48 min 2026-05
Daniela Amodei, Co-Founder and President of Anthropic: Building AI the Right Way
概要
Anthropic总裁Daniela Amodei在斯坦福GSB谈AI安全的激进责任感、创业历程与AI对就业和社会的深远影响
核心洞察
- Anthropic的AI安全哲学不是"限制能力"，而是"激进责任感"(radical responsibility)——从社交媒体公司无意间引发青少年饮食障碍的教训出发，在技术引爆之前预防系统性危害。Anthropic已因网络战争风险暂停发布Mythos级模型（Project Glasswing），即便客户强烈需求。
- AI目前以互补技能而非替代为主——Anthropic经济指数（81,000人定性调查）显示，仅客服领域出现明显替代。但Daniela坦言"我们不知道未来的具体形态"，CEO们已在私下问她"我女儿还该学CS吗"。
- AI普及率远低于硅谷想象：使用者偏大学学历、偏男性、有种族和财富偏差。全球南方反而最乐观，视AI为"公平化力量"；发达国家更焦虑。
- Claude的商业差异化正在从技术走向价值观——不放广告是基于"AI对话比社交媒体更私密"的刻意选择；注册为公益公司(PBC)是创立之初就确定的架构。Daniela认为"做好事与做好生意"的张力正在消解，这一代创始人越来越多地将两者融合。
- 贯穿全场的核心线索：Daniela从自己的文学背景到全球健康到Capitol Hill到硅谷的每一次转折，都遵循同一个决策框架——"兴趣×能力×影响力"的交集。这个框架同样驱动了Anthropic的创立、AI安全哲学、以及她对AI未来的判断：技术本身不是目的，人的福祉才是。
从文学学位到AI创业：好奇心与影响力驱动的跨界人生
核心要点：Daniela的职业路径看似随机实则一致——每次转折都是在"兴趣×能力×影响力"的交集处做选择。
- 2009年毕业于文学专业——"我有一个文学学位，没有任何技能。谁会雇我？"——恰逢金融危机，起点并不光鲜。
- 第一份职业方向是国际发展和全球健康，驱动力是"世界上有人出生就没有食物、水和药物，这不公平"。虽然后来不再直接做这个领域，但这段经历奠定了"做有意义的事"的底层价值观。
- 在Capitol Hill和竞选团队工作后，发现"一小群年轻、拼命的人确实能改变世界"——这一认知最终把她引向硅谷，因为创业公司有同样的模式但"钱多得多，也没那么苦"。
- 加入Stripe时公司只有约40人，在那里待了近6年。Capitol Hill的朋友们说"你走了去做什么？支付？"——现在回看当然是好决定，但当时看起来很疯狂。
- 自认为是"通才"——"如果你看我的履历，你会说'这位女士到底擅长什么？'她没有法学学位，也不是计算机科学家。"但她认为好奇心、跨学科学习能力、以及想要产生影响的驱动力是被低估的品质。
"I really think of myself as a generalist. If you look through my background, you would be like, 'What is this lady actually good at?'" —— Daniela Amodei
2018年加入OpenAI：技术门外汉如何找到自己的位置
核心要点：学会"技术语言"的关键不是天赋，而是不怕问问题+明确自己的比较优势。
- 2018年加入OpenAI时，研究团队在讨论神经网络、transformer和scaling laws。Daniela的两个"预训练"：Stripe六年的工程师协作经验，以及从小和物理学家兄弟Dario一起长大。
- 核心心法："不要害怕技术。术语和行话一开始确实让人overwhelm，但只要你不停提问，只要身边有人愿意耐心回答——我很幸运有这样的人——你最终能理解。"
- 知道自己的lane："我大概训练不了GPT，当然也训练不了GPT-2或GPT-3。但我带来了他们不擅长的东西。"理解自己的比较优势并融入更大的生态系统，需要人际技能和好奇心。
创立Anthropic：七人"跑向"一个愿景，而非"逃离"一家公司
核心要点：Anthropic的创立不是对OpenAI的否定，而是对"安全+责任+商业"如何共存的一次从零开始的实验。
- 2020年12月，7位联合创始人离开OpenAI。Daniela强调"我们是跑向(running towards)某个东西，而不是逃离(running away)某个东西"。
- 核心愿景：创建一个将安全和责任置于最前沿的组织。7个人在OpenAI都同时涉猎能力研究、安全研究和政策工作，但觉得在新公司里更容易实现这个愿景。
- 选择注册为公益公司(PBC)花了不少时间——这是一个刻意的架构选择：承认AI会创造巨大经济价值，但坚持"以正确的方式做这件事"。
"We're running towards something versus running away from something. We have this vision of wanting to create an organization where the values that matter to us around safety, around responsibility, were the forefront of what we were doing." —— Daniela Amodei
联合创始人关系：40年兄妹默契+"先去度假再创业"
核心要点：联合创始人关系的核心是"你们画的是同一张图"，而验证方式出人意料地朴素——一起旅行、共住一间房。
- 和Dario"吵架然后和好"已经练了近40年——"因为他是我哥，我小时候偷过他的玩具。"无论怎么冲突，都确定最终还是会爱对方。
- 7位创始人之间有深厚的历史基础：认识Jared 15年，认识Chris 15年；Tom和Sam曾是室友；Jared和Sam在Stanford读PhD时共事。全部创始人都曾在OpenAI向Daniela或Dario汇报。
- Daniela的"联合创始人测试"：如果你和你的联合创始人各自锁在一间房里，画出你们想要构建的东西——走出来时"一个人画了独角兽，另一个人画了鸭嘴兽"，那就不会有好结果。
- 终极建议："别急着创业，先一起去度假。分享一间房间。如果结束后你的感觉是'天呐我只想花更多时间跟你在一起'，太好了。如果你需要'一个假期来从假期中恢复'——那可能是错误的选择。"
AI安全＝对技术的"激进责任感"——社交媒体的教训与Mythos模型管控
核心要点：AI安全不是限制技术，而是利用"后发优势"——在前一代技术公司犯过的错误上学习，在问题爆发前预防。
- Daniela将AI安全定义为"对我们开发的技术承担一种激进的责任感(radical responsibility)"。
- 社交媒体类比：Facebook/Instagram/Snapchat/Twitter的创建者并不是故意要"引发青少年饮食障碍大流行"，但他们只优化增长指标而没有预想后果。AI公司的"特权"在于可以从这些教训中学习——"我们可以说'你们犯了这个错误，我们不会再犯'。"
- 安全工作的具体范围：防止化学和生物武器开发、网络战争防御、用户健康、儿童安全、虚假信息和选举诚信。
- Mythos级模型事件：Anthropic最近因网络战争潜力决定不发布Mythos级模型（Project Glasswing）。客户反应："我们都相信网络防御，但我真的很想用那个模型。"Daniela的回应是回归使命——"我们理解那个渴望，但在我们确信所有必要的安全补丁都完成之前，发布是不负责任的。"
"We're able to say, 'You guys made this mistake. We are not gonna make that mistake this time.' That is a huge privilege." —— Daniela Amodei
安全与商业的张力在"时间"而非"方向"
核心要点：企业客户不想要不安全的模型——"没有企业说'希望Claude多产生一些幻觉'"——但随着模型能力急速进化，安全评估所需的时间正在成为真正的摩擦点。
- Anthropic收入的大部分来自企业客户(businesses)，而企业天然风险厌恶——他们不想要不可预测或不可靠的AI技术。在很长一段时间里，安全与商业100%对齐。
- 但现在进入了一个新阶段：模型能力发展太快，张力变成了"时间"——不是模型不能做amazing的事，而是"我们还没完全理解风险有多严重"，需要更多时间做安全工作。
- Glasswing的决定"不舒服"——"告诉客户这件事很不舒服。他们说'我真的很想用那个模型'。"但Anthropic选择承受这种不舒服。
AI与就业：目前互补为主，但"没人知道未来的具体形态"
核心要点：Anthropic经济指数（81,000人定性调查）显示AI目前主要是"工作的赋能者"而非替代者，唯一明显的例外是客服——"如果你要给Comcast发邮件，可能永远不会再是人类回复了。"
- 81,000人定性调查（Anthropic称这是已知最大规模的此类研究）显示：AI使用目前主要表现为互补技能(complementary skills)，只有极少数情况出现替代，主要是客服领域。
- CEO的焦虑已从会议室蔓延到家庭：CEO们在商务会议聊到三分之二时，会"阴谋般地"倾身过来说——"我女儿在Stanford读大二，她本来打算学CS，她还应该学计算机吗？"
- Daniela的判断：软件开发者仍会存在，但"不会写那么多代码了"——开发者大量的时间花在与产品经理沟通、理解客户需求上，这些工作会扩展；能被AI轻松完成的部分会收缩，但整体将创造一个"全然不同的可能性范围"。
"A CEO will conspiratorially lean across the table and say, 'My daughter is a sophomore at Stanford. She was gonna be a CS major. Should she not major in computer science?'" —— Daniela Amodei
AI普及率远低于硅谷想象：全球南方最乐观，发达国家最焦虑
核心要点：硅谷以为所有人都在用AI，但现实是使用者偏大学学历、偏男性、存在种族和财富偏差——而全球南方反而最乐观，视AI为"公平化力量"。
- Anthropic数据显示：AI使用存在显著的人口统计学偏差——偏大学学历、男性多于女性、与种族和财富相关。即使在美国，很多人还不知道如何高效使用AI工具。
- 有趣的反差：发展中国家对AI"几乎普遍乐观"——"这是我们的时刻，也许能有一个公平化的力量让事情变得更公平"。但美国、欧洲和部分亚洲地区更焦虑——"我喜欢现在这样，不希望AI来打破。"
- Daniela的判断："我们其实还在比赛的起跑线上——发令枪刚响。"仍有大量机会积极塑造这项技术的使用方式、准入方式和内嵌的价值观。
"关掉大脑"的诱惑与Claude学习模式
核心要点：81,000人调查揭示了AI时代最需警惕的认知风险——不是AI不好用，而是太好用导致人们"关掉大脑"；Claude学习模式是Anthropic对这一问题的产品级回应。
- 调查中有一类独特的感受（"可能有一天会有一个专门的词来形容"）：不是像刷手机那种感觉，而是"我本来可以去想这个问题，但不想就太容易了——直接信任AI给的答案。"Daniela认为这是AI焦虑的真正源头。
- 两种使用方式的对比讲得很生动：一种是把作业丢给ChatGPT——"有一个词来形容这种行为，叫作弊(cheating)"（全场笑）；另一种是用Claude学习模式——"就像你有一个个性化的教授，了解你、理解你最想学什么，帮你走出困境。"
- 行业层面的选择：Daniela希望整个行业选择"让人变聪明"而非"让人关掉大脑"的方向。
人际技能将五倍升值——医生"床边态度"的启示
核心要点：当AI能完成大部分诊断工作时，医生的"床边态度"(bedside manner)将变得比现在重要五倍——这个逻辑适用于所有依赖人际交互的职业。
- 今天我们雇医生主要看诊断能力——"这里可能出了什么问题，最可能是哪个，让我做些检查。"AI很快就能做到这一点。
- 但AI做不到的是"看着你、检查你、理解你的感受、让你感觉好一点"。有医学文献表明：与医生关系好的患者，临床结果更好。可能的原因是医生会"多费心去理解你的情况，多跑一组意料之外的检查"。
- Daniela的推断：在AI能完成诊断工作的世界里，"床边态度"的重要性将是现在的五倍——因为你不再需要把它挤进医生需要具备的七种素质之一。
用Claude当管理教练和育儿顾问
核心要点：Daniela自己最有感触的Claude使用场景不是写代码，而是做管理和带孩子——"人人觉得AI不会来抢自己的工作，因为自己太特别了。我也犯了这个错。"
- 管理教练场景：Daniela上传3-4年的绩效review数据，Claude帮她发现她自己看不到的模式——"你们在这个问题上已经绕了三四年了，也许需要额外的coaching。"反方向也有用：上传下属对自己的上级反馈，Claude会"非常友好地说：'看起来你在这个方面过去一年没有改进。也许你需要额外的coaching，Daniela。'"
- Daniela承认自己有"每个人都觉得AI不会替代自己"的心理——"我想的是'人们喜欢人类。他们会想向我汇报，不会想向Claude汇报，对吧？'"——但发现Claude确实是强大的管理辅助工具。
- 育儿场景：两个孩子（快5岁和快1岁），Claude帮助她度过了厕所训练——"Claude做过的最棒的事就是帮我度过厕所训练。那不是一段愉快的经历，但Claude让它好了一点——有共情，非常可操作，还有一些图表。"
- 对AI辅助育儿的判断：每次Google"你的孩子是不是有什么问题"，答案都是"是"。Claude更有分寸，更能互动，对不知所措的家长特别有用。
AI泡沫：资本支出风险是真实的，但收入增长在VC历史上前所未有
核心要点：在"估值泡沫vs基建过度投资vs技术进步可持续性"三种含义中，Daniela认为最值得关注的是资本支出风险——这是一个对未来的巨额赌注，如果收入增长逆转就会出问题。
- AI是高资本支出业务：训练模型需要大量算力，算力供不应求，价格持续上涨。企业必须提前很久购买算力——"本质上是在对未来下注"。
- 坦率承认不确定性："在这些公司工作有点让人心惊——你在做一个精心计算的赌注，赌你能在未来还上这些钱。"除了Google因体量足够大之外，Anthropic和OpenAI都面临这个挑战。
- 乐观面：两家公司的收入增长"不可思议"——VC界的共识是"从来没见过这样的事，很难想象一家公司能在这么短的时间内达到这样的收入数字"。
- 但Daniela坚持风险真实存在："如果这个趋势有一天改变了，就会出问题。这些公司为未来买了很多算力，非常贵。归根结底这是一个赌注——我们当然认为赌对了，但我们绝对可能是错的。"
AI监管与数据隐私：Claude不放广告是刻意选择
核心要点：Daniela主张技术公司和监管者"手拉手"合作——因为AI公司看得到技术被滥用的实时数据，监管者知道如何建立可执行的框架——最怕的是这个议题被政治化。
- 监管立场：支持"合理的监管(sensible regulation)"作为AI故事的一部分，但反对"监管好vs创新好"的二元对立。有些监管领域确实没意义，但有些"绝对关键"。
- 最大的担忧是政治化："我最大的期望是这个对话不要被政治化——但我担心这已经发生了。"
- 数据隐私上的具体行动：Claude不放广告，原因是"人们和AI工具的对话比他们在Instagram上发的内容更私密"。这是一个基于价值观的商业决策。
- 医疗数据场景：Claude最常见的消费级使用场景之一是问医疗问题。Daniela自己的经验是"Claude在复杂医疗案例上比我的医生更准"，但她强调"绝对不要不经专业医疗人员确认就按AI说的做"。
附录：关键人/机构/产品/数据
| 项目 | 详情 |
|------|------|
| Daniela Amodei | Anthropic联合创始人兼总裁，Dario Amodei的妹妹，英国文学专业出身 |
| Dario Amodei | Anthropic联合创始人兼CEO，Daniela的哥哥，物理学家背景 |
| Anthropic | AI安全公司，注册为公益公司(PBC)，7位联合创始人于2020年12月创立 |
| 联合创始人团队 | Daniela、Dario + Jared、Chris、Tom、Sam + 第7人（未具名） |
| Stripe | Daniela入职时约40人，她待了近6年（约2012-2018） |
| OpenAI | Daniela 2018年加入，2020年12月离开 |
| Project Glasswing | Anthropic的Mythos级模型，因网络战争风险暂停发布 |
| 经济指数 | Anthropic发布的AI使用调查，81,000人定性调查，为已知最大规模 |
| Claude学习模式 | 教育场景产品功能，引导学生思考而非直接给答案 |
| The Guns of August | Daniela推荐的书，关于一战起源中个体人物和性格的角色 |
| Sparrow Systems | Anthropic差点使用的公司名 |
| 81,000人 | Anthropic定性调查的规模，涵盖Claude和其他AI工具用户 |
中文翻译 English Original
**Interviewer：** Daniela，欢迎来到 Stanford，欢迎来到 View from the Top。
**Interviewer:** Daniela, welcome to Stanford and View from the Top.
**Daniela Amodei：** 非常感谢邀请我来。
**Daniela Amodei:** Thank you so much for having me.
**Interviewer：** 我们今天非常高兴能和你在一起。大家也看到了，今天座无虚席。我先问一下在座的各位，请举手——你们谁在用 Claude？
**Interviewer:** We're so excited to be with you here today. As you can see, we have a full house, and I want to start by asking everyone in the audience, please raise your hand if you're a user of Claude.
