少。这也正是 Sam（Altman）和 Elon（Musk）等人的言论让我困扰的地方，因为他们用非黑即白的视角看世界。对我来说，不，世界不是那样的，它充满了细微差别，而这正是生而为人的奇妙之处。
[原文] [Daniel]: did you happen to practice this sort of communication in any way and gather feedback and uh um exercise this sort of muscle to be able to to speak and communicate clearly i mean even in the conversation that we're having right now with questions that you haven't seen you seem to speak with a degree of fluency and um preciseness that I've seen you speak in your TED talk and TED talks are very uh are famous for being so um intensely rehearsed um so I'm wondering is there any degree of practice to the sort of muscle that you have done
[译文] [Daniel]: 您有没有以某种方式练习过这种沟通技巧，收集反馈，锻炼这种“肌肉”，以便能够清晰地演讲和交流？我是说，即使在我们现在进行的这场对话中，面对您没见过的问题，您似乎也能以一种我在您的 TED 演讲中看到的那种流畅度和精确度来回答。而 TED 演讲以经过高强度的排练而闻名。所以我很好奇，您是否对这种能力进行过某种程度的刻意练习？
[原文] [Grady]: i'll answer that question but let me go back to TED talks because I gave a TED talk and I absolutely infuriated the TED people because I was trying to do some things out of the box there's a particular rhythm and pace and style and and and and storytelling within within TED and u I annoyed them by I would for my practices I you know memorized everything I didn't write everything down most of the stuff I just you know sort of do off the cuff and that'll come back my story in a moment and so in all my rehearsals it was always the same thing of course when I was on stage I feed off of the audience and so much to their dismay I let that come in and let the audience adapt with what I was doing
[译文] [Grady]: 我会回答这个问题，但让我先回过头说说 TED 演讲。我做过一次 TED 演讲，当时我彻底激怒了 TED 的工作人员，因为我试图做一些不按套路出牌的事。TED 有特定的节奏、配速、风格和讲故事的方式。我惹恼了他们，因为在练习时……你知道，我没有把所有东西都写下来死记硬背，大部分内容我都是即兴发挥的——这跟我接下来的故事有关。所以在所有的彩排中，每次都不太一样。当然，当我站在台上时，我会从观众那里汲取能量。让他们非常沮丧的是，我让这种互动介入了进来，并根据观众的反应调整了我的表现。
[原文] [Grady]: so where did that come from um I grew up in Amarillo Texas a small town of some 100 plus people a classic cold war town uh Amarillo was in the panhandle of the of of Texas uh it was known mostly for oil and gas and during the cold war end of World War II the beginning of cold war where the Russians were the bad guys Ammerlu Air Force Base was established because it was a place where strategic air command aircraft could come there it was also the place where we had the Pentax plant still exists and the Pantex plant is the one place in all the United States in which the triggers for all of our nuclear weapons are assembled and disassembled so it's got this curious rural uh character but also you know really lots of governmental things going on as well too
[译文] [Grady]: 那么这种能力从何而来？我在德克萨斯州的阿马里洛（Amarillo）长大，那是一个有十多万人的小城，一个典型的冷战时期城镇。阿马里洛位于德州的狭长地带（Panhandle），主要以石油和天然气闻名。在冷战期间——二战结束、冷战开始，俄国人是坏蛋的时候——阿马里洛空军基地建立了，因为战略空军司令部的飞机可以降落在那里。那里还有 Pantex 工厂（现在还在），那是全美国唯一一个组装和拆解所有核武器引信的地方。所以那里有一种奇特的乡村特质，但同时也发生着很多政府层面的大事。
[原文] [Grady]: my dad of all things was a sheriff so it's like you know wow i remember coming back from the Air Force Academy once on on some break you know the academy is a pretty pretty brutal place and he thought my hair was too long from that even but that's another story anyway in high school I took some uh some debate classes and what really helped me were some uh spontaneous lecture classes as well too where you'd be given a topic you had five minutes and you had to talk on it right then so I think that's when I began to develop my skills when I was what 14 15 uh I and I still remember my my debate instructor and my speech instructor i thank them constantly for that because that's where the foundation was laid
[译文] [Grady]: 我的父亲居然是个治安官（Sheriff），所以就像是……哇。我记得有一次放假从空军学院回来——你知道，军校是个相当残酷的地方——但他甚至觉得我的头发太长了（尽管已经很短了），但这又是另一个故事了。总之，在高中时，我参加了一些辩论课程。真正对我帮助很大的是一些“即兴演讲”课（spontaneous lecture classes）：给你一个题目，你有五分钟时间准备，然后必须立刻就此发表演讲。所以我认为那就是我开始发展这些技能的时候，大概是在我 14、15 岁的时候。我至今仍记得我的辩论老师和演讲老师，我不断地感谢他们，因为那是打下基础的地方。
📝 本节摘要：
在本章中，Grady 详细拆解了他处理不同类型工作的独特流程。对于会议演讲，他将其形容为“简单”且“未经蒸馏”的——几乎不排练，主要依靠脑海中的知识储备和极简的图片辅助。然而，写书对他而言则是一只“完全不同的野兽”。他分享了一种极具画面感（甚至有些复古）的知识管理法：建立私人网站搜集数万条笔记，然后将它们打印出来剪成纸条，坐在地板上进行物理分类以构建叙事结构。此外，他特别强调了其新书的目标读者是“充满求知欲的大众”，并致力于挖掘计算历史中被硅谷主流叙事所忽略的全球视角（如中国、印度、非洲的贡献），拒绝只写关于“已故白人男性”的历史。
[原文] [Daniel]: and we all appreciate you for it um do you have a process for writing and for delivering delivering a a good conference talk
[译文] [Daniel]: 我们都为此非常感激您。嗯，您对于写作和发表一场精彩的会议演讲，有什么特定的流程吗？
[原文] [Grady]: so let's separate those because there's a different process from giving a t conference talk versus writing a book two very different things and there's a third one which is you know writing a short form article about for about 10 years I wrote a column for it e software the first one was uh on architecture and I switched it over to on computing I did what 30 40 50 of those kinds of things
[译文] [Grady]: 让我们把这些分开来说，因为发表会议演讲和写书是完全不同的流程，这是两件截然不同的事。还有第三种，就是写短篇幅的文章。大概有10年时间，我为 IEEE Software 写专栏，最初是关于架构的，后来转为关于计算的。我大概写了30、40甚至50篇那样的东西。
[原文] [Grady]: so let's take each each of those for a a lecture at a a conference of some sort those turn out to be easy for me because in most cases I'm simply talking about the things that I'm currently working on so I will uh I I do very little research on them because it's mostly in my head i'll start out with a basic storyboard and then I'll do pictures in fact if you look at most of my slides there's very little text on them it's mostly a large picture and then I talk about them and I never rehearse i just sort of do it because those pictures remind me of what I need to say
[译文] [Grady]: 那么让我们逐一来看。对于某种会议上的讲座，这对我来说很容易，因为在大多数情况下，我只是在谈论我目前正在做的事情。所以我几乎不需要做什么研究，因为内容主要都在我脑子里。我会从一个基本的故事板开始，然后我会做图片。事实上，如果你看我的大部分幻灯片，上面几乎没有文字，主要是大幅的图片，然后我就看着它们讲。而且我从不排练，我就那样直接讲，因为那些图片会提醒我需要说什么。
[原文] [Grady]: uh you know a preparing for an hourlong lecture I don't know I might spend 30 minutes something like that and I don't mean to short change my audience but it's like you know this is what's in my head so you're getting getting raw undistilled grady
[译文] [Grady]: 呃，你知道，准备一个一小时的讲座，我不知道，我可能只会花30分钟之类的。我不是要有意敷衍我的听众，但这就像是，你知道，这就是我脑子里的东西，所以你们看到的是原汁原味、未经蒸馏的 Grady。
[原文] [Grady]: for the articles it's not unlike that especially with the things with it e software they gave me degrees of freedom to write what I wanted to so in most cases I would write something that I was currently researching and u and I I have this sort of portfolio of of notes and things like that so I'd go through and say "What do I want to say here what's triggered me?" And I'd write those down and the the story would sort of form itself
[译文] [Grady]: 对于文章，情况也差不多，尤其是给 IEEE Software 写的东西，他们给了我写我想写的东西的自由度。所以在大多数情况下，我会写一些我目前正在研究的东西。我有一种笔记之类的作品集，所以我会浏览一遍然后说：“我在这里想说什么？什么触发了我的灵感？”然后我会把这些写下来，故事就会自然而然地形成了。
[原文] [Grady]: now the book I'm working writing on currently very very different beast because you can see I've got all these books around me and I'm surrounded by these because there is so much I know I don't know uh let's take for example the book on uh on computing and um science mhm so I mean gosh there's an obvious connection between computing and science but gee one of the rabbit holes I went down was where did we start with uh the use of computing for meteorological and climate studies turns out it was all the way to back to John von Noman
[译文] [Grady]: 至于我目前正在写的书，那完全是一只非常、非常不同的野兽。你可以看到我周围全是这些书，我被它们包围着，因为我知道我有太多不知道的东西了。呃，就拿那本关于“计算与科学”的书来说吧。嗯，计算和科学之间显然有联系，但我钻进去的一个“兔子洞”是：我们是从哪里开始将计算用于气象学和气候研究的？结果发现这可以一直追溯到约翰·冯·诺依曼（John von Neumann）。
[原文] [Grady]: and there is this really interesting history of how we grew from just those ideas which actually fueled the creation of supercomputers which also helped in of all things the testban treaty during the Reagan days which is an interesting connection with uh with computing and conflict to even today where we're looking at less of these symbolic kinds of systems for meological studies to ones that are more neuralbased