**Daniela Amodei：** 说实话，你们是不是本来都会举手的？咱们坦诚一点。
**Daniela Amodei:** Would you all have raised your hands? Like, let's be real.
**Interviewer：** 好吧，不管怎样，你在这里有一群非常支持你的观众。你和你哥哥一起打造了全球最重要的 AI 公司之一，但你们两个人从小都没有为此做过规划。你的背景是人文学科，学的是英国文学，早期职业是在政治圈。能不能跟大家分享一下，你最初的职业规划是什么样的？
**Interviewer:** Okay. Still, you have a very supportive crowd down here. And you and your brother helped build one of the most important AI companies in the world, yet neither of you grew up planning for any of this. Your background is in the arts. You studied English literature, and your early career was in politics. Can you share more about what was your original career plan?
**Daniela Amodei：** 天哪。首先，你用了"规划"这个词，真是太客气了。我觉得我在人生的任何阶段都不会用这个词来描述自己的路径。我的经历其实是一个我从很多人身上都听到过的故事——那些最终创办了公司、或者做了一些意想不到的事情的人，往往都是跟着当下最让自己感兴趣的事情走的。就是找到一个交汇点：我擅长什么、我对什么感兴趣、什么事情能在这个世界上产生大的影响。
所以对我来说，大学毕业的时候——顺便说一下，我是2009年毕业的，那可不是什么毕业的好年份。你拿着一个文学学位，没有什么硬技能，心想"谁会来雇我？"但那个时候，我内心有一股很强的驱动力，想要让这个世界变得更好。我觉得这一直是我和 Dario 从小到大的一个共同特征。对我来说，起步阶段是在国际发展领域，做全球健康方面的工作。我当时真正想搞清楚的是：我们怎么才能建设一个更公平的世界？让每个人都能获得食物、水和医疗这些基本的东西。
虽然我现在直接做的已经不是这些了，但我觉得那段早期的经历给了我一个思考框架：你如何在这个世界上做有意义的事情？如何去做一件有份量的、有真正目标感的事情，而不是每周把五六十个小时花在无关紧要的东西上？所以从那之后，我的路径是很曲折的。我后来去了 Capitol Hill 工作，参加了一个竞选团队，最终又回到了硅谷。我本来就是旧金山人，然后开始在 Stripe 工作——当时这是一家没什么人听说过的小公司。我在 Capitol Hill 的朋友们都说："你要走了？去做什么？支付？"现在回头看当然是个很好的选择，但当时那家公司只有大约40个人。再后来就像滚雪球一样，先去了 OpenAI，然后联合创立了 Anthropic。
所以对我来说，大学毕业的时候——顺便说一下，我是2009年毕业的，那可不是什么毕业的好年份。你拿着一个文学学位，没有什么硬技能，心想"谁会来雇我？"但那个时候，我内心有一股很强的驱动力，想要让这个世界变得更好。我觉得这一直是我和 Dario 从小到大的一个共同特征。对我来说，起步阶段是在国际发展领域，做全球健康方面的工作。我当时真正想搞清楚的是：我们怎么才能建设一个更公平的世界？让每个人都能获得食物、水和医疗这些基本的东西。
虽然我现在直接做的已经不是这些了，但我觉得那段早期的经历给了我一个思考框架：你如何在这个世界上做有意义的事情？如何去做一件有份量的、有真正目标感的事情，而不是每周把五六十个小时花在无关紧要的东西上？所以从那之后，我的路径是很曲折的。我后来去了 Capitol Hill 工作，参加了一个竞选团队，最终又回到了硅谷。我本来就是旧金山人，然后开始在 Stripe 工作——当时这是一家没什么人听说过的小公司。我在 Capitol Hill 的朋友们都说："你要走了？去做什么？支付？"现在回头看当然是个很好的选择，但当时那家公司只有大约40个人。再后来就像滚雪球一样，先去了 OpenAI，然后联合创立了 Anthropic。
**Daniela Amodei:** Oh my God. First of all, so nice of you to use the word plan in there. I don't know that I necessarily would've described it that way at any point in time. I really think this is sort of a story that I hear from a lot of people who either wind up founding a company or just doing something a little bit unexpected or unconventional with their life. It was following what was most interesting to me at the time. And the intersection of what was I good at, what was I interested in, and what was going to have a big impact in the world.
So for me, I think coming out of college, that looked like— by the way, I graduated in 2009, which was not the most fun year to be a graduate. You were like, "I have a literature degree and no skills. Who will hire me?" But at the time, I felt this very strong pull towards wanting to make the world better. I think that was always a sort of defining feature of both me and Dario from a young age. And where that started out for me was in international development, working in global health. And I think my desire there was really to figure out how do we build a world that is fair, where everybody has access to basic things like food and water and medicine?
And even though that's not what I directly work on now, I think that early grounding sort of gave me this foundation for thinking about how do you do good in the world, right? How do you build something that is of consequence and has a real purpose behind where you're spending 50 or 60 of your hours per week, right? What you're working on. And so it was kind of a winding journey from there. I worked on Capitol Hill after that. I worked on a campaign. And eventually I ended up coming back to Silicon Valley. I'm originally from San Francisco and started working at Stripe, which was this tiny company that at the time nobody had heard of. My friends on Capitol Hill were like, "You're leaving to go do what? Payments?" Now it looks like a great decision, but at the time it was about 40 people. And then from there things really just snowballed into working at OpenAI and then co-founding Anthropic.
So for me, I think coming out of college, that looked like— by the way, I graduated in 2009, which was not the most fun year to be a graduate. You were like, "I have a literature degree and no skills. Who will hire me?" But at the time, I felt this very strong pull towards wanting to make the world better. I think that was always a sort of defining feature of both me and Dario from a young age. And where that started out for me was in international development, working in global health. And I think my desire there was really to figure out how do we build a world that is fair, where everybody has access to basic things like food and water and medicine?
And even though that's not what I directly work on now, I think that early grounding sort of gave me this foundation for thinking about how do you do good in the world, right? How do you build something that is of consequence and has a real purpose behind where you're spending 50 or 60 of your hours per week, right? What you're working on. And so it was kind of a winding journey from there. I worked on Capitol Hill after that. I worked on a campaign. And eventually I ended up coming back to Silicon Valley. I'm originally from San Francisco and started working at Stripe, which was this tiny company that at the time nobody had heard of. My friends on Capitol Hill were like, "You're leaving to go do what? Payments?" Now it looks like a great decision, but at the time it was about 40 people. And then from there things really just snowballed into working at OpenAI and then co-founding Anthropic.
**Interviewer：** 你跨越了多个领域，从来不被自己学过什么、做过什么所限制。这种"你的背景不必定义你下一步该做什么"的思维方式，是从哪里来的？
**Interviewer:** You've moved across fields without being constrained by what you studied or had done before. Where does that mindset come from, that your background doesn't have to define your next move?
**Daniela Amodei：** 这个问题很有意思。我觉得在某种程度上，我真的把自己定位成一个通才。如果你翻看我的履历，你可能会想："这个人到底擅长什么？"她没有法学学位，也不是计算机科学家。但我觉得有一种能力是被低估的：能够对很多不同的学科保持好奇心、持续学习，同时无论在哪个领域工作都有一个很强的"要产生影响"的底层驱动。我在 Anthropic 招聘的人身上经常看到这种特质。在更广泛的科技行业里，我合作过的真正优秀的人也是这样——他们好奇、聪明、愿意学习、乐于助人。说实话，这基本上就是创业公司里除了工程师以外每一个岗位的描述。
但对我来说，一直是兴趣和影响力在驱动我。我会想："有一件事让我觉得很不对——我出生在美国，生活中那些我们习以为常的基本条件我都拥有，但世界上有些人根本没有，不是因为他们做了什么，只是因为他们出生的地方不同、环境不同。我怎么才能让这件事更公平？我们作为一个全球社区，怎么才能做得更好？"然后我发现，"我现在产生的影响力还不够，我需要一些技能。"于是我去参加了一个竞选团队，然后发现，"一小群年轻、有干劲、拼命工作的人，真的可以改变世界。"从那里到硅谷其实是很自然的事——你发现在创业公司里也可以做到这一点，而且钱更多，日子也比竞选团队每周干80个小时要好过得多。
但我觉得核心还是那些底层品质——真正追随你的热情。因为当你真心在意自己正在做的事情时，无论是出于智识上的兴趣还是出于一种内在的意义感，你就会自发地想要做更多。
但对我来说，一直是兴趣和影响力在驱动我。我会想："有一件事让我觉得很不对——我出生在美国，生活中那些我们习以为常的基本条件我都拥有，但世界上有些人根本没有，不是因为他们做了什么，只是因为他们出生的地方不同、环境不同。我怎么才能让这件事更公平？我们作为一个全球社区，怎么才能做得更好？"然后我发现，"我现在产生的影响力还不够，我需要一些技能。"于是我去参加了一个竞选团队，然后发现，"一小群年轻、有干劲、拼命工作的人，真的可以改变世界。"从那里到硅谷其实是很自然的事——你发现在创业公司里也可以做到这一点，而且钱更多，日子也比竞选团队每周干80个小时要好过得多。
但我觉得核心还是那些底层品质——真正追随你的热情。因为当你真心在意自己正在做的事情时，无论是出于智识上的兴趣还是出于一种内在的意义感，你就会自发地想要做更多。
**Daniela Amodei:** So, you know, it's really interesting. I think in some ways I really think of myself as a generalist, right? If you sort of look through my background, you would be like, "What is this lady actually good at," right? She doesn't have a law degree. She's not a computer scientist. But I think this concept that... the ability to be curious and learn across a lot of disciplines, to have a strong foundation of wanting to have impact regardless of the area that you're working on. I think that's an underrated quality. And I see it a lot in the people that we hire at Anthropic. And the really talented people that I've worked with in the technology industry more broadly, people who are curious and smart and they want to learn and they want to be helpful. That is like the, first of all, the description of like every role except for like engineer at a startup.
But I think for me, it was always very, it was very interest and impact driven. So I was like, "Okay, there's something that feels really wrong to me about the fact that, you know, I was born in America. I had access to all of the sort of basic things that we take for granted in life. Some people around the world just weren't. That's just not where they were born. That's not what they were born into. How do I make that more fair? Like, how do we make that better as a broader community of the world?" And from there, I was like, "I'm just not having the level of impact that I want to. I need some skills." And so I went and I worked on a campaign, and I was like, "Wow," a small number of people that work really hard who are young and driven can really change the world." Not that surprising that eventually that led me to Silicon Valley, where you're like, "You can do that," but it turns out that at a startup, you have a lot more money, and it's a lot easier going than the kind of 80 hours a week of working on a campaign.
But I think those kind of core qualities around, really following your passion, because you just want to do more when you care about the thing that you're working on, either intellectually or from an implicit sense of meaning.
But I think for me, it was always very, it was very interest and impact driven. So I was like, "Okay, there's something that feels really wrong to me about the fact that, you know, I was born in America. I had access to all of the sort of basic things that we take for granted in life. Some people around the world just weren't. That's just not where they were born. That's not what they were born into. How do I make that more fair? Like, how do we make that better as a broader community of the world?" And from there, I was like, "I'm just not having the level of impact that I want to. I need some skills." And so I went and I worked on a campaign, and I was like, "Wow," a small number of people that work really hard who are young and driven can really change the world." Not that surprising that eventually that led me to Silicon Valley, where you're like, "You can do that," but it turns out that at a startup, you have a lot more money, and it's a lot easier going than the kind of 80 hours a week of working on a campaign.
But I think those kind of core qualities around, really following your passion, because you just want to do more when you care about the thing that you're working on, either intellectually or from an implicit sense of meaning.
**Interviewer：** 你的 AI 职业生涯始于2018年加入 OpenAI，当时它还是一个相对小规模的研究实验室。突然之间，你进入了一个大家都在讨论神经网络、Transformer、scaling laws 的环境。你是怎么学会说这门"语言"的？
**Interviewer:** And your career in AI started when you joined OpenAI in 2018, when it was still a relatively small research lab. Suddenly you were in rooms where people were talking about neural networks and transformers and scaling laws. How did you learn to speak that language?
**Daniela Amodei：** 我觉得我在两个维度上做了很好的准备。第一，我之前已经在 Stripe 待了将近六年，跟很多工程师合作过。研究和工程当然不同，但我觉得有一些重叠和基础。第二，我有一个非常有才华的、技术导向的物理学家作为兄弟，也就是我在 Anthropic 的联合创始人——当然还有另外五位联合创始人，他们也都是工程师或研究员。但我觉得最重要的一点是，这两段经历都让我培养了一种心态：不要害怕技术。归根结底，技术是一套非常受重视的技能，但它是任何人都可以学会的东西。
技术背后的基础知识，我觉得那些术语和行话一开始确实会让人觉得不知所措。但只要你不停地问问题，而且身边有足够耐心的人愿意回答你——我很幸运，我的生活中有这样的人——你就一直问，问到自己觉得能理解为止。另外一点是，要清楚自己的边界和别人的边界。研究员能做的很多事情我做不了，我大概训练不出 GPT，当然也训练不出 GPT-2 或 GPT-3。但我能带来他们不擅长的东西。所以我觉得，真正理解自己的比较优势在哪里、理解自己在更大的生态系统中扮演什么角色，这需要很多人际层面的技能。好奇心是一种天生的品质，但也是可以学习和训练的。我觉得这些东西综合在一起，真的让我能够在那种角色中取得成功。
技术背后的基础知识，我觉得那些术语和行话一开始确实会让人觉得不知所措。但只要你不停地问问题，而且身边有足够耐心的人愿意回答你——我很幸运，我的生活中有这样的人——你就一直问，问到自己觉得能理解为止。另外一点是，要清楚自己的边界和别人的边界。研究员能做的很多事情我做不了，我大概训练不出 GPT，当然也训练不出 GPT-2 或 GPT-3。但我能带来他们不擅长的东西。所以我觉得，真正理解自己的比较优势在哪里、理解自己在更大的生态系统中扮演什么角色，这需要很多人际层面的技能。好奇心是一种天生的品质，但也是可以学习和训练的。我觉得这些东西综合在一起，真的让我能够在那种角色中取得成功。
**Daniela Amodei:** So I think I was well-trained on two dimensions. I think the first is I had already spent almost six years at Stripe, and so I had worked with a lot of engineers. Obviously, research and engineering are different, but I think there was some overlap and some grounding. The second is that I grew up with a very talented technology-oriented physicist who's my sibling, and obviously my co-founder at Anthropic, among five others who are also all engineers or researchers. But I think the number one thing I would say is both of those experiences just instilled in me this sense to just not be afraid of technology, right? It was ultimately— it's a set of skills that are really highly prized, I think, but it's something that anybody can learn.
And the sort of basics behind it, I think the terminology and the jargon can feel overwhelming at first. But if you just ask enough questions, and if you have people who are kind enough to be patient with you— which I had in my life, I was very lucky for that. I just kept asking questions until I felt like I could understand it. And I think, you know, the second part of it was also just knowing my lane and their lane. So there were a lot of things that the researchers, I was like, I probably couldn't have trained GPT, I certainly couldn't have trained GPT-2 or GPT-3. But I brought things to the table that they didn't know how to do as well. And so I really think understanding what your comparative advantage is and knowing how you fit into the broader ecosystem, that takes a lot of skills that are kind of interpersonal. Curiosity, I think, is a very inherent skill but one that you can learn and train. And I think all of those sort of put together really put me in this position to be able to be successful in that type of role.
And the sort of basics behind it, I think the terminology and the jargon can feel overwhelming at first. But if you just ask enough questions, and if you have people who are kind enough to be patient with you— which I had in my life, I was very lucky for that. I just kept asking questions until I felt like I could understand it. And I think, you know, the second part of it was also just knowing my lane and their lane. So there were a lot of things that the researchers, I was like, I probably couldn't have trained GPT, I certainly couldn't have trained GPT-2 or GPT-3. But I brought things to the table that they didn't know how to do as well. And so I really think understanding what your comparative advantage is and knowing how you fit into the broader ecosystem, that takes a lot of skills that are kind of interpersonal. Curiosity, I think, is a very inherent skill but one that you can learn and train. And I think all of those sort of put together really put me in this position to be able to be successful in that type of role.
**Interviewer：** 2020年12月，你和你哥哥以及一群同事离开了 OpenAI。你和 Dario 为什么决定创办 Anthropic？
**Interviewer:** And in December of 2020, you, your brother, along with a group of colleagues, left OpenAI. Why did you and Dario decide to start Anthropic?
**Daniela Amodei：** 回答这个问题之前，让我先喝一大口水。
**Daniela Amodei:** I'm gonna take a big sip of water for that one.
**Interviewer：** 慢慢来。
**Interviewer:** Take your time.
**Daniela Amodei：** 我再喝一口。好，当时一共有七个人离开——我和 Dario，加上我们的五位联合创始人，之后又陆续有一些人跟过来。对我们来说，核心原因是我们真正想要这项技术最终产生什么样的影响。我们七个人虽然各不相同，但都是非常正直的人。我们都非常在意自己所构建的东西会带来什么影响。最终我们觉得，在一家新公司里去实现我们心目中的愿景，会比在原来的公司里更容易。
我经常说——我知道你们会很震惊——我其实被问这个问题的频率还挺高的。我们真的是在"奔向"某个东西，而不是"逃离"某个东西。我们脑海中有一个愿景，想要创建一个组织，在这里安全、责任感这些我们最看重的价值观，能够真正处于我们工作的最前端。所以我们选择注册为公益公司（public benefit corporation）。我们花了不少时间才找到合适的组织形式来表达这个理念：我们会是一个商业实体，我们认为人工智能会创造巨大的经济价值，但我们极其看重用正确的方式来做这件事。这一点把我们七个人凝聚在了一起。我们在 OpenAI 的时候，每个人都同时参与过能力（capabilities）和安全（safety）以及政策方面的工作，所以换一个新的形式来搭建这个架构，感觉更自然。
我经常说——我知道你们会很震惊——我其实被问这个问题的频率还挺高的。我们真的是在"奔向"某个东西，而不是"逃离"某个东西。我们脑海中有一个愿景，想要创建一个组织，在这里安全、责任感这些我们最看重的价值观，能够真正处于我们工作的最前端。所以我们选择注册为公益公司（public benefit corporation）。我们花了不少时间才找到合适的组织形式来表达这个理念：我们会是一个商业实体，我们认为人工智能会创造巨大的经济价值，但我们极其看重用正确的方式来做这件事。这一点把我们七个人凝聚在了一起。我们在 OpenAI 的时候，每个人都同时参与过能力（capabilities）和安全（safety）以及政策方面的工作，所以换一个新的形式来搭建这个架构，感觉更自然。
**Daniela Amodei:** I'll take one more, actually. So, there were seven of us that originally left, me and Dario and our five co-founders, and then a number of people that kind of came over shortly thereafter. And I think for us, it really came down to just the focus on what it was we really wanted the ultimate impact of the technology to be. And I think in our own very different ways, all seven of us are people that have— a lot of integrity. We're people who care a lot about the impact of what it is that we build. And I think eventually it felt like it was just easier for us to sort of create the type of vision that we saw outside of the company we were in and in a new company.
And I frequently say, because I know it'll shock you, I'm actually asked this question not infrequently. And... you know, we really, I think we're running towards something versus running away from something. Like, we have this vision in our heads of wanting to create an organization where the values that matter to us around safety, around responsibility, were the kind of forefront of what we were doing. That's why we chose to incorporate as a public benefit corporation. That took a while to figure out what was the right kind of form factor to express: Look, we are going to be a commercial entity. We think there's going to be a lot of economic value that's going to be created by artificial intelligence, but it's really important to us that we do this the right way. And I think that was something that kind of united the seven of us. We had all worked on a combination of both capabilities and safety and policy work when we were at OpenAI, and it felt like it was just easier to kind of create this structure in a new form.
And I frequently say, because I know it'll shock you, I'm actually asked this question not infrequently. And... you know, we really, I think we're running towards something versus running away from something. Like, we have this vision in our heads of wanting to create an organization where the values that matter to us around safety, around responsibility, were the kind of forefront of what we were doing. That's why we chose to incorporate as a public benefit corporation. That took a while to figure out what was the right kind of form factor to express: Look, we are going to be a commercial entity. We think there's going to be a lot of economic value that's going to be created by artificial intelligence, but it's really important to us that we do this the right way. And I think that was something that kind of united the seven of us. We had all worked on a combination of both capabilities and safety and policy work when we were at OpenAI, and it felt like it was just easier to kind of create this structure in a new form.
**Interviewer：** 你不仅和你哥哥一起创建了 Anthropic，还有另外五位联合创始人。在座的很多人即将第一次选择自己的联合创始人，我们都知道这件事经常会以失败告终。怎样才能让这种关系运转起来？
**Interviewer:** And you're building Anthropic not only with your brother, but with five other co-founders. Many of us here are about to choose co-founders for the first time, and we all know how often that ends badly. What does it take to make it work?
**Daniela Amodei：** 我觉得自己真的非常幸运。我们七个人在好几个维度上都是一个非常特别的组合。首先，人际关系——我知道这听起来不意外——但它比你想象的重要得多。你们之间会怎么处理冲突？举个例子，我和 Dario 已经吵架又和好将近40年了，因为他是我哥哥，我小时候偷过他的玩具。所以我们知道怎么一起度过冲突，而且不管怎样，最后我们肯定还是会爱对方的。
至于我们的联合创始人——我认识 Jared 大约15年了，认识 Chris 也有15年。Tom 和 Sam 曾经是室友。Jared 和 Sam 在 Stanford 读博的时候就一起共事过。所以有这样一段很长的共同历史。而且我和 Dario 实际上管理过所有其他联合创始人，他们要么向我们中的一个人汇报，要么向我们两个人都汇报过。我印象中大多数人在 OpenAI 时是向我们两个都汇报的。所以我们已经有了一套现成的框架——我们知道怎么给彼此反馈，知道怎么一起工作，了解彼此是什么样的人。
另外一件非常重要的事是，确保你们对"要做什么"有一个非常清晰且一致的认识。如果你把自己和联合创始人分别锁进一个房间，各自画出你们想要构建的东西，走出来的时候不能一个人画了独角兽、另一个人画了鸭嘴兽。那种情况——你们以为自己在做同一件事，但实际上不是——最后不会有好结果。对我们来说，我们很幸运，因为我们之前就在一个环境里，已经形成了"不太是这个方向，我们想做另一件事"的共识。所以我们在起步时就已经被筛选出来，在兴趣和方向上是一致的。
但我觉得，在力所能及的范围内多做一些压力测试是很重要的。与其直接一起创业，不如先一起去度个假，看看效果怎么样——跟他们同住一个房间，看感觉如何。如果你的反应是"我只想跟你多待一会儿"，那太好了。如果你的反应是"我需要再度一次假来从这次度假中恢复过来"，那可能就不是正确的选择。
至于我们的联合创始人——我认识 Jared 大约15年了，认识 Chris 也有15年。Tom 和 Sam 曾经是室友。Jared 和 Sam 在 Stanford 读博的时候就一起共事过。所以有这样一段很长的共同历史。而且我和 Dario 实际上管理过所有其他联合创始人，他们要么向我们中的一个人汇报，要么向我们两个人都汇报过。我印象中大多数人在 OpenAI 时是向我们两个都汇报的。所以我们已经有了一套现成的框架——我们知道怎么给彼此反馈，知道怎么一起工作，了解彼此是什么样的人。
另外一件非常重要的事是，确保你们对"要做什么"有一个非常清晰且一致的认识。如果你把自己和联合创始人分别锁进一个房间，各自画出你们想要构建的东西，走出来的时候不能一个人画了独角兽、另一个人画了鸭嘴兽。那种情况——你们以为自己在做同一件事，但实际上不是——最后不会有好结果。对我们来说，我们很幸运，因为我们之前就在一个环境里，已经形成了"不太是这个方向，我们想做另一件事"的共识。所以我们在起步时就已经被筛选出来，在兴趣和方向上是一致的。
但我觉得，在力所能及的范围内多做一些压力测试是很重要的。与其直接一起创业，不如先一起去度个假，看看效果怎么样——跟他们同住一个房间，看感觉如何。如果你的反应是"我只想跟你多待一会儿"，那太好了。如果你的反应是"我需要再度一次假来从这次度假中恢复过来"，那可能就不是正确的选择。
**Daniela Amodei:** So, you know, I think of myself as just extremely lucky. Like, the seven of us, I think, are a very special group on a couple of dimensions. I think the first thing I would say is, like— the interpersonal relationships, I know this is gonna sound not that surprising, matter a lot more than you think. Like, how do you guys have conflict together? So for example, Dario and I have been fighting and getting over it for almost 40 years. Because he's my brother, and I used to steal his toys, right? So we know how to work through conflict together, and there's no question we will love each other at the end of that, no matter what.
With our co-founders, it's like I've known Jared for like 15 years. I've known Chris for 15 years. Tom and Sam were roommates. Jared and Sam worked together at Stanford when they were in their PhD program. So there was this kind of long history. Both Dario and I actually managed— all of the other co-founders, and I think they had either reported to one of us or both of us. I think the majority had reported to both of us at OpenAI. So we already had this kind of existing structure and framework of we knew how to give each other feedback. We knew how to work together. We'd understood who we were as people.
And I think the other really important thing is just making sure you have a very strong sense of what it is you're trying to do, and that that picture is the same. Like, if you locked yourself and your co-founder in another room, and you wrote down or drew a picture of what it is you're trying to build, you're not gonna walk out and one has drawn a unicorn and the other has drawn a platypus, right? Like, that's the type of situation where you think you're doing the same thing, but I think it just doesn't end well. And I think for us, this kind of vision of what it is we wanted to build, in some ways we were lucky because we had been in an environment where we're like, "Ooh, it's not quite this. We wanna do this other thing." But we were in the same zone already. So we're sort of preselected for that level of interest.
But I think really being able to pressure test to the degree that you can. Instead of starting a company together, go on vacation together. Just see how that goes— share a room with them. Be like, "How did that go?" Um, and if you're like, "Man, all I wanna do is spend more time with you," great. If you're like, "Really, I'm gonna need a vacation to recover from my vacation," you just might— it might be the wrong choice.
With our co-founders, it's like I've known Jared for like 15 years. I've known Chris for 15 years. Tom and Sam were roommates. Jared and Sam worked together at Stanford when they were in their PhD program. So there was this kind of long history. Both Dario and I actually managed— all of the other co-founders, and I think they had either reported to one of us or both of us. I think the majority had reported to both of us at OpenAI. So we already had this kind of existing structure and framework of we knew how to give each other feedback. We knew how to work together. We'd understood who we were as people.
And I think the other really important thing is just making sure you have a very strong sense of what it is you're trying to do, and that that picture is the same. Like, if you locked yourself and your co-founder in another room, and you wrote down or drew a picture of what it is you're trying to build, you're not gonna walk out and one has drawn a unicorn and the other has drawn a platypus, right? Like, that's the type of situation where you think you're doing the same thing, but I think it just doesn't end well. And I think for us, this kind of vision of what it is we wanted to build, in some ways we were lucky because we had been in an environment where we're like, "Ooh, it's not quite this. We wanna do this other thing." But we were in the same zone already. So we're sort of preselected for that level of interest.
But I think really being able to pressure test to the degree that you can. Instead of starting a company together, go on vacation together. Just see how that goes— share a room with them. Be like, "How did that go?" Um, and if you're like, "Man, all I wanna do is spend more time with you," great. If you're like, "Really, I'm gonna need a vacation to recover from my vacation," you just might— it might be the wrong choice.
**Interviewer：** 我想回到你刚才说的一点。Anthropic 与 AI 安全紧密相关，但我想确保在座的每个人都真正理解这意味着什么。当你说 AI 安全时，你指的是什么？
**Interviewer:** I want to circle back to what you just said a few moments ago. Anthropic is deeply associated with AI safety, but I want to make sure that everyone understands what that actually means. So when you say AI safety, what do you mean?