[译文] [Grady]: 这其中有一段非常有趣的历史：我们是如何从那些想法发展起来的？这些想法实际上推动了超级计算机的诞生，而超级计算机居然还在里根时代的《禁止核试验条约》中发挥了作用——这是计算与冲突之间的一个有趣联系。直到今天，在气象研究中，我们要么较少关注那些符号类的系统，要么转向更多基于神经网络的系统。
[原文] [Grady]: so in those cases there's a lot that I know I don't know which is why I've surrounded myself with books like this uh especially for this particular book I set up my own website and over the last multiple years I've gosh I've read you know hundreds if not thousands of articles i've you know not read every word of all these books here but read certainly most of the important pieces of each of them and along the way I take notes which I put on this website and divided out chapter by chapter
[译文] [Grady]: 所以在这些情况下，有很多我知道自己不知道的东西，这就是为什么我把自己包围在这样的书堆里。特别是对于这本书，我自己建立了一个网站。在过去的很多年里，天哪，我读了数百甚至数千篇文章。你知道，我没有读完这里所有书的每一个字，但我肯定读了每本书中大部分重要的部分。在这个过程中我会做笔记，我把它们放在这个网站上，并按章节划分。
[原文] [Grady]: so when I go and write a chapter then what I'll do is I'll go back to those notes there are probably maybe 10,000 notes I have per chapter i will print out the interesting ones and then I'll cut them up and I'll sit in the middle of the floor and I begin to sort them and the story sort of follows itself from there
[译文] [Grady]: 所以当我要去写某一章时，我会回过头去看那些笔记。每一章我大概有10,000条笔记。我会把那些有趣的打印出来，然后把它们剪开。我会坐在地板中间，开始对它们进行分类，故事也就从那里顺其自然地展开了。
[原文] [Daniel]: is this uh that's how nice is this book um technical in nature so for a technical audience or for more of a lay person audience
[译文] [Daniel]: 这是……这太棒了。这本书本质上是技术性的吗？是面向技术读者的，还是更多面向普通读者的？
[原文] [Grady]: it is so I'm writing two books right now there's one book for the technical audience which is a book about software architecture and uh we can go into that detail later but the primary book I'm I'm working on the computing effort is really for the general public or as I like to describe it it is for the the intellectually curious public who wishes wish to know about the roots of computing and where it might heads where it might head
[译文] [Grady]: 是这样的，我现在正在写两本书。一本是给技术读者的，是关于软件架构的书，我们可以稍后再细聊那个。但我正在致力于“计算”项目的那本主要著作，实际上是面向大众的，或者如我喜欢描述的那样，它是面向那些“充满求知欲的大众”（intellectually curious public），那些希望了解计算的根源以及它可能去向何方的人。
[原文] [Grady]: uh so you know your mother and father might be interested in it i'm trying to write it at a level where it wouldn't overwhelm them but it could perhaps tell them you know why is my son doing this crazy stuff but hopefully it will speak to them
[译文] [Grady]: 呃，所以你知道，你的父母可能会对它感兴趣。我正试图把它写成这种水平：既不会让他们感到吃力，又能或许告诉他们“为什么我的儿子在做这些疯狂的事情”，希望能与他们产生共鸣。
[原文] [Grady]: the other thing I have to realize is that computing is is a global topic yeah there are stories that cover China and India and South America and Africa and I'm trying to be sensitive to those voices because I don't want to end up with something that's just you know old dead white guy stuff out of Silicon Valley if anything I've learned in my research the the story of computing is so much richer than the verbiage coming out of Silicon Valley right now and I want to honor the wonder and beauty of that
[译文] [Grady]: 我必须意识到的另一件事是，计算是一个全球性的话题。是的，这里有关于中国、印度、南美和非洲的故事。我正努力对这些声音保持敏感，因为我不希望最终写出来的东西只是，你知道，那些来自硅谷的“已故白人男性”的东西。如果说我在研究中学到了什么，那就是计算的故事远比现在硅谷流出的那些言辞要丰富得多，我想以此向那种奇迹与美丽致敬。
📝 本节摘要：
针对目前对“四人帮”（Gang of Four）设计模式过于僵化和繁琐的批评，Grady 直言这些批评是合理的。他强调，包括他一手缔造的 UML 在内，工具的初衷是为了辅助推理，绝大多数图表画完即扔，而非为了留存僵死的文档。Grady 给出了他经典的架构定义：“架构是设计决策的集合，其重要性由变更成本来衡量。” 他进一步引用了圣塔菲研究所（Santa Fe Institute）关于复杂系统的理论以及天文学家 George Ellis 的观点，指出模式应当像科学发现宇宙规律那样被视为“自然的向导”，而非绝对的教条，过度死守模式只会导致系统变得脆弱。
[原文] [Daniel]: well uh this uh conversation is probably going to be listened to by mostly programmers so software engineers so it would be uh uh I would be a miss if uh I didn't ask about software architecture and uh software engineering as it's changed over the years um software architecture it's a good idea yeah so one thing that I was curious about is that uh there has been a recent conversation uh going online in many different shapes and uh forms sure uh about the gang of four design patterns that they're taking a lot of heat for being very overly rigid or uh too much ceremony and uh taking a lot of uh heat these days do you think that's warranted or is there more nuance to it
[译文] [Daniel]: 嗯，这场对话的听众可能主要是程序员，也就是软件工程师。所以如果我不问问关于软件架构以及软件工程这些年来的变化，那就是我的失职了。软件架构，这可是个好话题。是的，所以我很好奇的一件事是，最近在网上有各种形式的讨论，关于“四人帮”（Gang of Four）的设计模式。它们正受到很多抨击，被认为过于僵化，或者形式主义太重（too much ceremony）。您认为这种抨击是合理的吗？还是说这其中有更多的细微差别？
[原文] [Grady]: it is it's warranted and it's the same thing is true with even the UML that to if you take it in very dogmatic ways then it loses the purpose of it uh I'll talk about patterns in a moment but let's go back to the UML i mean the words I had that I wrote at the very opening of the standard still hold true today it's a language meant for reasoning about systems and most of the diagrams you create from it you want to throw away because its value is is the scaffolding and understanding and reasoning about the decisions you make
[译文] [Grady]: 它是合理的，这种批评是合理的。即使对于 UML（统一建模语言）也是如此，如果你以非常教条的方式去使用它，那么它就失去了原本的目的。我一会儿再谈模式，让我们先回过头说说 UML。我是说，我在标准开篇写下的那些话今天依然适用：它是一种用于推演系统的语言，你用它创建的大多数图表其实你都应该把它们扔掉，因为它的价值在于作为“脚手架”，帮助你理解和推演你所做的决策。
[原文] [Grady]: uh you know of my definition of architecture all architecture is design but not all design is architecture architecture represents the set of significant design decisions that shape the form and function of a system where significant is measured by cost of change now I struggled with that definition for a long time because there were many other definitions the software engineering institute in particular had one and they were dealing with you know structure and shape and things like that you know I love those guys dearly but that's not what my experience was that architecture was all about decisions and the shape you ended up with was the consequence of those decisions and so I wanted to focus upon the process of that decision making
[译文] [Grady]: 呃，你知道我对架构的定义：所有的架构都是设计，但并非所有的设计都是架构。架构代表了塑造系统形式与功能的一系列重大设计决策，其中“重大”是由变更成本来衡量的。我在这个定义上纠结了很久，因为还有很多其他的定义。特别是软件工程研究所（SEI）有一个定义，他们关注的是结构、形状之类的东西。你知道，我很爱那些家伙，但这与我的经验不符。我的经验是，架构完全是关于决策的，而你最终得到的形状只是这些决策的后果。所以我想要聚焦于决策制定的过程。
[原文] [Grady]: the same thing therefore is true with regards to patterns all complex systems are full of patterns there's I'm deeply influenced by the work from uh the Santa Fe Institute on complexity and uh there's just a whole body of literature about the structure of complex systems i had the wonderful opportunity of of writing a paper with George Ellis george is the gentleman the astronomer who worked with uh with uh Stephen Hawking and they wrote the classic book uh the large scale structure of the universe something like that
[译文] [Grady]: 因此，对于模式（Patterns）而言也是同样的道理。所有复杂的系统都充满了模式。我深受圣塔菲研究所（Santa Fe Institute）关于复杂性研究的影响，那里有关于复杂系统结构的大量文献。我有幸曾与乔治·埃利斯（George Ellis）共同撰写过一篇论文。乔治是一位绅士，也是一位天文学家，曾与斯蒂芬·霍金（Stephen Hawking）共事，他们写过那本经典的书《时空的大尺度结构》（The Large Scale Structure of Space-Time）之类的。
[原文] [Grady]: and so George Ellis got wind of my stuff and talk about an amazing connection because here I'm a computing guy and here's an astronomer kind of guy and he he he noticed that I was interested in patterns and he he also has this view that if you look at uh astrophysics uh in much of science we grow that field based upon our discovery of patterns but those patterns should never be dogmatic because the universe nature is is subtle and wonderful and obscure that there are many patterns but there are not sharp edges to those patterns and I think that lesson can be applied and should be applied to the gang of four they are the smells we know we can look at a system we see these patterns but to be dogmatic about them leads us to systems that are far too brittle
[译文] [Grady]: 乔治·埃利斯听说了我的东西——说起这个惊人的联系，我是个搞计算的人，而他是个搞天文学的人——但他注意到我对模式感兴趣。他也持有这样一种观点：如果你观察天体物理学，或者科学的大部分领域，我们是通过发现模式来推动领域发展的。但这些模式绝不应该是教条的，因为宇宙和自然是微妙、奇妙且晦涩的。存在许多模式，但这些模式并没有那种非黑即白的“锋利边缘”。我认为这个教训可以也应该应用于“四人帮”的设计模式。它们是一种“味道”，我们观察一个系统时能看到这些模式，但如果对它们过于教条，就会导致我们的系统变得过于脆弱（brittle）。
📝 本节摘要：
针对主持人关于如何获得架构设计的“好品味”（Good Taste）的提问，Grady 提出了两个核心要素：一是跨领域的经验，避免陷入单一的技术栈（如仅局限于 AWS Web 系统），建议从非计算机领域（如伦敦地铁）或不同类型的系统（如 AlphaFold）中汲取灵感；二是理解“元模式”（Meta-patterns），即关注分离关注点、抽象层次和简洁性。为了加速这一学习过程，Grady 强烈建议阅读经典代码，例如比尔·阿特金森（Bill Atkinson）编写的 MacPaint 源码（被高德纳誉为最美的软件），以及通过研究 Photoshop 架构随外部约束变化的演进史来理解“代码即真理”之外的决策背景。
[原文] [Daniel]: okay now if you say if if you are talking about code smells and having a little bit of leeway into the the patterns themselves that leads into into a subject that I was very interested for quite a while which is um for the lack of a better term good taste in how you design your system and I was curious to uh to understand how one can un can get that sort of skill the sort of good taste because a programmer like me when I read a book like this I go like oh this works for here this is a piece of puzzle that works here if the archite if the software that I'm designing fits this sort of shape I can apply this pattern it's like a sticker just use that but as you say There are situ there there many many situations in which you don't need to apply that particular pattern exactly as it is or verbatim or with those exact particular details but then you leave some space off and I'm wondering what uh I I don't even know how to phrase this question but how can somebody learn this sort of skill and the sort of nuance this sort of um granularity to be able to say Oh this pattern works maybe 80% but here's here are the uh rough edges that we can smoothing out in this way how can one learn the skill
[译文] [Daniel]: 好的，既然您谈到了代码异味（Code Smells）以及在使用模式时要留有余地，这引出了一个我感兴趣很久的话题，那就是——姑且称之为——在设计系统时的“好品味”（Good Taste）。我很好奇一个人如何才能获得这种技能，这种“好品味”。因为像我这样的程序员，当我读这类书时，我会觉得：“哦，这在这里适用，这是一块拼图，如果我设计的软件符合这种形状，我就可以应用这个模式。”就像贴纸一样，直接贴上去就好。但正如您所说，在很多很多情况下，你不需要完全照搬那个特定的模式，或者逐字逐句地套用那些细节，你需要留出一些空间。我在想——我甚至不知道该怎么组织这个问题——但一个人如何才能学会这种技能，这种细微的差别，这种能够说出“哦，这个模式大概 80% 适用，但这里有些粗糙的边缘我们可以这样打磨一下”的分寸感？一个人怎么能学会这种技能？
[原文] [Grady]: well that's a great question and I think there are there are two elements to that in my experience the first is having experience across the domains of computing will help you not get in a rut i mean a lot of the books you read about architecture these days they're nice and fun but by and large they they they focus upon a particular architectural style it would be like saying "Oh I want to learn about civil architecture and the only book I give them is about brutalist architecture or or that particular thing." But it works it's okay and there are certain patterns that one can find with uh web centric systems of global elastic scale and these things have been codified in places like Amazon's AWS which is both good and bad because I mean the economics of AWS have sort of forced us into particular patterns that are hard to get out of and when you choose a piece of it it's the roach mo motel strategy in which once you're in it's really hard to get out and so you're stuck with all these other consequences of those kinds of things
[译文] [Grady]: 嗯，这是个很好的问题。根据我的经验，我认为这包含两个要素。第一是拥有跨计算领域的经验，这将帮助你不要陷入一种思维定势（rut）。我是说，如今你读到的很多关于架构的书，它们很好也很有趣，但大体上它们都聚焦于某种特定的架构风格。这就像是说：“哦，我想学习土木建筑。”但我就只给你一本关于野兽派建筑（Brutalist architecture）的书，或者只关于那一种东西的书。但这行得通，也没问题。在“全球弹性规模的以 Web 为中心的系统”中确实可以找到某些模式，这些东西已经被编纂在像亚马逊 AWS 这样的地方了。这既好又坏，因为我是说，AWS 的经济学在某种程度上迫使我们进入特定的模式，很难跳出来。当你选择了其中的一部分，那就是“蟑螂旅馆”（Roach Motel）策略——一旦进去了就真的很难出来。所以你就被困在那些选择带来的所有其他后果里了。
[原文] [Grady]: but if you look outside your domain and say "Hey I'm building this web system wouldn't it be interesting to learn what the architecture of of Alpha uh Fold was or let's look at the architecture of of um uh I don't know the London Underground." M so being exposed to those things I think will help any individual get out of their box and say there are alternative modalities of thinking and people have engineers have solved these problems with this completely different set of problems in which there are many different forces but wow that inspires me to think within this particular case maybe I should try this idea that I learned from over here it's that kind of cross fertilization that I think is is something that u really great engineers have this that's one of the two elements
[译文] [Grady]: 但如果你看看你的领域之外，说：“嘿，我正在构建这个 Web 系统，但去了解一下 AlphaFold 的架构，或者让我们看看，比如伦敦地铁的架构，难道不有趣吗？”所以，接触这些东西我认为会帮助任何人跳出他们的框框，意识到存在其他的思维模式，工程师们已经在完全不同的一组问题和约束力下解决了问题。但这会启发我在这个特定的案例中思考：“也许我应该尝试一下我从那边学到的这个想法。”我认为正是这种“交叉受精”（cross fertilization），是真正伟大的工程师所具备的特质。这是两个要素中的一个。
[原文] [Grady]: the other element is I think a recognition and an understanding of some meta patterns and that is the following set of things uh it is a it is a desire to build systems that have a good separation of concerns uh that you have layers of abstraction and you know all systems it it all of all of software engineering is one of largely of rising levels of abstraction and then lastly simplicity you take those three things plus patterns those are meta lessons that I think guide me whenever I see a situation and say well what would help me optimize those meta things and that helps me decide what patterns might fit into place
[译文] [Grady]: 另一个要素，我认为是对一些“元模式”（meta patterns）的认知和理解，即以下这些东西：构建具有良好“关注点分离”（separation of concerns）的系统的愿望；拥有“抽象层”（layers of abstraction）——你知道所有系统，整个软件工程在很大程度上就是关于不断提升抽象层次的；最后是“简洁性”（simplicity）。如果你掌握了这三件事再加上模式，这些就是我认为的“元课程”，它们指导着我。每当我看到某种情况时，我会问：“什么能帮助我优化这些元要素？”这能帮助我决定哪些模式可能适用。
[原文] [Daniel]: is there any way to accelerate that that skill acquisition and I feel that so uh throughout the conversation that I uh the conversations that I've had so far um and by my own thinking over the last uh year or two while thinking about this particular subject it feels to me that this skill of good taste can only be acquired through lots of exposure lots of experience and time is a very crucial factor so you cannot really short circuit your way into a very nuanced highquality skill of this sort and the because one of my uh main jobs is to uh to produce learning material for the things that I'm particularly good at uh one of my areas of focus is to how to accelerate that particular skill acquisition so my question to you is can this sort of skill be accelerated in any way
[译文] [Daniel]: 有什么方法可以加速这种技能的习得吗？通过目前的对话，以及我过去一两年对这个话题的思考，我感觉这种“好品味”的技能似乎只能通过大量的接触和经验来获得，时间是一个非常关键的因素。你不能真的抄近道去获得这种非常微妙、高质量的技能。因为我的主要工作之一就是为我擅长的领域制作学习材料，所以我关注的一个领域就是如何加速这种特定技能的习得。所以我想问您的是，这种技能可以以某种方式加速获得吗？
[原文] [Grady]: yes I think it can be and let me give you an analogy to become a great writer you do a lot of reading and it helps to read outside of your domain because you will learn rhythms and patterns and textures of color of language that you can then use to influence your particular style and you'll find your own voice along the way and so I would recommend one way to accelerate for an for an individual developer is go read lots of other code and it's available to you i mean you can go find the source code for ob for um uh Mac Paint Apple's Mac Paint it exists you can find it on the computer history museum it's actually a very beautiful piece of software um Donald canuth Don Canuth once remarked it's probably the most beautiful piece of software he has seen so doesn't take long to look at it and find it online
[译文] [Grady]: 是的，我认为是可以的。让我给你打个比方：要成为一名伟大的作家，你会进行大量的阅读，而且阅读你领域之外的作品很有帮助，因为你会学到语言的节奏、模式和色彩纹理，然后你可以用这些来影响你特定的风格，并在沿途找到你自己的声音。所以我建议，对于个人开发者来说，加速成长的一个方法就是去阅读大量其他的代码。这些代码你是可以获取到的。我是说，你可以去找到 MacPaint——苹果的 MacPaint——的源代码。它就在那里，你可以在计算机历史博物馆找到它。它实际上是一段非常优美的软件。唐纳德·克努特（Donald Knuth，高德纳）曾评价说，这可能是他见过的最美的软件。所以花不了多少时间就能在线找到并看看它。