**Daniela Amodei：** 这是一个很好的问题。我觉得"AI 安全"这个词在过去几年已经有些超载了，因为它变成了一个笼统的说法，在很多语境下被使用。但对我们来说，最高层次的定义就是：对我们正在开发的技术承担一种激进的责任感。我们经常用的一个类比是社交媒体公司——现在公开场合批评它们已经非常流行了，所以我也来批评一下。
基本上，如果你想象回到过去，我其实认为那些创建社交媒体公司的开发者，他们当初并不是说"我要去引发一场青少年女孩饮食障碍的大流行"。那不是他们的本意。但他们想的是："我要优化什么指标？我要建一家公司，我想看到极速增长，那就朝着这个方向去做。"在那个时候，这种前瞻性思考并不被认为是必要的，因为我们从来没有见过如此大规模的东西——这些公司增长得多快、多少人在多短的时间内就开始使用这些技术。
但你可以想象，如果你能回到过去，对他们说："你正在创办 Facebook 或 Instagram 或 Snapchat 或 Twitter，如果我真的去认真想一想这件事可能以哪些方式出问题呢？如果我能想清楚所有那些意料之外的负面效应，并且提前尝试去阻止其中一些呢？"对于 AI 来说，这有点不公平，因为我们之前已经有了一整代技术，它们已经犯过错了，而我们可以站在旁边说"我们不会再犯同样的错误。"但这是一种巨大的特权。我们能够说："你们犯了这个错误，我们这次不会再犯了。我们会很谨慎。我们会去思考，因为我们更了解这项技术，还有哪些其他的事情可能出问题？我们怎样去想象一个一切都顺利的世界，同时也想象一个一切都出问题的世界？"
对我们来说，安全意味着所有那些重大风险——防止化学和生物武器利用我们的技术被开发出来，这种可能性是存在的。还有网络安全。最近我们决定不发布 Mythos 级别的模型，因为存在网络战争的潜在风险，这件事被广泛报道了。此外还有大量围绕用户健康、儿童安全、虚假信息、选举诚信等方面的工作。这些不是我们新发明的。我们站在了前一代安全和安保团队的肩膀上——他们在历史上一些最有影响力的科技公司里做过这些工作——然后我们问自己："我们怎么向你们学习？怎么做得更好？"
基本上，如果你想象回到过去，我其实认为那些创建社交媒体公司的开发者，他们当初并不是说"我要去引发一场青少年女孩饮食障碍的大流行"。那不是他们的本意。但他们想的是："我要优化什么指标？我要建一家公司，我想看到极速增长，那就朝着这个方向去做。"在那个时候，这种前瞻性思考并不被认为是必要的，因为我们从来没有见过如此大规模的东西——这些公司增长得多快、多少人在多短的时间内就开始使用这些技术。
但你可以想象，如果你能回到过去，对他们说："你正在创办 Facebook 或 Instagram 或 Snapchat 或 Twitter，如果我真的去认真想一想这件事可能以哪些方式出问题呢？如果我能想清楚所有那些意料之外的负面效应，并且提前尝试去阻止其中一些呢？"对于 AI 来说，这有点不公平，因为我们之前已经有了一整代技术，它们已经犯过错了，而我们可以站在旁边说"我们不会再犯同样的错误。"但这是一种巨大的特权。我们能够说："你们犯了这个错误，我们这次不会再犯了。我们会很谨慎。我们会去思考，因为我们更了解这项技术，还有哪些其他的事情可能出问题？我们怎样去想象一个一切都顺利的世界，同时也想象一个一切都出问题的世界？"
对我们来说，安全意味着所有那些重大风险——防止化学和生物武器利用我们的技术被开发出来，这种可能性是存在的。还有网络安全。最近我们决定不发布 Mythos 级别的模型，因为存在网络战争的潜在风险，这件事被广泛报道了。此外还有大量围绕用户健康、儿童安全、虚假信息、选举诚信等方面的工作。这些不是我们新发明的。我们站在了前一代安全和安保团队的肩膀上——他们在历史上一些最有影响力的科技公司里做过这些工作——然后我们问自己："我们怎么向你们学习？怎么做得更好？"
**Daniela Amodei:** Yeah, this is a great question. And I think this is a term that's gotten almost a little bit overloaded in the past few years because it's kind of a catch-all that I think has been reached for in a lot of contexts. But I think to us, the kind of highest level framing is just taking a form of radical responsibility for the technology that we're developing. And the analog that we often use or that we often point to is— social media companies, which, like, it's so in vogue to, like, crap all over them in public now, so I'm gonna do that.
But basically, you know, like if you imagine going back in time, I actually think developers who created these technology companies, they weren't like, "I'm setting out to cause a pandemic of eating disorders for teenage girls." Right? That was not their intention. But they were like, "What are the metrics that I'm trying to optimize for?" Right? "I'm trying to build a company. I would like to see rapid, rapid scale growth. Like, let's just build towards that." And there wasn't this sort of... at the time it just wasn't necessary 'cause we'd never seen something on the level of scale that we have seen today, how quickly these companies grew, how many people adopt the technologies so quickly.
But if you could imagine, you could go back in time and say, like, "Wow, you're starting Facebook or Instagram or Snapchat or Twitter," and you're like, "What if I really tried to think through all of the ways that this could go wrong?" Like, what? What if I could think about what all the unintended externalities are, and really just kind of tried in advance to prevent some of those from happening? And it's a little bit unfair because in AI, we've had this whole generation of technologies before us where they've gotten to fuck up, and we've gotten to be like, "Ha ha, we're not gonna do that thing again." But that is a huge privilege. Right? We're able to say, "Okay, you guys made this mistake. We are not gonna make that mistake this time." Right? We're gonna be careful, and we're going to say, how do we make sure to think about the other things that might not go wrong because we understand the technology better? How can we? How can we imagine a world where everything goes right, but also a world where everything goes wrong?
And I think for us, safety means all of the big stuff, so preventing chemical and biological weapons from being developed using our technology, which, by the way, they could have the potential to do. But also cyber, right? Cyber warfare. We've been in the news a lot lately about our decision to not release our Mythos-class model because of the potential for cyber warfare. There's also a lot of work that happens around things like user wellness, child safety, a lot of misinformation, election integrity work. This is not something new that we invented. We've been able to stand on the shoulders of previous safety and security teams who worked on this at some of the other most consequential technology companies in history and say, "How do we learn from you? How do we do this better?"
But basically, you know, like if you imagine going back in time, I actually think developers who created these technology companies, they weren't like, "I'm setting out to cause a pandemic of eating disorders for teenage girls." Right? That was not their intention. But they were like, "What are the metrics that I'm trying to optimize for?" Right? "I'm trying to build a company. I would like to see rapid, rapid scale growth. Like, let's just build towards that." And there wasn't this sort of... at the time it just wasn't necessary 'cause we'd never seen something on the level of scale that we have seen today, how quickly these companies grew, how many people adopt the technologies so quickly.
But if you could imagine, you could go back in time and say, like, "Wow, you're starting Facebook or Instagram or Snapchat or Twitter," and you're like, "What if I really tried to think through all of the ways that this could go wrong?" Like, what? What if I could think about what all the unintended externalities are, and really just kind of tried in advance to prevent some of those from happening? And it's a little bit unfair because in AI, we've had this whole generation of technologies before us where they've gotten to fuck up, and we've gotten to be like, "Ha ha, we're not gonna do that thing again." But that is a huge privilege. Right? We're able to say, "Okay, you guys made this mistake. We are not gonna make that mistake this time." Right? We're gonna be careful, and we're going to say, how do we make sure to think about the other things that might not go wrong because we understand the technology better? How can we? How can we imagine a world where everything goes right, but also a world where everything goes wrong?
And I think for us, safety means all of the big stuff, so preventing chemical and biological weapons from being developed using our technology, which, by the way, they could have the potential to do. But also cyber, right? Cyber warfare. We've been in the news a lot lately about our decision to not release our Mythos-class model because of the potential for cyber warfare. There's also a lot of work that happens around things like user wellness, child safety, a lot of misinformation, election integrity work. This is not something new that we invented. We've been able to stand on the shoulders of previous safety and security teams who worked on this at some of the other most consequential technology companies in history and say, "How do we learn from you? How do we do this better?"
**Interviewer：** Anthropic 是一家 AI 安全公司，但同时也必须创造收入。你怎么管理这两者之间的张力？
**Interviewer:** Anthropic is an AI safety company that also has to generate revenue. How do you manage the tension between the two?
**Daniela Amodei：** 我们也经常被问到这个问题。总体来说，这两者产生冲突的频率比你想象的要低很多。我想说的是，大多数企业——而且企业客户是我们收入的主要来源——并不想要不安全的模型。他们不会说"我们希望 Claude 产生更多幻觉"或者"如果 Claude 在你提问的时候能输出有害内容就好了。"所以我觉得，直到最近为止，安全和商业完全是一致的。做安全对商业真的是好事，因为企业客户天然是风险规避型的。他们不想要那些高度不可预测或不可靠的 AI 技术。
话虽如此，我觉得我们现在正在进入一个新的阶段——模型的能力发展得非常快，张力出现在时间上。不是说模型不能做出惊人的事情，而是我们在现阶段还没有完全搞清楚风险到底有多严重、到底有哪些风险、以及我们如何去减轻这些风险。这有时意味着我们会采取一些不太寻常的行动，就像我们在 Project Glasswing 上做的那样——"这个新一代模型，我们当然希望能直接发布给所有客户，他们都想用。但我们还不够有信心，我们需要更多时间来做安全方面的工作。"但这是让人不舒服的。跟客户这样说是不舒服的。客户的反应是："我们都支持网络防御，但我真的很想用那个模型。"我觉得在这个时候，我们只能回归使命。我们说："我们理解你的需求，我们也希望尽快把这项技术交到你手上，但在我们确信所有需要修补的地方都已经修补好之前，贸然发布是不负责任的。"
话虽如此，我觉得我们现在正在进入一个新的阶段——模型的能力发展得非常快，张力出现在时间上。不是说模型不能做出惊人的事情，而是我们在现阶段还没有完全搞清楚风险到底有多严重、到底有哪些风险、以及我们如何去减轻这些风险。这有时意味着我们会采取一些不太寻常的行动，就像我们在 Project Glasswing 上做的那样——"这个新一代模型，我们当然希望能直接发布给所有客户，他们都想用。但我们还不够有信心，我们需要更多时间来做安全方面的工作。"但这是让人不舒服的。跟客户这样说是不舒服的。客户的反应是："我们都支持网络防御，但我真的很想用那个模型。"我觉得在这个时候，我们只能回归使命。我们说："我们理解你的需求，我们也希望尽快把这项技术交到你手上，但在我们确信所有需要修补的地方都已经修补好之前，贸然发布是不负责任的。"
**Daniela Amodei:** We're asked this question a lot too. And I think, in general, the two don't come into conflict as much as you would expect. So what I would say is, most businesses in particular, which is, by the way, the majority of our revenue comes from businesses, are not looking to have models that are unsafe, right? They're not like, "Wow, we would love for Claude to hallucinate more," or— "it would be great if Claude just produced like harmful outputs when you're asking it a question." And so I think until like pretty recently, these things were just like 100% in alignment, right? Yeah. You're like, "It's actually really good for business to be safe because businesses are correctly, they're risk-averse, right? They're like, 'We don't want AI technologies that are going to be super unpredictable or unreliable.'"
That said, I think we're now entering an era where the capabilities of the models are developing so rapidly that the tension is about time. So it's not necessarily the case that the models can't do amazing things. It's just we don't fully understand at this stage, and I think it will be more the case going forward, how serious are the risks? What are all of the risks, and how do we help mitigate them? That sometimes means that we take slightly unusual actions like we did with Project Glasswing and say, "This new class of model, it would be great if we could just release this to all of our customers. They all would love to use it, but we're just not confident enough yet. Like, we need a little more time to do some of the work to make the model safer to use." But that's uncomfortable. Right? It's uncomfortable to say that to your customers, right? They're like, "Look, we all believe in cyber defense, but I really want access to that model." And I think this is the place where we just come back to the mission, right? We say, "Okay, we understand that desire. We want to get this technology to you as quickly as possible, but it is irresponsible of us to release it until we are confident that all of the patching that needs to be done has been done."
That said, I think we're now entering an era where the capabilities of the models are developing so rapidly that the tension is about time. So it's not necessarily the case that the models can't do amazing things. It's just we don't fully understand at this stage, and I think it will be more the case going forward, how serious are the risks? What are all of the risks, and how do we help mitigate them? That sometimes means that we take slightly unusual actions like we did with Project Glasswing and say, "This new class of model, it would be great if we could just release this to all of our customers. They all would love to use it, but we're just not confident enough yet. Like, we need a little more time to do some of the work to make the model safer to use." But that's uncomfortable. Right? It's uncomfortable to say that to your customers, right? They're like, "Look, we all believe in cyber defense, but I really want access to that model." And I think this is the place where we just come back to the mission, right? We say, "Okay, we understand that desire. We want to get this technology to you as quickly as possible, but it is irresponsible of us to release it until we are confident that all of the patching that needs to be done has been done."
**Interviewer：** 我们不能否认，一提到 AI，人们有很多恐惧。人们担心 AI 意味着更少的工作，因为对人类判断力的需求会下降。这种恐惧是合理的吗？
**Interviewer:** We can't deny that when it comes to AI, there's a lot of fear. Fear that AI means fewer jobs because of less need for human judgment. Is that fear valid?
**Daniela Amodei：** 我觉得这其实是一个非常复杂的问题。我的判断是——这对在座的各位来说可能不意外——AI 会改变可获得的工作类型和人们所从事的工作类型。今天，有一些五年前因为 AI 还不存在的工作岗位。我也认为未来会有一些工作岗位因为 AI 而消失。但到目前为止我们观察到的是——根据我们的经济指数（economic index），研究人们实际上是怎么使用人工智能技术的——目前主要呈现为互补型技能，AI 是在赋能工作，而不是替代工作，只有极少数情况是替代的，主要是客服领域。抱歉，如果你以后要给 Comcast 发邮件的话，你大概永远不会再遇到一个真人了。不过说实话，五年前可能也已经差不多了。
所以我的预期是，会出现很多"类似于"今天已有工作的新型工作，但它们和今天的工作并不完全一样。我们还不知道这些变化具体是什么形态。目前讨论最多的是编程，也就是软件开发者。在商务会议上，我们可能正在认真讨论 Claude 的业务，然后谈到三分之二的时候，某个 CEO 会压低声音、用一种"咱俩私下说"的语气凑过来问："我女儿在 Stanford 读大二，她本来要学 CS，还应该学计算机吗？"说实话，我们不知道答案。
但我猜测，软件开发者这个职业仍然会存在，但他们不会再写那么多代码。软件开发者做的事情远不止坐在键盘前写代码——他们要和产品经理沟通，要和客户密切合作。我觉得这部分工作的比重会扩大，而那些更容易被 AI 完成的部分会缩小。但我的感觉是，这将极大地拓展什么是可能的。
所以我的预期是，会出现很多"类似于"今天已有工作的新型工作，但它们和今天的工作并不完全一样。我们还不知道这些变化具体是什么形态。目前讨论最多的是编程，也就是软件开发者。在商务会议上，我们可能正在认真讨论 Claude 的业务，然后谈到三分之二的时候，某个 CEO 会压低声音、用一种"咱俩私下说"的语气凑过来问："我女儿在 Stanford 读大二，她本来要学 CS，还应该学计算机吗？"说实话，我们不知道答案。
但我猜测，软件开发者这个职业仍然会存在，但他们不会再写那么多代码。软件开发者做的事情远不止坐在键盘前写代码——他们要和产品经理沟通，要和客户密切合作。我觉得这部分工作的比重会扩大，而那些更容易被 AI 完成的部分会缩小。但我的感觉是，这将极大地拓展什么是可能的。
**Daniela Amodei:** I think this is actually a very complicated question. My sense is that this probably will not surprise people here. AI is going to change what types of jobs are available and what types of jobs people do. So today, I think there are jobs that exist that did not exist five years ago because of AI. I also think there will be some jobs that will not exist in the future because of AI. But what we have seen so far today is that— according to our economic index, where we study how are people actually using artificial intelligence technology right now, it mostly looks like complementary skills, right? So you have artificial intelligence as an enabler of work. You don't have it as a replacer of work, except in a very vanishingly small number of cases, which is mostly customer service. Sorry if you have to email Comcast. It will never be a human again, probably. But I don't know that that was actually different five years ago.
So I think in reality, what I expect is there will be a number of types of work that will feel a lot like they rhyme with a job that exists today, but they're not necessarily the same as a job that exists today. And I think we just don't know the shape of what all of that is, right? Today, I think the thing that's most talked about is coding, right? Software developers. I always, you know, in business meetings, right, people will say to me, like, we're talking shop about Claude, and then like two-thirds of the way through the conversation, like a CEO will kind of conspiratorially lean across the table and say, you know, "My daughter is a sophomore at Stanford. Like, 'What should she study?' Right? Like, she was gonna be a CS major. Should she not major in computer science?" And I think the truth is we don't know.
But my guess is software developers will still exist, but like they won't write as much code. Right? Like a lot of what software developers do is much bigger than just hands on keyboard. They're talking to product managers. They're working closely with customers. And I think the percentage of that work is going to expand. And I think the sort of things that can be more easily done by AI, that will contract, but my sense is that is going to create a very different scope of what's possible.
So I think in reality, what I expect is there will be a number of types of work that will feel a lot like they rhyme with a job that exists today, but they're not necessarily the same as a job that exists today. And I think we just don't know the shape of what all of that is, right? Today, I think the thing that's most talked about is coding, right? Software developers. I always, you know, in business meetings, right, people will say to me, like, we're talking shop about Claude, and then like two-thirds of the way through the conversation, like a CEO will kind of conspiratorially lean across the table and say, you know, "My daughter is a sophomore at Stanford. Like, 'What should she study?' Right? Like, she was gonna be a CS major. Should she not major in computer science?" And I think the truth is we don't know.
But my guess is software developers will still exist, but like they won't write as much code. Right? Like a lot of what software developers do is much bigger than just hands on keyboard. They're talking to product managers. They're working closely with customers. And I think the percentage of that work is going to expand. And I think the sort of things that can be more easily done by AI, that will contract, but my sense is that is going to create a very different scope of what's possible.
**Interviewer：** 在教育、领导力和社会层面，需要发生什么变化，才能让人们感到准备好了、感到兴奋，而不仅仅是焦虑？
**Interviewer:** But what needs to happen in education, leadership, society, that people feel prepared and excited and not just anxious?
**Daniela Amodei：** 我觉得这里有几个层面。第一，我们需要从谦逊出发，承认我们不知道所有的答案，但同时要去做研究。在 Anthropic，我们一直力求在这些问题上做到尽可能地透明。我们一直说："我们没有所有的答案，我们需要去研究这些问题，这样才能告诉大家我们看到了什么趋势。"有时候人们会说："你们也太消极了吧，总在说未来可能会怎样怎样。"但我觉得更重要的是尽早开始这个对话，因为我们不想让人们措手不及。我们发布经济指数，就是要说"这是人们今天使用人工智能的方式"，因为我们希望人们了解这一切可能的走向。所以第一步是，我们必须尽可能地就现实是什么达成共识。
第二步是，我们需要在很多不同层面上发挥创造力和实验精神。AI 怎样才能真正成为一种让人们凝聚在一起的基础力量，而不仅仅是"我在工作中用一下这个工具"——工作当然很重要，但我觉得在某种程度上，我们需要重新思考工作、意义和社会生活之间的关系范式。我觉得这些东西在未来都会看起来非常不同，而我们需要开始练习和准备。
第三点，也是超出任何一家科技公司单独能做到的范围的：这将成为一个社会和政治议题。如果人们感觉自己的工作正在被 AI 取代，他们会在意的。人们现在就已经在意了。民调数据显示，人们说"我对人工智能将对我的未来、对我孩子的未来意味着什么感到焦虑。"不只是那个在会议桌对面悄悄问我的 CEO 会这么想。所以我认为，需要在政府的不同层面、与公民社会、与大学一起进行更广泛的讨论，去思考：这意味着什么？在人工智能能够完成人类今天所做的很多事情的世界里，我们想要建设一个什么样的世界？
第二步是，我们需要在很多不同层面上发挥创造力和实验精神。AI 怎样才能真正成为一种让人们凝聚在一起的基础力量，而不仅仅是"我在工作中用一下这个工具"——工作当然很重要，但我觉得在某种程度上，我们需要重新思考工作、意义和社会生活之间的关系范式。我觉得这些东西在未来都会看起来非常不同，而我们需要开始练习和准备。
第三点，也是超出任何一家科技公司单独能做到的范围的：这将成为一个社会和政治议题。如果人们感觉自己的工作正在被 AI 取代，他们会在意的。人们现在就已经在意了。民调数据显示，人们说"我对人工智能将对我的未来、对我孩子的未来意味着什么感到焦虑。"不只是那个在会议桌对面悄悄问我的 CEO 会这么想。所以我认为，需要在政府的不同层面、与公民社会、与大学一起进行更广泛的讨论，去思考：这意味着什么？在人工智能能够完成人类今天所做的很多事情的世界里，我们想要建设一个什么样的世界？
**Daniela Amodei:** So I think there's a few things here. I think the first one is we need to start and lead from a place of like humility and not knowing the answers, but doing the research. And I think at Anthropic, something that we've always aimed to do is be as kind of radically transparent about all of this as we possibly can be, right? We've always said, "Look, we don't have all the answers. We do need to study this so that we're able to tell people what we see coming." And I think, you know, sometimes fairly, people can say like, "Wow, you guys are just— this is like a lot of negativity," right? You're like, "Here's what we think might happen in the future." But I think it's more important that we start the conversation sooner because we don't want people to be caught off guard, right? We're publishing the economic index to say, "Here's how people are using artificial intelligence today," because we want people to have an understanding of, like, where do we think this is going? So I think that's step one, is we have to actually all agree on what reality is to the degree that's possible.
I think step two is we have to be creative and experimental at many different layers. So how do we have artificial intelligence really be something that is a grounding and a unifying force for people outside of just like, "Oh, I'm using it at my kind of job," right? Which is really important, but I think in some ways we need to sort of be rethinking the paradigm of this connection between work and meaning and, like, social life. Like, all of these things I think are going to look very different in the future, and we need to practice.
And then I think the third one, which is really outside of the realm of what a technology company can do alone, is this is going to become a social and political issue. People are going to care if it feels like their jobs are being displaced by AI. People already care about this, right? It comes up in polling. People are like, "I have anxiety about what artificial intelligence is going to mean for my future, for my kids' future," right? That CEO that's leaning across the table from me, it's not just them. And so I think there is a broader discussion that needs to happen at many different levels of government with civil society, with universities, by the way, to say like, "What does this mean? What is the type of world that we wanna be able to build where artificial intelligence is capable of doing many of the things that humans do today?"
I think step two is we have to be creative and experimental at many different layers. So how do we have artificial intelligence really be something that is a grounding and a unifying force for people outside of just like, "Oh, I'm using it at my kind of job," right? Which is really important, but I think in some ways we need to sort of be rethinking the paradigm of this connection between work and meaning and, like, social life. Like, all of these things I think are going to look very different in the future, and we need to practice.
And then I think the third one, which is really outside of the realm of what a technology company can do alone, is this is going to become a social and political issue. People are going to care if it feels like their jobs are being displaced by AI. People already care about this, right? It comes up in polling. People are like, "I have anxiety about what artificial intelligence is going to mean for my future, for my kids' future," right? That CEO that's leaning across the table from me, it's not just them. And so I think there is a broader discussion that needs to happen at many different levels of government with civil society, with universities, by the way, to say like, "What does this mean? What is the type of world that we wanna be able to build where artificial intelligence is capable of doing many of the things that humans do today?"
**Interviewer：** 归根结底，这都是关于普及率的。在 Stanford，我们每天都沉浸在 AI 里，但 Stanford 和硅谷并不是整个世界。目前是什么在阻碍 AI 在这个圈子之外的普及？
**Interviewer:** At the core, it is all about adoption. Here at Stanford, we live and breathe AI, but Stanford and Silicon Valley are not the whole world. What currently hinders AI adoption outside of this bubble?
**Daniela Amodei：** 这个问题提得非常好。确实很有意思，因为对于我们 Anthropic 来说，我相信在 Stanford 也是这样，AI 是唯一人们想聊的话题。当然，别人跟我聊的也只有 AI，但那可能是我的问题。不过你说的完全对——即使在美国的其他地方，人们对 AI 整体上还不太适应，甚至不一定知道怎么高效地使用。你会看到一些令人印象深刻的数据，说有多少人在用 AI 工具，但这里面有人口统计学的因素。使用者通常是受过大学教育的人，不完全是，但男性多于女性，还涉及种族构成和财富水平。放眼全球来看，分布更是极不均匀的。
有意思的是，把这些数据和我们收集的其他数据放在一起看——发展中国家的人对 AI 的态度比高收入国家的人乐观得多。全球南方几乎一致认为："这对我们来说是一个巨大的机会。"也许这就是一个能让事情变得更公平的均衡力量。但在美国、欧洲和亚洲部分地区，人们的焦虑感更强。他们觉得："我喜欢现在的生活。我不想让 AI 来打乱这一切，这听起来对我来说不太好。"面对这些信息该怎么做？说实话，我也不知道。
但我觉得有意思的是，围绕这项技术，在获取和普及方面存在着不同的问题。而且我们实际上仍然处于非常早期的阶段。这一点在硅谷的泡泡里很容易被忽略。我们觉得"大家都在用了"——所有的软件工程师都在用 Claude Code、用 Codex。但在全世界范围内，这远不是大多数开发者的状态。所以我觉得，这场赛跑才刚刚鸣枪起跑。在如何积极地塑造这项技术的使用和开发方式、在获取途径上、以及在技术最终会融入什么样的价值观上，仍然有很大的空间。
有意思的是，把这些数据和我们收集的其他数据放在一起看——发展中国家的人对 AI 的态度比高收入国家的人乐观得多。全球南方几乎一致认为："这对我们来说是一个巨大的机会。"也许这就是一个能让事情变得更公平的均衡力量。但在美国、欧洲和亚洲部分地区，人们的焦虑感更强。他们觉得："我喜欢现在的生活。我不想让 AI 来打乱这一切，这听起来对我来说不太好。"面对这些信息该怎么做？说实话，我也不知道。
但我觉得有意思的是，围绕这项技术，在获取和普及方面存在着不同的问题。而且我们实际上仍然处于非常早期的阶段。这一点在硅谷的泡泡里很容易被忽略。我们觉得"大家都在用了"——所有的软件工程师都在用 Claude Code、用 Codex。但在全世界范围内，这远不是大多数开发者的状态。所以我觉得，这场赛跑才刚刚鸣枪起跑。在如何积极地塑造这项技术的使用和开发方式、在获取途径上、以及在技术最终会融入什么样的价值观上，仍然有很大的空间。
**Daniela Amodei:** You know, I think this is a really, really great call-out. It's so interesting because it feels like, you know, certainly to us at Anthropic, and I'm sure at Stanford too, it's like the only thing people want to talk about is AI. Certainly, the only thing people want to talk to me about is AI, but that's probably a me problem. But I think that feeling, you're totally right, that even in other parts of America, it's not something that people are comfortable with yet. On the whole, and it's not even something that people necessarily know how to use with very high fluency, right? So you read these really astounding, super impressive numbers about how many people are using AI tools. But there's a demographic component to that, right? So it's generally people who are college-educated, not uniquely. It's more men than women. There's racial demographics involved. There's wealth demographics involved, and if you look around the world, it's really not equally distributed.
And so I think what's interesting is that you pair that with some other data that we've collected, which is that people in developing countries are much more optimistic about AI than people in higher income countries. So the Global South is almost universally like, "Wow, this is a huge opportunity for us." Right? This is the moment where perhaps we could have an equalizing force that will make things more fair. But I think in the US and in Europe and in parts of Asia, people have a lot more anxiety. They're like, "I like things the way they are. I don't want AI to come in and disrupt that. That doesn't sound as good to me." What do we do with this information? I have no idea.
But I think it's really interesting to say, like, there are different questions of access and adoption around the technology. And I think we are actually still very early in the game. And that's the thing that I think can be missed in Silicon Valley, in our bubble, that we're like, "It's already," everybody who's a software engineer is like, "I'm using Claude Code, I'm using Codex." That is not the vast majority of developers in the world at large. And so I think the race is still, like we're at the— like the gun just went off to start the race. And I think there's a lot of opportunity to still positively shape how this technology is going to be used and developed, what access looks like, and just what the values that are baked into it are going to ultimately be.
And so I think what's interesting is that you pair that with some other data that we've collected, which is that people in developing countries are much more optimistic about AI than people in higher income countries. So the Global South is almost universally like, "Wow, this is a huge opportunity for us." Right? This is the moment where perhaps we could have an equalizing force that will make things more fair. But I think in the US and in Europe and in parts of Asia, people have a lot more anxiety. They're like, "I like things the way they are. I don't want AI to come in and disrupt that. That doesn't sound as good to me." What do we do with this information? I have no idea.
But I think it's really interesting to say, like, there are different questions of access and adoption around the technology. And I think we are actually still very early in the game. And that's the thing that I think can be missed in Silicon Valley, in our bubble, that we're like, "It's already," everybody who's a software engineer is like, "I'm using Claude Code, I'm using Codex." That is not the vast majority of developers in the world at large. And so I think the race is still, like we're at the— like the gun just went off to start the race. And I think there's a lot of opportunity to still positively shape how this technology is going to be used and developed, what access looks like, and just what the values that are baked into it are going to ultimately be.
**Interviewer：** 那让我们展望一下未来，AI 被更广泛地普及了。如果我们开始把太多事情委托给 AI，我们可能会失去什么？
**Interviewer:** So let's forward to the future where AI is far more widely adopted. What are the things we risk losing if we start delegating too much to AI?