[原文] [Grady]: and I think what you'll learn is that in the presence of constraints that those developers had what is the kind of structure they ended up with and you'll learn two things first is wow it's a really visible structure that it's not difficult to see the essence of the architecture there because it's just in the code itself and the second is it's probably very different than the kinds of things one write the code is the truth but it's not the whole truth and so reading things around that code the comments and such as to how they ended up with that is another great way to uh to to learn
[译文] [Grady]: 我想你会学到的是，面对那些开发者所面临的约束条件，他们最终采用了什么样的结构。你会学到两件事：第一，哇，这是一个非常显而易见的结构，不难看出架构的本质，因为它就在代码本身里；第二，它可能和你写的东西非常不同。“代码是真理”，但它不是全部的真理。所以阅读代码周边的东西，比如注释，以及关于他们是如何最终得出那个结果的资料，是另一种很好的学习方式。
[原文] [Grady]: here's a great example I had an opportunity to work with the u to interview the two architects of Photoshop one of the gentlemen was the one of the original architects the other was brought on later at Adobe they had two of them they split their responsibilities because one was responsible for the Mac codebase the other for the PC PC codebase it was really fun to see the two of them but they walked me through the architectural changes that took place over the years and that is an interesting story into itself because you'll find that external forces required Adobe to change the architecture of Photoshop four five six times but at its essence some of the original abstractions are still there and endure and so you know if you want to learn how to write lots of great code read the code of others go read some of uh some of Lionus's kernel code it's beautiful stuff
[译文] [Grady]: 这里有一个很好的例子。我有机会采访了 Photoshop 的两位架构师。其中一位是最初的架构师之一，另一位是后来加入 Adobe 的。他们有两个人，分担了职责，因为一个负责 Mac 代码库，另一个负责 PC 代码库。看着他们两个非常有趣。但他们带我回顾了多年来发生的架构变化，这本身就是一个有趣的故事。因为你会发现，外部力量迫使 Adobe 改变了 Photoshop 的架构四、五、六次，但在本质上，一些最初的抽象仍然存在并延续着。所以你知道，如果你想学习如何编写大量优秀的代码，去读别人的代码。去读读 Linus（Torvalds）的内核代码，那是很美的东西。
[原文] [Daniel]: what makes beautiful code to you great question beautiful code flows it resonates it tells a story it is clear it is simple it is not obscure there's such pressure I think among many individuals especially those using co-pilot like tools to write the tightest smallest code they might write but the problem is it's going to be impossible for somebody to follow on and understand it and more likely people will throw it away because it is so so difficult to understand
[译文] [Daniel]: 对您来说，什么构成了优美的代码？
[译文] [Grady]: 好问题。优美的代码是流动的（flows），它能产生共鸣，它在讲述一个故事。它是清晰的，它是简洁的，它不晦涩。我认为许多人，特别是那些使用 Copilot 类工具的人，面临着一种压力，要写出最紧凑、最小的代码。但问题是，这样别人就不可能跟进并理解它，更有可能的是，人们会把它扔掉，因为它太难理解了。
📝 本节摘要：
针对主持人关于 Scala、函数式编程及“形式化方法”（Formal Methods）的提问，Grady Booch 给出了极其务实的回答。他指出，形式化方法仅适用于极少数（不到 1%）“人命关天”（如医疗设备、航空控制）的系统，因为其成本极高。他强调，即使是完美的数学证明也无法解决现实世界的物理干扰（如宇宙射线翻转比特位）。虽然他支持强类型系统（如 Ada 和 Scala）来辅助架构设计，但他认为这只能解决一部分问题。他还以微软利用形式化方法验证 Windows 设备驱动程序为例，说明了其在特定封闭场景下的有效性。
[原文] [Daniel]: yep uh I wanted to ask about formal methods because you me you mentioned that where do formal methods matter and because there is a discussion in the programming languages space I'm a particular fan of uh Scola and functional programming and Scola has particularly powerful type system where a lot of things are checked by the compiler and you want to make sure that things are correct before they run many people in the in in this community believe that it's such a that's an indictment on software engineering that we are still at the stage where the only way to learn whether a program is correct is to run it uh so and they very much advocate for u uh formal methods and formal verification and stuff where are these important and are there situations in which formal methods are really not
[译文] [Daniel]: 是的。我想问问关于形式化方法（Formal Methods）的问题，因为您提到了它。形式化方法在哪些地方重要？因为在编程语言领域有一个讨论——我是 Scala 和函数式编程的特别粉丝，Scala 有一个特别强大的类型系统，很多东西都是由编译器检查的，你想在运行之前确保东西是正确的。这个社区里的许多人认为，如果我们仍然处于“只有运行程序才能知道它是否正确”的阶段，那是对软件工程的一种控诉。所以他们非常提倡形式化方法和形式化验证（Formal Verification）之类的东西。这些在哪些地方是重要的？是否存在形式化方法真的不适用的情况？
[原文] [Grady]: another set of great questions and you introduce the issues of typing as well in my experience over the decades uh and I look at the the number of systems which I've been engaged with formal methods you know now pop up in a fraction of a percent of the systems that I encounter and in most cases it appears in those places where human lives are at stake where a failure means people die and it's in those circumstances where you can't fail that formal methods are so important
[译文] [Grady]: 这又是一组好问题，而且你还引入了类型（typing）的问题。根据我几十年的经验，看看我接触过的系统数量，形式化方法现在只出现在我遇到的系统中极小的一部分（不到百分之一）。在大多数情况下，它出现在那些人命关天的地方，在那里的失败意味着由于某种原因人会死亡。正是在这种不能失败的情况下，形式化方法才如此重要。
[原文] [Grady]: so it's it goes back to the economics and the moral issues we want to apply these kinds of techniques in those cases where the cost of failure are so high because remember that formal methods work really well for certain things but they don't deal with the problems of the reality of the external world so I might build a formal system that tells me or I might use a formal system that says hey I've got this thing that you know helps me drive this car and functionally it may be wonderful and perfect but the problem is not so much you know and I feel confident that it that the formal methods have helped me there but the problem is I can't account for every edge case
[译文] [Grady]: 所以这要追溯到经济学和道德问题。我们希望在失败成本极高的情况下应用这类技术。因为请记住，形式化方法在某些事情上效果很好，但它们无法处理外部世界现实的问题。所以我可能会构建一个形式化系统，或者使用一个形式化系统，它告诉我：“嘿，我有这个东西可以帮我驾驶这辆车。”在功能上它可能是美妙且完美的，我对形式化方法在那里的帮助感到自信。但问题在于，我无法顾及到每一个边缘情况（edge case）。
[原文] [Grady]: and so systems that are are are functionally proven you have to recognize their limits upon what they can do no functional system that I have ever encountered when no formal system of of reasoning or model or model checking system deals with things like oh my gosh a cosmic ray has just flipped the bit on this particular ROM this is outside the scope of most systems of formal methods most formal methods don't deal with time and space kinds of things of the real world so from the very beginnings we know our mathematical abilities are limited we could do it theoretically but the systems are so complex there aren't enough you know seconds in the life of the universe to do that
[译文] [Grady]: 所以对于那些在功能上被证明过的系统，你必须认识到它们能做的事情是有限的。我从未遇到过任何功能系统，没有任何形式化的推理系统、模型或模型检查系统能够处理诸如“天哪，一道宇宙射线刚刚翻转了这个特定 ROM 上的一个比特位”这样的事情。这超出了大多数形式化方法系统的范围。大多数形式化方法不处理现实世界中的时间和空间这类事情。所以从一开始我们就知道，我们的数学能力是有限的。理论上我们可以做到，但系统太复杂了，宇宙生命中的秒数都不够用来做这些计算。
[原文] [Grady]: now typing is a different issue because strong typing and I'm a great proponent of it having you know developed having been involved with the creation of of ADA back in the day which was a strongly typed language strong typing helps you in gluing pieces together but it doesn't necessarily help you with the functional aspects the operational aspects of a system there's a whole different set of mathematics you need to do there so Scala is great typing is great i think it's wonderful but even so that only addresses a piece of the hard parts of the system
[译文] [Grady]: 现在的类型（Typing）是一个不同的问题。因为强类型——我是它的坚定支持者，你知道我当年参与了 Ada 语言的开发，那是一种强类型语言——强类型有助于你将各个部分粘合在一起，但它不一定能在系统的功能方面、操作方面帮助你。那里你需要做一套完全不同的数学运算。所以 Scala 很棒，类型系统很棒，我认为这很精彩，但即便如此，那也只解决了系统困难部分的一小块。
[原文] [Grady]: so short answer is to your question yeah I if if the system if the cost of failure is high then there probably is a role for formal methods but even so we know that formal methods have a limitation to them and one of the great uses I've seen for formal methods is Microsoft used it uh to validate device drivers for various forms of Windows great stuff because you have a complete functional specification you knew you could you knew when things were right or not and so functional methods or formal methods were absolutely perfect for that but there are very few cases where that is true
[译文] [Grady]: 所以简单回答你的问题：是的，如果系统的失败成本很高，那么形式化方法可能有一席之地，但即便如此，我们知道形式化方法也有其局限性。我见过的形式化方法的一个伟大应用是微软用它来验证各种版本 Windows 的设备驱动程序。那是很棒的东西，因为你有一个完整的功能规范，你知道什么时候是对的什么时候是错的，所以形式化方法对于那种情况绝对是完美的。但在很少有其他情况下这也是成立的。
📝 本节摘要：
在本章中，Grady Booch 首先对 Python 和 JavaScript 等动态语言在关键系统中的滥用表示担忧，将其类比为当年的 Visual Basic——易于上手但极易产生脆弱的“遗留代码”。他特别警告说，用 Python 控制核武器将是“极度糟糕的主意”。随后，话题转向 LLM 生成的代码。Grady 指出，研究表明 LLM 生成的代码充满了安全漏洞和脆弱性。虽然专家（如 Andrej Karpathy）能辨别好坏，但对于缺乏经验的开发者，这只会导致系统中堆积大量的“垃圾代码（Bullshit）”。更深层的问题在于，许多开发者缺乏“切身利益”（Skin in the Game），只想着在股票归属（Vest）后离职，因此没有动力去打磨代码质量。最后，他批评了微软、谷歌和 OpenAI 的过度营销，认为目前企业因 FOMO（错失恐惧症）而盲目撒下的“AI 精灵粉尘”，最终大多会经历市场的残酷洗牌，就像早期笔记本电脑形态的演变一样，最终回归理性。
[原文] [Daniel]: do you see um the JavaScripts and the Pythons warranted to the degree that we see them uh currently so there there is a a lot of complaint about the quality of the code being pushed to production and a lot of that criticism is directed at the tools and the languages that are used to push that because Python and JavaScript in particular are relatively easy to pick up and so it's very easy to push faulty code that hasn't been checked and so on and so forth dynamic typing also plays a role and so a lot of criticism is directed at those languages and those tools and I'm wondering to what degree do you think those tools and th that that style of writing and pushing code is legitimate
[译文] [Daniel]: 您认为 JavaScript 和 Python 目前的受捧程度是合理的吗？现在有很多关于推送到生产环境的代码质量的抱怨，很多批评都指向了用来推送代码的工具和语言。因为特别是 Python 和 JavaScript 相对容易上手，所以很容易推送那些未经检查的有缺陷的代码等等。动态类型也起了一定作用。所以很多批评都指向了这些语言和工具。我很好奇，您在多大程度上认为这些工具以及那种编写和推送代码的风格是合理的？
[原文] [Grady]: visual Basic had the same problem in that Visual Basic back in the day was a language was transformative it meant that people who were not just full-time programmers could all of a sudden build full-fledged applications under Windows and you would see entire businesses being built around it problem is that it was very fragile and so the code ended up being very disposable
[译文] [Grady]: Visual Basic 曾经也有同样的问题。当年的 Visual Basic 是一门变革性的语言，它意味着那些不只是全职程序员的人突然可以在 Windows 下构建功能齐全的应用程序，你会看到整个业务都建立在它之上。问题在于它非常脆弱，所以那些代码最终变得非常“一次性”（disposable）。
[原文] [Grady]: i think the same thing is true with JavaScript and JavaScript and Python that it's easy to build simple things but the problem is once you've written a line of code all of a sudden becomes legacy and you're building up this weight of code that's difficult to change over time so you know for lightweight things I I Python's my primary language right now it's great because I throw away most of the stuff I do it's not like I have a million people uh using what I'm creating here and that's fine so it's fine for me because it's very malleable but as I start building systems again like air traffic control systems or pacemakers or things like that then you know you want to apply it a little bit more discipline and maybe those aren't the right choice of languages for it i would not recommend using Python for nuclear weapon for example that would be a a really profoundly bad idea
[译文] [Grady]: 我认为 JavaScript 和 Python 也是如此。构建简单的东西很容易，但问题是，一旦你写下一行代码，它突然就变成了“遗留代码”（legacy），而且你正在积累这种难以随时间修改的代码包袱。所以你知道，对于轻量级的东西——Python 是我现在的主要语言，它很棒，因为我做的大部分东西我都扔掉了，又不是有一百万人于使用我在这里创造的东西，这没问题。所以对我来说这很好，因为它可塑性很强。但如果我重新开始构建系统，比如空中交通管制系统或心脏起搏器之类的，那么你知道，你会想要应用更多的纪律，也许那些语言并不是正确的选择。例如，我不会推荐用 Python 来控制核武器，那将是一个极度、极度糟糕的主意。
[原文] [Daniel]: um and that ties into the AI bit that uh LLMs are now generating a lot of code and most of that code is generated in the most popular languages and the most popular languages being JavaScript and Python the LLMs are going to generate even more of that and uh I wanted to ask if you could speak to that is there a danger not necessarily from the point of view of because we've talked about this uh a little bit and uh we might expand further on the topic of uh replacement of software engineers and whatever but I wanted to ask you about the quality of the code pushed by LLMs in these languages that could become dangerous if applied to the wrong systems could you speak to the quality of code being pushed by LLMs
[译文] [Daniel]: 嗯，这这就联系到了 AI 的部分。LLM（大语言模型）现在正在生成大量代码，而大部分代码都是用最流行的语言生成的，也就是 JavaScript 和 Python。LLM 将会生成更多这样的代码。我想问问您能否谈谈这一点。是否存在危险？我指的不仅仅是取代软件工程师——这个我们已经聊过一点，也许以后会深入探讨——我想问的是 LLM 用这些语言推送的代码质量。如果应用到错误的系统中，这可能会变得危险。您能谈谈 LLM 推送的代码质量吗？
[原文] [Grady]: so there have been a number of really formal studies that suggest that LLMs produce code that has far more security holes than ones developed by humans that they produce code that is uh is very fragile that breaks in strange ways so it's it's hard to use and so there's this seductive problem in that if I'm a developer that's never you know used this particular API or something then my gosh these these tools cursor co-pilot or co-pilot or whatever any number from which you could choose they accelerate that and so for a novice who knows nothing about these things it's a great way to build prototypes very very quickly
[译文] [Grady]: 已经有许多非常正式的研究表明，LLM 生成的代码比人类开发的代码包含更多的安全漏洞，它们生成的代码非常脆弱，会以奇怪的方式崩溃，所以很难使用。这就存在一个诱惑性的问题：如果我是一个从未用过某个特定 API 的开发者，天哪，这些工具——Cursor、Copilot 或者你可以选择的任何其他工具——它们加速了这个过程。所以对于一个对这些一无所知的新手来说，这是快速构建原型的绝佳方式。
[原文] [Grady]: and in the hands of somebody like Andruff who of course developed the idea of VI V by programming in the in the hands of an expert he's someone who knows when the code is is utterly wrong so he's in a very good place where he can he can discriminate the problem is this group in the middle the people who are trying to build systems that endure and using code along the way uh I fear for that that I think most of these kinds of systems uh I worry about the degree of uh bullshit that they produce in our code and and uh it's it's a clear and present problem and I know how we I don't know we changed that to be very honest
[译文] [Grady]: 而在像 Andrej（注：指 Andrej Karpathy）这样的人手中——当然是他提出了“软件 2.0”或通过编程来编程的想法——在专家手中，他知道代码什么时候是完全错误的，所以他处于一个很好的位置，他可以进行辨别。问题在于中间的这群人，那些试图构建能够持久运行的系统并在此过程中使用代码的人。我为此感到担忧。我认为大多数这类系统……我担心它们在我们的代码中制造出的“废话”（bullshit）程度。这是一个显而易见且迫在眉睫的问题。坦率地说，我不知道我们要如何改变这种状况。
[原文] [Daniel]: is there any way uh do you have any recommendations for okay this this the problem of in having people intentionally add discipline to the code that they produce be it with copilot or their own code is enough of a hard topic but at the individual developer do you have any tips for how to add discipline and how to add guard rails and um maybe um systems not necessarily techn technological systems but mental systems uh for them to avoid producing stupid code with real consequences
[译文] [Daniel]: 有没有什么办法……您有没有什么建议？让人们有意地为他们生成的代码增加纪律——无论是用 Copilot 还是自己写的——这已经是一个足够难的话题了。但对于个人开发者，您有什么建议来增加纪律和护栏吗？也许不是技术系统，而是心理系统，以避免产生具有真实后果的愚蠢代码？
[原文] [Grady]: i think some developers will never have the desire to do that because they're so happily turnurning out code for which they have no responsibility once it's out there and so there is no incentive either economic or scientific or moral for that matter that would incentivize them to provide that discipline and I think LLMs in particular have enabled the growth of that particular segment of the industry so I worry about that
[译文] [Grady]: 我认为有些开发者永远不会有这种愿望，因为他们正快乐地制造出那些一旦发布就不再需要负责的代码。因此，无论是在经济上、科学上，还是在道德上，都没有激励机制去促使他们提供这种纪律。我认为特别是 LLM 助长了行业中这一特定群体的增长。所以我对此感到担忧。
[原文] [Daniel]: yeah so it seems to be a skin in the game problem yeah you're right how much skin in the game do they have
[译文] [Daniel]: 是的，所以这似乎是一个“切身利益”（skin in the game）的问题。是的，你说得对，他们有多少切身利益呢？
[原文] [Grady]: yes and especially with the job hopping we see in so many places in the valley it's like I'm going to write this code and you know I'm not going to be working with this company after I vest so why should I care that's a real problem in the end you know capitalism wins out that these companies is they're going to you know they eventually will die truth truth and reality are are harsh and they'll eventually come to grips with that kind of reality but right now we're in a funny place I think in the industry where it's so easy to produce things that of that are of such poor quality because there's so much money chasing these companies right now