**Daniela Amodei：** 我们在 Anthropic 做了一项非常大规模的定性调查，据我们所知是有史以来最大的定性研究。我们与81,000人谈了他们对人工智能的使用。其中一些是 Claude 用户，一些是其他 AI 工具的用户。有意思的是，人们对 AI 的感受非常多样。根据你在哪里、你做什么工作，看法差异很大。有些人说："AI 让我做到了以前觉得自己做不到的事情。"就拿我自己来说——我以前觉得我不可能搭一个网站，但现在用 Claude，我发现"天哪，这也太简单了"，我就点几个按钮，Claude 就帮我搭好了一个网站。如果我自己来做的话，大概要花一年，而且效果肯定不怎么样。
但也有一些人表达了另一种感受——目前还没有一个专门的词来形容它，但我觉得未来可能会有，也许在某种语言里会诞生这样一个词——就是"我不再动脑了，因为我不需要了。"这和刷手机的那种感觉不一样。它更像是：我本来可以去思考这个问题的，我本来可以自己想清楚的，但不去想实在太容易了，我就直接相信了 AI 给我的东西。我认为这实际上是人们对 AI 焦虑的一个深层来源。人类天生有学习的欲望，有好奇心，想要拓展自己的认知边界。AI 在某种程度上能够促进这种学习，但如果使用不当，也可能抑制它。我自己有时候也会这样——"我可以自己去查一下弄清楚的，但我就直接问 AI 了，然后盲目相信它给的答案是对的。"顺便说一下，它并不总是对的。有时候 Claude 会出错。说这话有点大逆不道，但这是事实。
我觉得这里的焦虑在于：我们怎么设置一些护栏，让人不太容易无脑依赖 AI——不是说完全不能这样做，但你得刻意去做才行。我们和大学合作的一些工作可能是一个有意思的缩影。我们有一个叫"学习模式"（learning mode）的概念，在座的可能有人用过，不知道 GSB 这边有没有接入。一种方式是你把作业丢进 ChatGPT——我故意用这个例子——然后它直接给你答案。这有一个专门的词来形容，叫作弊。另一种方式是你使用 Claude 的学习模式。你说"我卡住了，我在写这篇论文，但格式上有些地方总感觉不对"，然后 Claude 就像一个耐心的辅导老师，几乎像是一个了解你、理解你最想学什么、知道这门课对你为什么重要的私人教授。它会说："让我帮你走出困境。你要不要回去一起重读这个部分？我们能不能讨论一下这个问题？"我觉得这种版本的 AI 能让你变得更聪明，能拓展你认为自己能学的东西的范围。而另一种就是关掉大脑。我希望我们这个行业会选择前者而不是后者。
但也有一些人表达了另一种感受——目前还没有一个专门的词来形容它，但我觉得未来可能会有，也许在某种语言里会诞生这样一个词——就是"我不再动脑了，因为我不需要了。"这和刷手机的那种感觉不一样。它更像是：我本来可以去思考这个问题的，我本来可以自己想清楚的，但不去想实在太容易了，我就直接相信了 AI 给我的东西。我认为这实际上是人们对 AI 焦虑的一个深层来源。人类天生有学习的欲望，有好奇心，想要拓展自己的认知边界。AI 在某种程度上能够促进这种学习，但如果使用不当，也可能抑制它。我自己有时候也会这样——"我可以自己去查一下弄清楚的，但我就直接问 AI 了，然后盲目相信它给的答案是对的。"顺便说一下，它并不总是对的。有时候 Claude 会出错。说这话有点大逆不道，但这是事实。
我觉得这里的焦虑在于：我们怎么设置一些护栏，让人不太容易无脑依赖 AI——不是说完全不能这样做，但你得刻意去做才行。我们和大学合作的一些工作可能是一个有意思的缩影。我们有一个叫"学习模式"（learning mode）的概念，在座的可能有人用过，不知道 GSB 这边有没有接入。一种方式是你把作业丢进 ChatGPT——我故意用这个例子——然后它直接给你答案。这有一个专门的词来形容，叫作弊。另一种方式是你使用 Claude 的学习模式。你说"我卡住了，我在写这篇论文，但格式上有些地方总感觉不对"，然后 Claude 就像一个耐心的辅导老师，几乎像是一个了解你、理解你最想学什么、知道这门课对你为什么重要的私人教授。它会说："让我帮你走出困境。你要不要回去一起重读这个部分？我们能不能讨论一下这个问题？"我觉得这种版本的 AI 能让你变得更聪明，能拓展你认为自己能学的东西的范围。而另一种就是关掉大脑。我希望我们这个行业会选择前者而不是后者。
**Daniela Amodei:** Yeah. We did this very large qualitative survey at Anthropic. I think it's the largest qualitative study ever done that we know about. We talked to 81,000 people about their use of artificial intelligence. And so some of them were Claude users, some of them were users of other AI tools. And what was interesting is, you know, people have a lot of different feelings about AI, right? And again, you know, cuts differently depending on where you are and what you do. But some people are like, "I've never been able..." You know, it enabled me to do things that I never thought I could do," right? I think— for myself as an example, like I didn't think I could build a website, right? And now using Claude, I'm like, "Oh man, that's so easy." Like I just click a couple buttons and Claude builds a website for me. What? That would've taken me probably a year if I tried to do it by myself, and it would not have been a very good website.
There's some people though who express a feeling of— there's not like a specific term for it, but I think there might be, or there might be in another language one day. Like I don't engage my brain because I don't have to. Yeah. So it's not the same feeling as like scrolling on your phone, but it's like I could have reached for this idea, I could have thought through it, but it was so much easier to not do it and to just trust what the AI tool was giving me. And I think this is the source, I actually believe, of a lot of the anxiety around AI. As humans like to, I think, have an inherent desire to learn, to be curious, to want to expand the aperture of things that they know about. And AI, in some ways, enables that, but if used incorrectly, can sort of disable that, right? It's like I've done this sometimes, right? And I was like, "Oh, I could look this up and figure it out myself, but I'll just ask an AI tool, and then I'll blindly trust that what it says is correct." It's not always correct, by the way. Sometimes Claude is wrong. Heretical to say, but a fact.
And I think the anxiety there is around how do we actually set some guardrails in place so that it's just, it's not impossible to do that, but you actually have to really be trying to do it, right? Like, I think some of the work that we do with universities is maybe an interesting microcosm for this. We have this concept of learning mode. Maybe some of you even use it. I don't know if we're at the GSB yet. But faculty and professors and students— like, one version of this is like you put your homework in ChatGPT, I'm gonna use that one instead, and you're like, "Ha ha, it just gave me the answer." There's a word for that. It's called cheating, right? And you're like, "That was great." There's another version— yeah, where you use Claude in learning mode, and you're like, "I don't... I'm... I'm stuck," right? "I'm trying to write this essay and I, like, there's something about the format that doesn't feel right to me." And Claude is this sort of— patient tutor. It's almost like you have an individualized professor who, like, knows you and understands, like, what you most wanna learn and why this class is important to you. It's like, "Let me help get you unstuck," right? "Do you wanna go back and read this section together? Could we talk through this?" I think that's the version where, like, these tools can make you smarter. They can make you expand the set of things that you think you can learn. And then I think there's the version that's just turn your brain off, and I'm hoping that as an industry we're going to choose to do the second versus the first.
There's some people though who express a feeling of— there's not like a specific term for it, but I think there might be, or there might be in another language one day. Like I don't engage my brain because I don't have to. Yeah. So it's not the same feeling as like scrolling on your phone, but it's like I could have reached for this idea, I could have thought through it, but it was so much easier to not do it and to just trust what the AI tool was giving me. And I think this is the source, I actually believe, of a lot of the anxiety around AI. As humans like to, I think, have an inherent desire to learn, to be curious, to want to expand the aperture of things that they know about. And AI, in some ways, enables that, but if used incorrectly, can sort of disable that, right? It's like I've done this sometimes, right? And I was like, "Oh, I could look this up and figure it out myself, but I'll just ask an AI tool, and then I'll blindly trust that what it says is correct." It's not always correct, by the way. Sometimes Claude is wrong. Heretical to say, but a fact.
And I think the anxiety there is around how do we actually set some guardrails in place so that it's just, it's not impossible to do that, but you actually have to really be trying to do it, right? Like, I think some of the work that we do with universities is maybe an interesting microcosm for this. We have this concept of learning mode. Maybe some of you even use it. I don't know if we're at the GSB yet. But faculty and professors and students— like, one version of this is like you put your homework in ChatGPT, I'm gonna use that one instead, and you're like, "Ha ha, it just gave me the answer." There's a word for that. It's called cheating, right? And you're like, "That was great." There's another version— yeah, where you use Claude in learning mode, and you're like, "I don't... I'm... I'm stuck," right? "I'm trying to write this essay and I, like, there's something about the format that doesn't feel right to me." And Claude is this sort of— patient tutor. It's almost like you have an individualized professor who, like, knows you and understands, like, what you most wanna learn and why this class is important to you. It's like, "Let me help get you unstuck," right? "Do you wanna go back and read this section together? Could we talk through this?" I think that's the version where, like, these tools can make you smarter. They can make you expand the set of things that you think you can learn. And then I think there's the version that's just turn your brain off, and I'm hoping that as an industry we're going to choose to do the second versus the first.
**Interviewer：** 如果必须排优先级的话，在 AI 驱动的世界里，哪些人类技能最有可能变得更加重要？
**Interviewer:** So if you had to prioritize, what human skills are most likely gonna be more important in an AI-driven world?
**Daniela Amodei：** 我有自己的看法。我觉得，正如我们刚才讨论的，很多特定的、任务导向的事情——比如我是一个金融分析师、我是一个开发者、我是一个文字编辑——这些工作会发生很大的变化。其中很多工作我相信 AI 工具将能够完成。但归根结底有一个非常真实的现象：人类喜欢和其他人在一起。我们喜欢相互陪伴，喜欢彼此学习，喜欢创造，喜欢花时间去理解对方。我们是社会性动物。我觉得在一个 AI 能够完成更多日常生产性工作的世界里，这些技能将变得重要得多，也会更加被珍视。因为归根结底，如果你在一个工作环境中，你可以让 Claude 帮你写一堆代码，但你还是会选择去找那个能够向你解释"为什么某个东西坏了"或"我们为什么要用这种方式来构建一个工具"的开发者。
把这个推广到科技行业之外，我经常举的例子是医学。今天我们雇佣医生，是因为他们擅长诊断。你说"我不舒服，能告诉我怎么了吗？"然后你基本上是在付钱请这个医生说"你可能有这几种问题，最有可能的是这个，我来开一些检查。"AI 会变得非常擅长做这件事。但 AI 做不到的是真正地看着你、检查你的身体，以及理解你的感受、帮助你感觉好一些。有相当多的医学文献表明，那些与医生关系良好、喜欢自己医生的病人，临床结果比那些不喜欢自己医生的病人要好。这很难完全解释，但可能的原因是——那个医生会更努力地去理解你到底怎么了，也许会开一些意想不到的检查。我觉得这些技能——所谓的"床边态度"（bedside manner）——在一个你不再需要把它塞进七项衡量医生是否合格的标准里的世界中，重要性会提升五倍。
把这个推广到科技行业之外，我经常举的例子是医学。今天我们雇佣医生，是因为他们擅长诊断。你说"我不舒服，能告诉我怎么了吗？"然后你基本上是在付钱请这个医生说"你可能有这几种问题，最有可能的是这个，我来开一些检查。"AI 会变得非常擅长做这件事。但 AI 做不到的是真正地看着你、检查你的身体，以及理解你的感受、帮助你感觉好一些。有相当多的医学文献表明，那些与医生关系良好、喜欢自己医生的病人，临床结果比那些不喜欢自己医生的病人要好。这很难完全解释，但可能的原因是——那个医生会更努力地去理解你到底怎么了，也许会开一些意想不到的检查。我觉得这些技能——所谓的"床边态度"（bedside manner）——在一个你不再需要把它塞进七项衡量医生是否合格的标准里的世界中，重要性会提升五倍。
**Daniela Amodei:** So I have my own views on this, which is I think, like we talked about, a lot of specific, like task-oriented things, right? Like I'm a financial analyst, or I'm a developer, or I'm a copy editor. Those jobs are going to change a lot. And a lot of that work I think will be able to be done by AI tools. But I think ultimately there's this very real phenomenon, which is that humans like to be with other humans. We like to spend time together with each other. We like to learn from each other. We like to be creative. We like to spend time understanding the other person, and we're social creatures, right? And I think I'm imagining that in a world where AI is able to do a lot more of the productive day-to-day work that we do, those skills are going to become a lot more important and much more prized. Because ultimately, if you're in a work environment and you're like, "Well, I could just ask Claude to like write a bunch of code for me," you are going to choose to talk to the developer that's going to explain to you like why something broke. Or why we chose to build, you know, a tool the way that we did.
And I think sort of expanding this outside of the realm of just the technology industry, the example I often use is in medicine, right? Today. We hire doctors who are really good diagnosticians. We're like, "Hey, can you tell me what's wrong with me? I don't feel well." And you're basically paying this doctor to say, "Here's a set of things that could be wrong with you. This is the one that's the most likely. Let me run some tests." Guess what? AI is going to get really good at doing that. But the thing that an AI tool can't do is actually look at you and examine you, and also help understand how you're feeling and help you feel better, right? There's a reasonable body of medical literature that indicates that people have a good relationship with their doctor, they just like their doctor, have better clinical outcomes than people who do not like their doctor. That's really hard to explain, but like, what's probably going on? Probably the doctor, like— tries a little bit harder to understand what's wrong with you. Maybe they run a set of tests that were unexpected. And I think those skills, right, that bedside manner is going to be like five times more important in a world where you're not trying to cram that into one of seven things that you're looking for to make a doctor qualified to treat you.
And I think sort of expanding this outside of the realm of just the technology industry, the example I often use is in medicine, right? Today. We hire doctors who are really good diagnosticians. We're like, "Hey, can you tell me what's wrong with me? I don't feel well." And you're basically paying this doctor to say, "Here's a set of things that could be wrong with you. This is the one that's the most likely. Let me run some tests." Guess what? AI is going to get really good at doing that. But the thing that an AI tool can't do is actually look at you and examine you, and also help understand how you're feeling and help you feel better, right? There's a reasonable body of medical literature that indicates that people have a good relationship with their doctor, they just like their doctor, have better clinical outcomes than people who do not like their doctor. That's really hard to explain, but like, what's probably going on? Probably the doctor, like— tries a little bit harder to understand what's wrong with you. Maybe they run a set of tests that were unexpected. And I think those skills, right, that bedside manner is going to be like five times more important in a world where you're not trying to cram that into one of seven things that you're looking for to make a doctor qualified to treat you.
**Interviewer：** 当你思考未来的时候，你个人最期待哪些 AI 应用场景？
**Interviewer:** And when you are thinking about the future, what AI use cases are you most personally excited about?
**Daniela Amodei：** 对我个人来说——我的职业就是做管理者，我大部分时间都在跟人打交道。有一种现象，也许有一天会有人给它起个名字：每个人都觉得 AI 不会取代自己的工作，因为自己太特别了。我对此也深感内疚。我以前觉得"人们喜欢人。他们当然想向我汇报，不会想向 Claude 汇报的。"但实际上，Claude 作为管理教练、帮助你成为更好的领导者方面，真的非常强大。
我是这样用 Claude 的——我们在 Anthropic 要写绩效评估，跟我合作的很多人已经向我汇报三四年了。一般来说，你还是同一个人，他们还是同一个人，你给他们反馈，但过去六个月真正改变了多少呢？Claude 在帮我发现关于某个人的模式方面特别有用。如果你回顾三到四年的合作历程，Claude 会指出"你们在某个话题上已经绕了三四年了"——也许这个人需要一些额外的辅导，或者也许他们需要你之外的某个人来帮忙。这种东西在日常工作中很容易被忽略，因为你就在其中、身在局中。
反过来，Claude 也很擅长给你反馈。我会把我的下属对我的向上反馈都上传上去，然后 Claude 有时候会非常客气地说："听起来你在这件事上过去一年都没有进步。也许你应该额外找一些辅导，Daniela。"我觉得 Claude 在教练和帮助人们成为最好版本的自己方面的能力，无论是在职场还是在个人生活中，如果做得谨慎，都可以非常强大。
另外一个——我有两个小孩，一个快五岁了，一个快一岁了。说真的，Claude 做过的最棒的一件事就是帮我度过了如厕训练。那段经历真的不好受。Claude 让整个过程稍微好了那么一点——它很有同理心，给的建议非常具体可操作，还有一些图表，细节我就不展开了。但真的非常有用。我觉得 Claude 帮助那些手忙脚乱的父母这一点，将来会非常有价值。因为网上有太多不靠谱的信息了。每次你搜"我的孩子是不是有什么问题"，答案永远是"有"。Claude 则要理性得多，而且可以互动交流，我觉得这一点非常有帮助。
我是这样用 Claude 的——我们在 Anthropic 要写绩效评估，跟我合作的很多人已经向我汇报三四年了。一般来说，你还是同一个人，他们还是同一个人，你给他们反馈，但过去六个月真正改变了多少呢？Claude 在帮我发现关于某个人的模式方面特别有用。如果你回顾三到四年的合作历程，Claude 会指出"你们在某个话题上已经绕了三四年了"——也许这个人需要一些额外的辅导，或者也许他们需要你之外的某个人来帮忙。这种东西在日常工作中很容易被忽略，因为你就在其中、身在局中。
反过来，Claude 也很擅长给你反馈。我会把我的下属对我的向上反馈都上传上去，然后 Claude 有时候会非常客气地说："听起来你在这件事上过去一年都没有进步。也许你应该额外找一些辅导，Daniela。"我觉得 Claude 在教练和帮助人们成为最好版本的自己方面的能力，无论是在职场还是在个人生活中，如果做得谨慎，都可以非常强大。
另外一个——我有两个小孩，一个快五岁了，一个快一岁了。说真的，Claude 做过的最棒的一件事就是帮我度过了如厕训练。那段经历真的不好受。Claude 让整个过程稍微好了那么一点——它很有同理心，给的建议非常具体可操作，还有一些图表，细节我就不展开了。但真的非常有用。我觉得 Claude 帮助那些手忙脚乱的父母这一点，将来会非常有价值。因为网上有太多不靠谱的信息了。每次你搜"我的孩子是不是有什么问题"，答案永远是"有"。Claude 则要理性得多，而且可以互动交流，我觉得这一点非常有帮助。
**Daniela Amodei:** Oh, man. I think for me personally, I mean, I'm a career manager, so I spend most of my time with people. And we have this sort of, I don't know, again, it's maybe one day there's gonna be a word for it, but there's this phenomenon where everybody thinks that AI is not gonna come for their job because they're so special. And I am extremely guilty of this. I was like, "People love people. People are gonna wanna report to me. They're not gonna wanna report to Claude, obviously." Um, but I actually think Claude is incredibly powerful as a management coach and as a person to help you be a better leader.
So I use Claude. We write performance reviews at Anthropic, and I've uploaded, you know, a lot of the people that I work with have reported to me for, like, you know, three or four years, so— in general, it's like you're the same person, they're the same person. You give them feedback, but like how much has really changed just like in the past six months? But I think Claude has been really powerful in helping me to spot patterns about somebody. So if you, you know, more data is better, but if you look back over the course of three to four years of time of working with somebody and you're like, "Wow, you know, you guys have been circling around like this topical issue for the past three to four years," like maybe they need some additional coaching. Um, or maybe they need somebody sort of outside of you. It's just the type of thing that I think tends to get missed because you're just in it day to day.
And in the opposite direction, Claude is great at giving you feedback. So like I upload all of my reports' upward feedback for me, and sometimes Claude will kind of— very kindly be like, "It sounds like you haven't improved on this in the past year. Like, maybe you should get some extra coaching, Daniela." But I think those, I think Claude's ability to kind of coach and help people be the best versions of themselves. I think there's a version of that that makes sense in the workplace, in people's personal lives, that I think could be done kind of, you know, quite carefully, but that I think could be really powerful.
And then the second is I have two little kids, so I have an almost five-year-old and almost one-year-old. And I have to tell you, number one best thing Claude has ever done is help me through potty training. That was not a fun experience. And Claude made it just like a little bit... it was empathetic, like very actionable. There were some diagrams. I don't need to tell you guys. But it was really, really useful, and I think Claude's ability to help, in particular, overwhelmed parents is going to be really powerful because there's so much bad information. It's like every time you Google, "Is something wrong with your kid?" The answer is yes. And I think Claude is a lot more measured and can be interactive in a way that I think is really, really helpful.
So I use Claude. We write performance reviews at Anthropic, and I've uploaded, you know, a lot of the people that I work with have reported to me for, like, you know, three or four years, so— in general, it's like you're the same person, they're the same person. You give them feedback, but like how much has really changed just like in the past six months? But I think Claude has been really powerful in helping me to spot patterns about somebody. So if you, you know, more data is better, but if you look back over the course of three to four years of time of working with somebody and you're like, "Wow, you know, you guys have been circling around like this topical issue for the past three to four years," like maybe they need some additional coaching. Um, or maybe they need somebody sort of outside of you. It's just the type of thing that I think tends to get missed because you're just in it day to day.
And in the opposite direction, Claude is great at giving you feedback. So like I upload all of my reports' upward feedback for me, and sometimes Claude will kind of— very kindly be like, "It sounds like you haven't improved on this in the past year. Like, maybe you should get some extra coaching, Daniela." But I think those, I think Claude's ability to kind of coach and help people be the best versions of themselves. I think there's a version of that that makes sense in the workplace, in people's personal lives, that I think could be done kind of, you know, quite carefully, but that I think could be really powerful.
And then the second is I have two little kids, so I have an almost five-year-old and almost one-year-old. And I have to tell you, number one best thing Claude has ever done is help me through potty training. That was not a fun experience. And Claude made it just like a little bit... it was empathetic, like very actionable. There were some diagrams. I don't need to tell you guys. But it was really, really useful, and I think Claude's ability to help, in particular, overwhelmed parents is going to be really powerful because there's so much bad information. It's like every time you Google, "Is something wrong with your kid?" The answer is yes. And I think Claude is a lot more measured and can be interactive in a way that I think is really, really helpful.
**Interviewer：** Daniela，在我们把问题交给学生之前，当你想到下一代 AI 领导者和建设者——他们今天就在这个房间里——你最希望他们从你的经历中带走什么？
**Interviewer:** Daniela, before we turn to student questions, when you think about the next generation of AI leaders and builders who are with us here today in this room, what's the one thing you hope they take from your journey?
**Daniela Amodei：** 我能说两点吗？
**Daniela Amodei:** I would say, can I do two?
**Interviewer：** 请讲。
**Interviewer:** Go ahead.
**Daniela Amodei：** 好。第一点，听起来真的很老套，我都差点不想说了。但我真心认为，追随你真正在意的东西、你真正热爱的东西，是你能做的最重要的事情。好想法多得是，但如果你没有那种"这件事必须存在于这个世界上"的燃烧般的感觉，没有那种"我会不惜一切代价去做成它"的冲劲——在不好玩的时候、在很痛苦的时候，这种感觉才是支撑你的东西。你必须能够对自己说"我记得这件事为什么对我重要、为什么它是重要的"，无论是对你个人，还是因为你想看到某种改变在世界上发生。在 Anthropic，肯定有过那样的时刻——这不是最有趣的部分，有些阶段很艰难。能够把它联系回你当初为什么决定做这件事、为什么它对你重要，我觉得这极其关键。
第二点，我觉得对这一代人来说尤其适用——过去五到十年出现了一个观念：做商业和做好事不必是矛盾的。这是一个非常新的理念，也是非常特别的。我对这一代创始人和创造者印象深刻，他们在用这种方式思考问题——创新和社会影响的结合。Stanford 在这方面一直做得很出色。但我觉得这在更广的范围内是一个全新的概念，而且现在对这种理念的接受度更高了。过去可能会有一种感觉——只有那些冷酷、无情的人才能做成生意。我不觉得这是对的。而且我越来越觉得，做好事的愿望和做出好成绩之间有很强的正相关。
第二点，我觉得对这一代人来说尤其适用——过去五到十年出现了一个观念：做商业和做好事不必是矛盾的。这是一个非常新的理念，也是非常特别的。我对这一代创始人和创造者印象深刻，他们在用这种方式思考问题——创新和社会影响的结合。Stanford 在这方面一直做得很出色。但我觉得这在更广的范围内是一个全新的概念，而且现在对这种理念的接受度更高了。过去可能会有一种感觉——只有那些冷酷、无情的人才能做成生意。我不觉得这是对的。而且我越来越觉得，做好事的愿望和做出好成绩之间有很强的正相关。
**Daniela Amodei:** Okay. I would say the first is, it sounds so trite, and so I almost, like, I almost won't say it. But I truly think following something that you really care about, something that you are passionate about, is the most important thing you can do. There are so many great ideas, and if you don't feel the sort of like burning feeling of "This is a thing that needs to exist in the world," and I will just run through walls to be able to do it. It's just it matters for the times when it's not fun. And when it sucks. Like, you just have to be able to say like, "I remember why this matters to me. I remember why it's important." Whether it's to me personally or it's because of a type of change that I wanna see happen in the world. There will be times, there certainly have been at Anthropic, times where we were like, "This is not the most fun part," right? There are parts that are tough, and so being able to relate it back to why you decided to do this in the first place and why it matters to you, I think is so important.
And then the second, I would say, is I think for this generation in particular, and I think really in the past, you know, five to 10 years, this concept that like— being in business doesn't have to be in tension with doing good. I think that is a very new idea. And I think it is really special. And I have been so impressed at the sort of generation of founders and just like creators who are thinking in that way, right? Like, there's this kind of marriage of... innovation and social impact. I think Stanford has always been exceptional at this. But I think that is a very new concept. And I think there's more appetite for it today, right? I think there can sort of be this feeling of only the kind of mean, sucky people can build a business. I just don't think that's true. And I think increasingly I feel that the desire to do good is a strong correlate with actually doing well.
And then the second, I would say, is I think for this generation in particular, and I think really in the past, you know, five to 10 years, this concept that like— being in business doesn't have to be in tension with doing good. I think that is a very new idea. And I think it is really special. And I have been so impressed at the sort of generation of founders and just like creators who are thinking in that way, right? Like, there's this kind of marriage of... innovation and social impact. I think Stanford has always been exceptional at this. But I think that is a very new concept. And I think there's more appetite for it today, right? I think there can sort of be this feeling of only the kind of mean, sucky people can build a business. I just don't think that's true. And I think increasingly I feel that the desire to do good is a strong correlate with actually doing well.
**Interviewer：** 把问题交给学生们。
**Interviewer:** Turn to the students.
**Brandon：** 嗨，Daniela。我是 Brandon，这里的 MBA 二年级学生。谢谢你来。关于我们是否处于 AI 泡沫，人们通常说这话时指的是三件不同的事：公司估值、企业在基础设施上的资本支出、以及 AI 进步的速度是否可持续。这三个里面你最担心哪一个，又觉得人们在哪一个上面错得最离谱？
**Brandon:** Hi, Daniela. My name is Brandon. I'm a second-year MBA student here. Thank you for joining us. There's a debate about whether we're in an AI bubble, and people usually mean three different things when they say that. Company valuations, how much companies are spending on infrastructure, or whether the pace of AI progress is actually sustainable. Which of those three are you most worried about, and which do you think people are most wrong about?
**Daniela Amodei：** 好问题。确实需要区分不同类型的"泡沫"——气泡和玻璃泡泡是很不一样的，我理解你的意思。我觉得最——不确定用"最担心"是否准确——但我认为对这个行业来说最合理的担忧是：这是一个高资本支出的行业，这本身就带来一些风险。这些你们可能都已经知道了，但训练这些模型非常昂贵，需要大量算力（compute），而算力的供给是稀缺的。当需求很大、供给很少——我不是经济学教授，但我觉得这意味着价格会上涨。作为这些公司的命脉，算力必须提前很久购买，所以你本质上是在押注未来。你在想"我们认为我们在某个时间点需要这么多算力"。这是一笔非常大的支出，在任何一家这样的公司工作都会让人有些提心吊胆。
如果有人告诉你他们不紧张——也许 Google 除外，因为它是上市公司、钱多到花不完。但我觉得无论是 Anthropic 还是 OpenAI，我们都是在做一个经过计算的赌注——我们赌自己最终能在一段时间内把这笔钱赚回来。我们当然对此非常乐观。两家公司的收入增速令人难以置信。我们经常听到风投们说"从来没有见过这样的事情"——很难想象一家公司在这么短的时间内达到这样的收入规模。
但如果情况发生变化，那就会有问题。两家公司都已经为未来购买了大量的算力，非常昂贵。所以我觉得这可能是那个最合理的担忧。担心这件事并不疯狂。我们当然认为自己处于一个非常好的位置，整个行业也是。但这随时可能改变。重要的是要记住，这归根结底是一个赌注——是整个行业在说"这将带来很大的回报"。但我们完全有可能是错的。
如果有人告诉你他们不紧张——也许 Google 除外，因为它是上市公司、钱多到花不完。但我觉得无论是 Anthropic 还是 OpenAI，我们都是在做一个经过计算的赌注——我们赌自己最终能在一段时间内把这笔钱赚回来。我们当然对此非常乐观。两家公司的收入增速令人难以置信。我们经常听到风投们说"从来没有见过这样的事情"——很难想象一家公司在这么短的时间内达到这样的收入规模。
但如果情况发生变化，那就会有问题。两家公司都已经为未来购买了大量的算力，非常昂贵。所以我觉得这可能是那个最合理的担忧。担心这件事并不疯狂。我们当然认为自己处于一个非常好的位置，整个行业也是。但这随时可能改变。重要的是要记住，这归根结底是一个赌注——是整个行业在说"这将带来很大的回报"。但我们完全有可能是错的。
**Daniela Amodei:** That's a great question. Yeah, we need different... we need air bubbles and glass bubble. They're very different. I see what you're saying. I think the one that probably... I don't know if I would say I'm the most worried about, but I think is valid as a concern about this industry in general, is it is a high capital expenditure business. And that inherently brings some risks along with it. This is probably stuff you all already know, but it is really expensive to train these models. It takes a lot of compute, and that compute is in scarce supply. And when you put those things together, there's a lot of demand, there's not a lot of supply. I'm not an economics professor, but I think that means the price goes up. And so, the compute that is sort of the lifeblood of these companies— you have to buy really far in advance, and so you're essentially making a bet on the future. You're like, "We think we're going to need this much compute at this period of time." That's a really big expenditure to make, and it is a little bit harrowing to work at any of these companies.
I think if someone doesn't tell you that, maybe with the exception of Google because they're a public company, they have so much money. But I think certainly for Anthropic, for OpenAI, you're kind of making a calculated bet that you're going to be able to pay that money back over time. We obviously are very bullish on this. I think the revenue from both of those companies is like... unbelievable. It's something that I think, you know, we hear all the time, like venture capitalism, nothing like this has ever happened before, right? It's impossible to imagine a business, you know, getting to the kind of revenue numbers that are being talked about on such a short timescale.
And if that ever were to change, there would be a problem, right? Both of these companies have bought a lot of compute for the future. It's very expensive. And so I think that is probably the risk that feels... it's not crazy to be worried about that. We obviously think we're in a very good position. I think the industry as a whole is. But that could change any time. And I think it's important to remember that, like, that's ultimately a bet, right? It's ultimately the industry thinking, "Hey, this is gonna have a lot of returns to it." But like, we could absolutely be wrong.
I think if someone doesn't tell you that, maybe with the exception of Google because they're a public company, they have so much money. But I think certainly for Anthropic, for OpenAI, you're kind of making a calculated bet that you're going to be able to pay that money back over time. We obviously are very bullish on this. I think the revenue from both of those companies is like... unbelievable. It's something that I think, you know, we hear all the time, like venture capitalism, nothing like this has ever happened before, right? It's impossible to imagine a business, you know, getting to the kind of revenue numbers that are being talked about on such a short timescale.