[译文] [Grady]: 是的，特别是在硅谷很多地方我们看到的跳槽现象。就像是：“我要写这段代码，反正我在股票归属（vest）之后就不会在这家公司工作了，所以我为什么要在这个？”这是一个真正的问题。最终，你知道，资本主义会胜出，这些公司……它们最终会死掉。真理和现实是残酷的，它们最终不得不面对那种现实。但现在我们正处于行业的一个滑稽阶段，制造如此低质量的东西太容易了，因为现在有太多的钱在追逐这些公司。
[原文] [Grady]: i'm I'm thinking right now of a a particular project i can't mention the name of the company let's just say that there's they're trying to apply LLMs to a particular customer support problem and to begin with they're already starting off wrong because there's great FOMO in this company oh my gosh our competitors are using it let's sprinkle some AI pixie dust into what we're doing and they failed to step back and say what are we trying to do for our end users and customers but instead throwing the AI into it to see what happens and see what sticks
[译文] [Grady]: 我现在正如想到一个特定的项目——我不能提那家公司的名字——就说他们试图将 LLM 应用于特定的客户支持问题吧。从一开始他们就错了，因为这家公司存在巨大的 FOMO（错失恐惧症）：“天哪，我们的竞争对手在用它，让我们往我们做的事情里撒点‘AI 精灵粉尘’（AI pixie dust）吧。”他们没有退后一步问：“我们要为最终用户和客户做什么？”反而是把 AI 扔进去，看看会发生什么，看看能保留住什么。
[原文] [Daniel]: do you think there's an overhype with uh the current AI wave
[译文] [Daniel]: 您认为当前的 AI 浪潮存在过度炒作吗？
[原文] [Grady]: absolutely there's tremendous overhype at the moment and not just overhype but just too much money chasing it and you know Microsoft and Google and Open AI problem is they're cramming it down our throats it's impossible to even do the simplest kind of things without saying "By the way here's this new AI feature you can have with it." Well I don't want that feature but you're forcing me to use it and that's annoying
[译文] [Grady]: 绝对的，目前存在巨大的过度炒作。不仅仅是炒作，而是有太多的钱在追逐它。你知道，Microsoft、Google 和 OpenAI 的问题在于，他们正把它强塞进我们的喉咙。甚至做最简单的事情时，也不得不听到：“顺便说一句，这是你可以使用的新 AI 功能。”好吧，我不想要那个功能，但你在强迫我使用它，这很烦人。
[原文] [Grady]: the great thing about science and the thing about computing is it over time truth wins out and it becomes a very you know homoasius kicks in and we'll eventually find the right right standard right now the weather veins flapping all over the place but we'll get there it's probably a uh a bit of search happening until we get to that that Absolutely yeah i mean a good analogy is think back to the very early days of of laptops when they were first created i mean the idea of a laptop came from um from Alan K and his idea of the dynook and when the form factor of the laptop came out you saw all sorts of experimentation with different ports and different ways of you know using cursors and the like and little displays and the like but ultimately we settled down to a pretty standard form... the same thing is going to happen with AI i believe
[译文] [Grady]: 科学和计算的伟大之处在于，随着时间的推移，真理会胜出。就像体内平衡（homeostasis，此处口误为homoasius）会起作用，我们最终会找到正确的标准。现在风向标到处乱转，但我们会到达那里的。这可能是一个搜索过程，直到我们到达那里。绝对是。我是说，一个很好的类比是回想笔记本电脑的早期时代。笔记本电脑的想法来自艾伦·凯（Alan Kay）和他的 Dynabook 概念。当笔记本电脑的形态出现时，你看到了各种各样的实验：不同的接口、不同的使用光标的方式、小显示屏等等。但最终我们稳定在了一个相当标准的形态……我相信 AI 也会发生同样的事情。
📝 本节摘要：
在本章中，Grady Booch 否认了大语言模型（LLM）仅仅是软件抽象层级提升的观点，将其定义为一种根本不同的“统计学怪兽”。他以 AlphaGo 为例，说明 AI 确实能通过探索人类因传统而忽视的“山谷”来创造新策略，但他强调 LLM 在架构上根本无法构建“真理系统”。Grady 指出，真理源于具身性（Embodiment）——即人类拥有的欲望、情感以及与物理世界的互动，而 LLM 仅拥有基于文本的统计概率。他引用了莉莉·汤普林（Lily Tomlin）关于“现实只是共同的幻觉”的戏剧台词，以及马丁·布伯（Martin Buber）的“我与你”（I-Thou）概念，论证了人类感知的独特性，并断言尽管心智可能在理论上是可计算的，但目前的 LLM 架构绝不是通向那里的路径。
[原文] [Daniel]: are LLMs from uh the point of view of a a programmer so uh you've said this not only in this conversation but uh uh in other conversations as well that software has generally moved towards higher levels of abstraction um are and I have an opinion on this but I'm not going to uh I'm not I'm going to ask you first uh are LLMs another layer of abstraction or are they something else
[译文] [Daniel]: 从程序员的角度来看，LLM（大语言模型）……您不仅在这次对话中，在其他对话中也说过，软件通常朝着更高层次的抽象发展。我也对此有看法，但我不想……我还是先问您：LLM 是另一层抽象吗？还是某种别的东西？
[原文] [Grady]: i would say they're something else because they represent the pendulum swinging between uh discrete deterministic symbolic systems to systems that are more uh statistical in nature i mean that's what LLMs are they're a they're a statistical beast they allow us to build these multi-dimensional networks and give us ways to search through them that's you know from an architectural perspective that's what they are
[译文] [Grady]: 我会说它们是某种别的东西，因为它们代表了钟摆在“离散的、确定性的符号系统”与“本质上更具统计性的系统”之间的摆动。我是说，这就是 LLM 的本质，它们是一只统计学怪兽。它们允许我们构建这些多维网络，并给了我们搜索它们的方法。你知道，从架构角度来看，这就是它们。
[原文] [Grady]: um hang on I'm going to bookmark this notion because I'm going to go down a rabbit hole and this goes back to Alph Go which is another great example of that not an LLM but a neural system uh when Alph Go beat the human players uh one of the gentlemen I think from Korea he was you know he was visibly shaken by this and he said oh my gosh like I I've seen the hand of God in this one
[译文] [Grady]: 嗯，等一下，我要把这个概念先标记一下，因为我要钻进一个“兔子洞”了。这要回溯到 AlphaGo，那是另一个很好的例子——它不是 LLM，而是一个神经系统。当 AlphaGo 击败人类棋手时，其中一位先生——我想是来自韩国的（注：指李世石）——你知道，他显然对此感到震惊，他说：“天哪，我仿佛在其中看到了上帝之手。”
[原文] [Grady]: I like to think of it this way that LLMs in that particular architecture have allow us to create a multi-dimensional space and in the case of you know things like protein fold building or or or chess games and the like or just producing coherent conversation they're all related in the sense that tradition and experience and and human factors have led us down particular paths within this multi-dimensional landscape
[译文] [Grady]: 我喜欢这样思考：LLM 在那种特定架构下，允许我们创建一个多维空间。无论是在蛋白质折叠构建、国际象棋游戏之类的案例中，还是仅仅产生连贯的对话，它们都是相关的。因为传统、经验和人为因素引导我们在这一多维景观中走上了特定的路径。
[原文] [Grady]: the thing that LLMs do statistically is they are not burdened by the history and and such that have led us to these paths so when Alph Go made these remarkable Mu uh moves it's because they explored hills and valleys that we had not seen as humans because our past had led us away from them but statistically they were there we simply had not shined a light to them and that's a lot what that's where I think a lot of the interest comes from LLMs because they produce astonishingly coherent output
[译文] [Grady]: LLM 在统计上所做的是，它们没有背负那些引导我们走上特定路径的历史包袱。所以当 AlphaGo 做出那些非凡的举动时，是因为它们探索了我们作为人类未曾看到的“山丘和低谷”，因为我们的过去引导我们要远离它们。但从统计学上讲，它们就在那里，只是我们没有把光照向它们。我认为这就是 LLM 引发大量兴趣的原因，因为它们能产生惊人连贯的输出。
[原文] [Grady]: the problem of course is they're not grounded in reality that if you've often heard me say um our large language models transformer-based architectures are architecturally incapable of building systems of truth because they have no basis of truth within them and any truth you might see for them is accidental and only statistically interesting yet they are truthy they're truthy enough that they fool a lot of people to the point where well we've already seen it diagnosably people are going down the path of using LLMs and and replacing their loved ones with it they're finding God in the process this is Eliza on steroids
[译文] [Grady]: 当然，问题在于它们没有扎根于现实。如果你经常听我说，你会知道我认为我们的大语言模型、基于 Transformer 的架构，在架构上是无法构建“真理系统”的，因为它们内部没有真理的基础。你在它们身上看到的任何真理都是偶然的，仅仅在统计学上是有趣的。然而它们很“似是而非”（truthy），它们太像真的了，以至于欺骗了很多人，甚至到了这种地步——正如我们已经看到的——人们开始用 LLM 替代他们的亲人，他们在过程中寻找上帝。这简直是类固醇加强版的 Eliza（注：早期的心理治疗聊天机器人）。
[原文] [Daniel]: this is a a very interesting take because uh LM's being so being uh statistic they have no notion of truth and I was curious to uh gain your perspective on how humans have at least the perception that we can discover the truth... so I was I was wondering about your perspective on this ability to discover truth how is a human fundamentally different in this search
[译文] [Daniel]: 这是一个非常有趣的观点。因为 LLM 是统计性的，它们没有真理的概念。我很好奇您对人类如何拥有——至少是感知——发现真理能力的看法……我想知道您对这种发现真理能力的看法，人类在这个探索过程中有何根本不同？
[原文] [Grady]: a great question humans are also impacted by a context that LLMs do not have we have desires we have subjective experiences we have goals we have sense of love and beauty and all these emotions that an LLM does not have mhm and I believe they shape us in ways individually and culturally because of what society has shaped around it in a particular place that provide a context that an LLM is simply not able to encompass
[译文] [Grady]: 问得好。人类受到一种 LLM 所不具备的语境（context）的影响。我们有欲望，我们有主观体验，我们有目标，我们有对爱和美的感知，以及所有这些 LLM 没有的情感。嗯，我相信这些因素在个人和文化层面上塑造了我们，因为社会在特定地点围绕这些因素塑造了我们，这提供了一种 LLM 根本无法包含的语境。
[原文] [Grady]: remember also that we humans are embodied and so we live in that complex world of noise and gaps and these kinds of things whereas if I'm an LLM the only world I've seen is what's been presented to me largely through text and through pictures which are very self- selected uh mostly coming from the internet and the like so they see a fraction of the observable space of what it means to be human so I think that's that's one thing that distinguishes us
[译文] [Grady]: 还要记住，我们人类是“具身的”（embodied），所以我们生活在那个充满噪音、缝隙和这类事物的复杂世界里。而如果我是个 LLM，我看到的世界仅仅是呈现给我的东西，主要是通过文本和图片，而这些都是经过高度筛选的，大多来自互联网之类的。所以它们只看到了“生而为人”意味着什么这一可观察空间的一小部分。所以我认为这是区分我们的一点。
[原文] [Daniel]: ... I tend to agree with you uh with respect to the criticism of the hubris and the um uh hypocrisy of creating things that we don't really understand... to what degree could an AI given your experience be presented with an objective function that is measurable in similar ways to ourselves and call it a truth-seeking AI
[译文] [Daniel]: ……我倾向于同意您对那种傲慢和伪善的批评，即创造我们并不真正理解的东西……根据您的经验，AI 在多大程度上可以被赋予一个像我们一样可测量的目标函数，并称之为“寻求真理的 AI”？
[原文] [Grady]: a meta meta con comment I should say and the fact that we're going down this path is what just delights me about the whole topic of computing because notice how we are eventually led to these deep philosophical questions of what it means to be human... you may be surprised to know that my studies and for the last six or so years I've been studying with a set of neuroscientists to understand the architecture of the brain i've come to the conclusion that the mind is computable now Roger Penrose would disagree with me but that's a topic for a completely different podcast i think Roger is wrong in that regard that's okay but it then says well gee maybe the things that we do are human are indeed computable but uh we're probably generations away from that because LLMs are certainly not the right architecture to do that
[译文] [Grady]: 我要说一个“元元评论”（meta meta comment）。事实上，我们正沿着这条路走下去，这正是整个计算话题让我感到高兴的地方，因为请注意，我们最终是如何被引向这些关于“生而为人意味着什么”的深刻哲学问题的……你可能会惊讶地发现，在过去六年中，我一直在与一群神经科学家一起研究，试图理解大脑的架构。我得出的结论是：心智是可计算的。罗杰·彭罗斯（Roger Penrose）会不同意我的观点，那是完全另一个播客的话题了，我认为罗杰在这方面是错的，那没关系。但这进而说明，好吧，也许我们要做的那些属于人类的事情确实是可计算的。但呃，我们可能离那个目标还有几代人的距离，因为 LLM 肯定不是实现那一目标的正确架构。
[原文] [Grady]: there's this great play by Lily Tomlin and her wife um I think it's titled uh the search for signs of intelligent life in the universe and in it uh Lily plays the the character in which he says uh reality is just a shared hallucination and I think there's great truth in that that what we see as humans I mean we're we have as in humanity we have a particular view upon what the cosmos is and what it is not as Sean has said we are the way that the universe can know itself but even so we are looking at it through the lens of our own humanity
[译文] [Grady]: 莉莉·汤普林（Lily Tomlin）和她的伴侣有一部很棒的戏剧，我想题目是《寻找宇宙中智慧生命的迹象》（The Search for Signs of Intelligent Life in the Universe）。在剧中，莉莉扮演的角色说：“现实只是一个共同的幻觉。”（Reality is just a shared hallucination）。我认为这其中有很大的真理。作为人类，我们看到的……我是说，作为人类，我们对宇宙是什么、不是什么有着特定的看法。正如萨根所说，我们是宇宙认识自身的方式。但即便如此，我们仍是透过我们自身人性的透镜在观察它。
[原文] [Grady]: and this is not to say that there aren't truths outside of our species to go down another rabbit hole uh I do a lot of work with with a number of citation researchers researchers and humpback whales and I swear if you ever get to a blue whale largest creature mammal ever to walk the earth well swim the earth and look them in the eye there's this there's this concept of the eye thou which Martin Buber speaks of in which you look into the eye of another and you see the presence of another soul the idea of having a theory of mind of others we know that something else is going on there so there's probably there probably is some you know deep shared truth among anything that's sentient
[译文] [Grady]: 这并不是说在我们物种之外没有真理。再钻个兔子洞，我与许多鲸类研究人员以及座头鲸做了很多工作。我发誓，如果你曾经接近过蓝鲸——地球上行走过，哦不，游过的最大的哺乳动物——并看着它们的眼睛。这就有一种马丁·布伯（Martin Buber）所说的“我与你”（I-Thou）的概念，当你看着另一个生命的眼睛时，你看到了另一个灵魂的存在。这种拥有对他人的“心智理论”（theory of mind）的想法，让我们知道那里发生了别的事情。所以在任何有感知的事物之间，可能都存在着某种深刻的共同真理。
[原文] [Grady]: but then if we go back to my notion that the mind is indeed computable can LLMs get us there well the problem with LLMs is because they are purely statistical consider what they're trained on they're trained upon the corpus of things we have fed them that are only a fraction of humanity... but what is truth well gosh now we're getting into some deeply philosophical things i think the answer there is uh a truth is something that whose whose reality can be tested in its falseness and LLMs don't of course this is the nature of science that we have hypotheses and we can come up with ways that test their val validity or not and that's how science progresses
[译文] [Grady]: 但如果回到我的观点，即心智确实是可计算的，LLM 能带我们到达那里吗？LLM 的问题在于它们纯粹是统计性的。想想它们是在什么上训练的，它们是在我们喂给它们的语料库上训练的，那只是人类的一小部分……但什么是真理？天哪，现在我们进入了深刻的哲学领域。我认为答案是：真理是某种其现实性可以在其证伪性中被检验的东西。当然，LLM 做不到这一点。这是科学的本质，我们有假设，我们可以想出办法来检验它们的有效性与否，这就是科学进步的方式。
[原文] [Grady]: llms don't do that because they're just simply saying statistically I've got this multi-dimensional space you've given me this huge context which is basically a vector into it and u hey you know it's close but that's not necessarily the only truth it just so happens statistically what happens if I point in there but not objectively because it's not measured to the outside world and that's where the embodiment thing comes into place our truth comes from our embodiment as conscious beings in the world llms don't have that so we're in many ways we as humans are speaking very different languages in the LLMs their truth is not the same truth as what we as humans have
[译文] [Grady]: LLM 不做这个，因为它们只是简单地说：“从统计上讲，我有这个多维空间，你给了我这个巨大的上下文（基本上是一个指向其中的向量），嘿，你知道它很接近。”但这不一定是唯一的真理，它只是碰巧统计上指向了那里，而不是客观的，因为它没有在这个外部世界中被测量。这就是“具身性”（embodiment）发挥作用的地方。我们的真理来自于我们作为世界上有意识的存在的具身性。LLM 没有这个。所以在很多方面，我们作为人类说着与 LLM 非常不同的语言，它们的真理与我们人类拥有的真理是不一样的。
📝 本节摘要：
在访谈的最后，针对程序员对 AI 取代工作的普遍焦虑，Grady Booch 进行了有力的回应。他将当前的变革类比为当年从汇编语言到高级语言的过渡——虽然消灭了“打孔员”等旧工种，但通过提升抽象层次，让更多人掌握了计算能力。他举例说明一位夏威夷教师如何利用 LLM 为不同年龄段的学生重写故事，强调了“人”在主观判断中的核心地位。Grady 驳斥了 AI 毁灭人类的末世论（P(doom) 趋近于零），但他严肃警告了自动化偏见（如种族歧视、厌女症）和推卸责任的现实风险。最终，他引用人类历史上多次克服危机的韧性，以及国际象棋特级大师 Tunde Onakoya 的故事，重申软件是“小地方诞生伟大事物”的杠杆，鼓励年轻一代利用这一工具改变世界。
[原文] [Daniel]: i wanted to ask about your optimism in the face of this uh these really seismic changes in in the technology space because a lot of programmers I see they're scared by this new technology by AIS of whatever generation of whatever species we might end up producing that uh it's going to take a lot of uh our work as uh software engineers and stuff and uh I've seen you on on multiple occasions express a lot of optimism and excitement about this particular field about software engineering so uh it's by any stretch of the imagination probably the most exciting time in software engineering and I was curious to understand why or how that feeling came to be uh in the face of reasons for being um fearful of of these new technologies
[译文] [Daniel]: 我想问问您面对技术领域这些真正的剧变时所持有的乐观态度。因为我看到很多程序员对这种新技术感到害怕，无论我们最终制造出哪一代或哪种物种的 AI，他们担心这会夺走我们要作为软件工程师的大量工作。而我在多个场合看到您对这个特定领域、对软件工程表达了很多乐观和兴奋。这无论如何都可能是软件工程最激动人心的时刻。所以我很好奇，面对那些令人恐惧的新技术理由，这种感觉是如何产生的？
[原文] [Grady]: well let me talk about the optimism and then I'll go to the issue of fear so um in many ways my career grew up from the transition from assembly language to high order programming languages so I saw the the rise of compilers and the like and the rise of high orderer languages put a lot of you know individual developers in certain jobs whole classes of jobs they didn't exist anymore you don't need you know flowchart people and you don't need key punch folks and so the rise of things like compilers all of a sudden made computation uh accessible to a vastly larger number of folks
[译文] [Grady]: 好吧，让我先谈谈乐观，然后再谈谈恐惧的问题。在很多方面，我的职业生涯是在从汇编语言到高级编程语言的过渡中成长起来的。所以我见证了编译器的兴起之类的变化。高级语言的兴起让很多——你知道——从事特定工作的个人开发者，甚至整类工作都不复存在了。你不再需要画流程图的人，也不再需要打孔员了。所以像编译器这类事物的兴起，突然之间让计算变得对更多的人触手可及。
[原文] [Grady]: and that's why I say that the history of computing history of software engineering is one of rising levels of abstraction the same thing is happening here so it destroys certain classes of jobs but it creates opportunities for completely new classes of people that simply would have not existed before
[译文] [Grady]: 这就是为什么我说，计算的历史、软件工程的历史，就是一部不断提升抽象层次的历史。同样的事情正在这里发生。所以它摧毁了某些类别的工作，但它为全新类别的人群创造了机会，而这些人以前根本不可能存在。
[原文] [Grady]: uh I've got um a friend who is a school teacher and she's been using LLMs to take a story in Hawaii in particular and recast that story to different age groups to elementary school to junior high to high school that's great so all of a sudden she now has a tool that allows her to take the essence of an idea and express it across a wider set of folks much faster than she could ever have done and that's great but remember there's a human in the loop she's the one who is able to assess the truthfulness of that and furthermore these are all stories so there's no objective truth in them but there is subjectiveness and with a human in the middle of it you know it's great
[译文] [Grady]: 呃，我有一位朋友是学校老师。她一直在使用 LLM 把一个故事——特别是在夏威夷的故事——重写给不同的年龄组，比如给小学生、初中生和高中生。这太棒了。所以突然之间，她现在拥有了一个工具，允许她提取一个想法的本质，并以比以往任何时候都更快的速度向更广泛的人群表达出来。这很好，但请记住，这里有一个“人在回路中”（human in the loop）。她是那个能够评估其真实性的人。而且，这些都是故事，所以其中没有客观真理，但有主观性。只要有人在中间把关，这就很棒。
[原文] [Grady]: so this is a classic example of where I think LLMs can provide some real useful value the danger of course is when I start using LLMs not in objective ways or sorry not in subjective ways but in objective ways like gosh I'm going to read your resume and decide whether or not I'll even consider you for an interview well we know they're full of biases we know that there's no accountability within them and so there's there's incredible danger that we can now automate racism and misogyny and bias and all these kinds of things and dismiss it because we are h we as humans uh uh remove ourselves from the process of responsibility that's I believe where the danger lies
[译文] [Grady]: 所以这是一个经典的例子，说明我认为 LLM 在哪里可以提供真正有用的价值。当然，危险在于当我开始不以主观方式，而是以客观方式使用 LLM 时。比如：“天哪，我要用它读你的简历，并决定是否考虑给你面试机会。”嗯，我们知道它们充满了偏见，我们知道它们内部没有问责机制。所以存在着巨大的危险，即我们现在可能会自动化种族主义、厌女症和偏见以及所有这类事情，并对其不屑一顾，因为我们作为人类将自己从责任过程中移除了。我相信这才是危险所在。
[原文] [Grady]: in an existential way there's a clear and present danger associated with the use of LLM is because we see them it's not I think it's going to you know kill humanity that's I don't worry about that my P doom as they say is asotically approaching zero um so I don't get I don't get worried about those things but it's the clear and present things where I see it introducing biases and and and this upon people today and how do we counter that well laws and and shining light upon it like what you and I are doing here to express to the politicians and the public that there are clear and presence dangers to them that's how we that's how we deal with it
[译文] [Grady]: 在存在主义层面上，LLM 的使用确实存在着“迫在眉睫的危险”（clear and present danger），因为我们看到了它们。我不认为它会——你知道——毁灭人类。我不担心那个。就像他们说的，我的“末日概率”（P(doom)）是渐进趋近于零的。所以我并不担心那些事情。但我看到它给今天的人们带来了偏见和其他问题，这是迫在眉睫的。我们如何对抗它？嗯，通过法律，以及像你我在这里所做的那样将其曝光，向政治家和公众表达这对他们有明确和现实的危险。这就是我们要处理它的方式。
[原文] [Grady]: last thing I'll mention back to my optimism um there's a great book titled 1000 which is basically look what was the world like at the change of that first millennium if you think about it this is a time of of science wasn't really you know big we didn't even really understand science then uh lots of myths and and and these things fueled fueled the world and there was concerns real real concerns that at the change of that millennium that it was representing the end of the world of course it didn't happen we lived through that
[译文] [Grady]: 关于我的乐观主义，我要提到的最后一件事。有一本很棒的书叫《1000年》（The Year 1000），它基本上是在看第一个千禧年更替时世界是什么样子的。如果你仔细想想，那个时代科学还不是很发达，我们甚至不真正理解科学，很多神话之类的东西驱动着世界。当时有担忧——真正的担忧——认为千禧年的更替代表着世界的终结。当然，这并没有发生，我们挺过来了。
[原文] [Grady]: go back even further in some millions of years ago um ge or or uh archaeologists and anthropologists will tell us there is a period of time that maybe 18,000 of our protohumans existed and we now all 8 billion of us rose from that 18 or so thousand that persisted so this leads me to suggest that from the year thousand to the middle ages and all that to those times we as a human species have remarkable ability to adapt and to change and to thrive and we've done it for over the millennia and I have no reason to believe that we won't continue to do that kind of thing and that by the way is what distinguishes us in many ways from the AIs that I see being created because they are not so adaptable and resilient they are still so very brittle and so very narrow in what they can do
[译文] [Grady]: 甚至追溯到更久远的数百万年前，考古学家和人类学家会告诉我们，曾有一段时间，我们的原始人类可能只剩下 18,000 人左右。而我们现在所有的 80 亿人都是从那坚持下来的约 18,000 人繁衍而来的。这让我认为，从公元 1000 年到中世纪以及所有那些时代，我们作为一个人类物种，拥有非凡的适应、改变和繁荣的能力。我们在几千年来一直这样做，我没有理由相信我们不会继续这样做。顺便说一句，这也是在很多方面将我们与我所看到的 AI 区分开来的地方，因为它们没有那么强的适应性和韧性，它们在能做的事情上仍然非常脆弱和狭隘。
[原文] [Daniel]: i think uh this is uh as good a place to uh start wrapping the conversation because uh on the optimism side there's never been a more exciting and uh full of opportunities time to uh create good software for uh to create more beauty and more utility for people and um and and I would also add that software is an incredible lever uh that the ability ility for a single person in the you know wherever they might be um they can make incredible changes
[译文] [Daniel]: 我想这是一个结束对话的好地方。因为在乐观的一面，这绝对是创造优秀软件、为人们创造更多美和实用价值的最激动人心且充满机遇的时刻。而且我也要补充一点，软件是一个不可思议的杠杆。它是那种能力，让任何地方的一个人，都能做出不可思议的改变。
[原文] [Grady]: there's a dear gentleman I' I've met him only once virtually uh tundday oh and his last name escapes me he's a chess grandmaster who really grew up from the the some of the slums in Africa and he's done this marvelous mission of bringing chess to to to kids the phrase he uses great things can come from small places this is especially true of software that the levers we have with software are astonishing in their ability uh to what they can do
[译文] [Grady]: 有一位亲爱的先生，我只在网上见过他一次，Tunde Onakoya（注：Grady口误为Oh），我想不起他的姓了。他是一位国际象棋特级大师，真正是从非洲的贫民窟中成长起来的。他完成了一项了不起的使命，将国际象棋带给孩子们。他使用的那句话是：“伟大的事物可以来自微小的地方。”（Great things can come from small places）。这对软件来说尤其如此，我们拥有的软件杠杆在能力上是惊人的。
[原文] [Grady]: i will take that and uh I also intend to propagate this because I I uh one of my other um project is teaching kids to code and I great I see kids running away with running with ideas uh to such a degree that it astonishes me and uh when when I was a kid I wish I had these opportunities so I'm very very excited for them and I would like to help them however I can so um this is a nice mission to have I would say uh it is Grady it's been an honor to have you thank you so much i want to be respectful of your time i have a million of other questions but you you are very generous with your time and um I wanted to say thank you for your work and uh for everybody else we're going to leave links to the video and u until next time thank you for listening great it's been a pleasure my pleasure aloha
[译文] [Grady]: 我会记住这一点，我也打算传播这一点，因为我的另一个项目是教孩子们写代码。我很棒地看到孩子们带着想法奔跑，那种程度让我惊讶。当我还是个孩子的时候，我希望我也能有这些机会。所以我为他们感到非常非常兴奋，我也愿意尽我所能帮助他们。
[Daniel]: 嗯，这真是一个美好的使命。Grady，能邀请到您是我的荣幸，非常感谢。我想尊重您的时间。我还有一百万个问题，但您已经非常慷慨地付出了您的时间。我想对您的工作表示感谢。对于其他人，我们会留下视频链接。下次见，谢谢收听。
[Grady]: 好的，这是我的荣幸。Aloha（夏威夷语：再见/你好）。