And if that ever were to change, there would be a problem, right? Both of these companies have bought a lot of compute for the future. It's very expensive. And so I think that is probably the risk that feels... it's not crazy to be worried about that. We obviously think we're in a very good position. I think the industry as a whole is. But that could change any time. And I think it's important to remember that, like, that's ultimately a bet, right? It's ultimately the industry thinking, "Hey, this is gonna have a lot of returns to it." But like, we could absolutely be wrong.
**Yash：** 嗨，Daniela。我是 Yash，MBA 学生。我的问题是，Anthropic 认为政府监管和 AI 创新之间的正确平衡是什么？你希望全球各国政府在哪些方面做得不一样？
**Yash:** Hi, Daniela. I'm Yash, MBA student. My question is, what does Anthropic believe is the right balance between government regulation and AI innovation? And what do you wish governments around the world were doing differently?
**Daniela Amodei：** 好问题。我觉得这个领域的讨论——很遗憾，跟当下政治氛围的大趋势一样——很难做到有细微差别的讨论。这很可惜，因为这确实是一个需要细致讨论的问题。合理的监管必然会成为人工智能故事的一部分。AI 是一种与以往任何技术都非常不同的技术。即便是上一代技术，在我个人看来，也本可以从更多的监管中受益。
话虽如此，我们也清楚，公司需要有一定的自由空间去尝试新事物，才能打造出下一代人们想要使用和采纳的优秀产品。我觉得我对这场讨论最大的期望是它不要被政治化——但我担心这已经发生了。变成了"监管坏、创新好"或者"创新坏、监管好"的二元对立。实际上这非常复杂。有些监管领域确实没什么道理，但也有些监管领域对于确保技术以有利于人们的方式开发、防止坏事发生，是绝对关键的。
我的理想愿景是，科技公司和监管机构能够携手合作。因为我们掌握着技术可能被滥用的信息——我们每天都在看到。我们有安全团队、安保团队在监测人们如何试探这些模型，真正的风险是什么。而监管者知道如何提供一个可以被遵循和执行的框架体系。也许这过于乐观了，但我仍然对这个未来抱有希望——双方能找到共同点，一起思考"我们怎么确保能够开发下一个还不存在的惊人技术、下一个 Google 或 Meta 级别的公司，同时又设立一些常识性的监管来保护人们。"
话虽如此，我们也清楚，公司需要有一定的自由空间去尝试新事物，才能打造出下一代人们想要使用和采纳的优秀产品。我觉得我对这场讨论最大的期望是它不要被政治化——但我担心这已经发生了。变成了"监管坏、创新好"或者"创新坏、监管好"的二元对立。实际上这非常复杂。有些监管领域确实没什么道理，但也有些监管领域对于确保技术以有利于人们的方式开发、防止坏事发生，是绝对关键的。
我的理想愿景是，科技公司和监管机构能够携手合作。因为我们掌握着技术可能被滥用的信息——我们每天都在看到。我们有安全团队、安保团队在监测人们如何试探这些模型，真正的风险是什么。而监管者知道如何提供一个可以被遵循和执行的框架体系。也许这过于乐观了，但我仍然对这个未来抱有希望——双方能找到共同点，一起思考"我们怎么确保能够开发下一个还不存在的惊人技术、下一个 Google 或 Meta 级别的公司，同时又设立一些常识性的监管来保护人们。"
**Daniela Amodei:** Great question. So I think this is an area where I think the conversation about it has been, you know, sadly, as sort of is the case, I think, today, just in the sort of political climate, it's like hard to have a nuanced discussion. And I think that's a shame because I think it is a really nuanced question. I think sensible regulation will need to be part of the story for artificial intelligence, right? It's a very different technology than any other technology that's ever been built before. And even the last generation of technologies probably could have benefited from slightly more regulation, in my personal opinion.
That being said, we're not blind to the fact that you need to have some room to maneuver as a company to be able to try things if you want to come up with the next generation of incredible products that people want to use and that are adopted. And I think this is one of these things that my actual most kind of critical hope for the conversation is that it doesn't become politicized, which I fear it already has, right? It's sort of like, oh, like regulation bad, innovation good, or innovation bad, regulation good. I think it's just really complicated. I think there are areas of regulation that just don't make a lot of sense, and then I think there are areas of regulation that are absolutely critical to make the technology be developed in a way that is good for people and that will prevent bad things from happening to the people that rely on it every day and to the broader world.
My hope is that in an ideal world, what that would look like is technology companies and regulators, like, working hand-in-hand. Because we have the information about how the technology can be abused because we see it every day, right? We have safeguards teams and security teams that look at, like, how are people poking on this, and what are the actual risks? But regulators know how to provide a framework and a system that can actually be followed and enforced. And so... maybe it's overly optimistic, but I still hold out hope that that future is possible, where I think the two sides can find common ground and say like, "How do we ensure that we're able to develop the next, you know, amazing technology that doesn't even exist yet, or the next company that's going to be like, whatever, the next Google or the next Meta?" Um, but that we just put some common sense regulations in place that help protect people.
That being said, we're not blind to the fact that you need to have some room to maneuver as a company to be able to try things if you want to come up with the next generation of incredible products that people want to use and that are adopted. And I think this is one of these things that my actual most kind of critical hope for the conversation is that it doesn't become politicized, which I fear it already has, right? It's sort of like, oh, like regulation bad, innovation good, or innovation bad, regulation good. I think it's just really complicated. I think there are areas of regulation that just don't make a lot of sense, and then I think there are areas of regulation that are absolutely critical to make the technology be developed in a way that is good for people and that will prevent bad things from happening to the people that rely on it every day and to the broader world.
My hope is that in an ideal world, what that would look like is technology companies and regulators, like, working hand-in-hand. Because we have the information about how the technology can be abused because we see it every day, right? We have safeguards teams and security teams that look at, like, how are people poking on this, and what are the actual risks? But regulators know how to provide a framework and a system that can actually be followed and enforced. And so... maybe it's overly optimistic, but I still hold out hope that that future is possible, where I think the two sides can find common ground and say like, "How do we ensure that we're able to develop the next, you know, amazing technology that doesn't even exist yet, or the next company that's going to be like, whatever, the next Google or the next Meta?" Um, but that we just put some common sense regulations in place that help protect people.
**Jackie Kimmel：** 嗨，Daniela，感谢你来这里。我叫 Jackie Kimmel，是 MSX 学生。我的问题是，AI 正在获取越来越敏感的个人数据，比如我们的健康数据。你觉得个人应该做些什么来保护自己的隐私？
**Jackie Kimmel:** Hi, Daniela. Thanks for being here. My name's Jackie Kimmel. I'm an MSX student. My question is, AI is gaining access to increasingly sensitive personal data, such as our health data. What do you think individuals should actually be doing to protect their privacy?
**Daniela Amodei：** 我觉得这也是一个非常好的问题。首先你可能想不到，用 Claude 问医疗问题是一个多么常见的使用场景——这可能是我们最常见的日常使用场景之一。我自己也一直在用。"我孩子哪里不对劲了""我自己哪里不舒服，帮帮我"。我觉得这个问题有两个方面。第一，保护用户数据、谨慎使用用户数据，其实是公司的责任。这一点极其重要。我直说吧，我认为人们应该追究公司是否在认真保护他们的数据。因为这些数据非常私密。比如我们决定不在 Claude 中放广告，部分原因就是基于这样一个认知：AI 技术是不同的。人们与 AI 工具之间的对话，比他们放在 Instagram 或任何社交媒体上的内容都要私密得多。所以第一点是，有了这种认知，科技公司就有更大的责任去保护你的数据。
第二点，从个人角度来说，我不确定自己有完美的答案。我可以告诉你，很多人确实在用 AI 问医疗问题。从安全的角度来说，我的建议是不要把模型在医疗方面说的话当作绝对正确的。以我自己的经验，在复杂的医疗案例上，Claude 给出正确答案的次数比我的医生还多。但我永远不会在没有咨询持证医疗专业人士的情况下根据 AI 的建议采取行动。我们非常坦诚地说过，模型有时候会编造信息，会搞混，它不认识你，也没法给你做检查。所以保持一定程度的健康怀疑是完全正确的。
但你可以把它想象成，你有一个朋友是个很不错的医生，但不是专科医生。比如你要去看专科，你想在和医生的对话中得到一些引导。Claude 在这方面是一个很好的工具，它能帮你想到一些你可能不知道的潜在情况。但我最强烈的建议是：请不要仅仅因为某个 AI 工具说了"去做 X"就照做——我知道你们都太聪明了不会这样——但还是请带着一些怀疑的态度，去和专业人士沟通。
第二点，从个人角度来说，我不确定自己有完美的答案。我可以告诉你，很多人确实在用 AI 问医疗问题。从安全的角度来说，我的建议是不要把模型在医疗方面说的话当作绝对正确的。以我自己的经验，在复杂的医疗案例上，Claude 给出正确答案的次数比我的医生还多。但我永远不会在没有咨询持证医疗专业人士的情况下根据 AI 的建议采取行动。我们非常坦诚地说过，模型有时候会编造信息，会搞混，它不认识你，也没法给你做检查。所以保持一定程度的健康怀疑是完全正确的。
但你可以把它想象成，你有一个朋友是个很不错的医生，但不是专科医生。比如你要去看专科，你想在和医生的对话中得到一些引导。Claude 在这方面是一个很好的工具，它能帮你想到一些你可能不知道的潜在情况。但我最强烈的建议是：请不要仅仅因为某个 AI 工具说了"去做 X"就照做——我知道你们都太聪明了不会这样——但还是请带着一些怀疑的态度，去和专业人士沟通。
**Daniela Amodei:** I think this is also an excellent question. The first thing that I'll say is you would be surprised at how common of a use case it is to ask Claude medical questions. I think it's one of our most common just casual use cases. People... I use it all the time, right? I'm like, "This is wrong with my kid." "This is wrong with me. Help me." I think there's two sides to this. I think the first is it's actually— it's actually the company's job to use and protect your data with care. I think that's extremely important. I bluntly think people should hold companies accountable for using their data carefully, because it is very personal. And I think, for example, our decision to not put ads in Claude is partly predicated on this belief that AI technology is just different, right? People have conversations with AI tools that are much more personal than even what you would put on an Instagram account or... on social media of any form. And so I think the first is, like, with that knowledge comes more responsibility for the technology companies to actually protect your data.
The second is, I think from a personal perspective, like, I don't know that I have the perfect answer. I can certainly tell you, like, a lot of people use it for medical questions. I would really think about safety from the perspective of, like, don't take the models on faith about medical things. In my own experience, Claude has been right more often than my doctors about complex medical cases, and I would never do something without checking with a licensed medical professional, right? We are very open about the fact that, like, the models make things up sometimes. They get confused. They don't know you. They can't examine you, right? Just like I was saying on stage. So I think, like, having some healthy skepticism is extremely correct.
But think of it as like if you had a friend who was a really good doctor who maybe wasn't a specialist. So you're like, "I'm seeing a specialist and I would like help kind of being guided in this conversation with my doctor." I think Claude is a great tool for that. It's great for helping you think of, like, things that might be going on that you might not know. But I think my number one recommendation is please do not just do medical things that any AI... I mean, I know you're all too smart to do this, but like any AI tool just says, like, "Go do X." Like— look at it with some skepticism and actually talk to a professional.
The second is, I think from a personal perspective, like, I don't know that I have the perfect answer. I can certainly tell you, like, a lot of people use it for medical questions. I would really think about safety from the perspective of, like, don't take the models on faith about medical things. In my own experience, Claude has been right more often than my doctors about complex medical cases, and I would never do something without checking with a licensed medical professional, right? We are very open about the fact that, like, the models make things up sometimes. They get confused. They don't know you. They can't examine you, right? Just like I was saying on stage. So I think, like, having some healthy skepticism is extremely correct.
But think of it as like if you had a friend who was a really good doctor who maybe wasn't a specialist. So you're like, "I'm seeing a specialist and I would like help kind of being guided in this conversation with my doctor." I think Claude is a great tool for that. It's great for helping you think of, like, things that might be going on that you might not know. But I think my number one recommendation is please do not just do medical things that any AI... I mean, I know you're all too smart to do this, but like any AI tool just says, like, "Go do X." Like— look at it with some skepticism and actually talk to a professional.
**Interviewer：** 感谢学生们的提问。Daniela，在我们让你离开之前，我们还有 View from the Top 的快速问答传统环节。
**Interviewer:** Thank you to the students. And Daniela, before we let you go, we have to do our View from the Top rapid fire tradition.
**Daniela Amodei：** 太好了。
**Daniela Amodei:** Love it.
**Interviewer：** 准备好了吗？
**Interviewer:** Are you ready?
**Daniela Amodei：** 完全准备好了。
**Daniela Amodei:** I'm so ready.
**Interviewer：** 好，开始。如果回到大学，你会选什么专业？
**Interviewer:** Okay, let's go. What would you major in if you were back in college?
**Daniela Amodei：** 如果我说商科，是不是就可以直接跳过这个问题了？如果重新选的话，我大概还是会选文学。我知道这听起来很疯狂，但我就是喜欢读书。
**Daniela Amodei:** Can I, if I say business, does that just get me out of this entirely? Um, what would I major in? I would probably major in literature again. I know that sounds crazy. I like to read. Oh. Yeah.
**Interviewer：** 和你哥哥一起工作，你最喜欢的是什么？
**Interviewer:** What's your favorite thing about working with your brother?
**Daniela Amodei：** 感恩节晚餐。开玩笑的。我想说的是，我们对彼此太了解了，有时候我们能够对对方说一些公司里其他任何人都说不出口的话。有些话只有我们能说得出来，别人觉得自己说不了。
**Daniela Amodei:** Oh, Thanksgiving dinner. No, I'm joking. I would say that we know so much about each other, and we're able to have... like, sometimes we're able to say things to each other that nobody else in the company can. Yeah. Right? Like, sometimes we can get away with something that no one else feels like they could.
**Interviewer：** 那最不喜欢的呢？和你哥哥一起工作最不喜欢的？
**Interviewer:** What about least favorite thing— about working with your brother?
**Daniela Amodei：** 感恩节晚餐。不是啦。我觉得最难的是需要在个人关系和工作关系之间做切割。我们每周会安排专门的时间在办公室之外一起待着。但这确实不容易。我们做兄妹的时间远比做联合创始人的时间长，以后做兄妹的时间也会远比做联合创始人的时间长。确保在工作之外也持续浇灌这段关系，很重要。
**Daniela Amodei:** Thanksgiving dinner. No. I think just like the needing to also have separation between our personal relationship and our work relationship. So we build in time every week when we hang out outside of the office, but I think that's hard. Just so we were siblings for a long time before we were co-founders. We will be siblings for a long time after we're co-founders, and making sure to just continue to water that relationship outside of work.
**Interviewer：** 你在办公室图书馆发现的最喜欢的书是什么？
**Interviewer:** Favorite book you found at the library in your office?
**Daniela Amodei：** 我不确定我在那里发现了什么新书——好吧也许我应该多去翻翻。但我在那里重新想起了一本喜欢的书，这也许算数吧。有一本书叫《八月炮火》（The Guns of August）。我不知道在座有没有对一战感兴趣的——我看到一些茫然的表情，可能没有。这是一本很棒的书，如果你对一战感兴趣的话。我在图书馆里拿起来重读了一遍，上一次读应该是大学刚毕业之后。它是一个关于个人和人格特质的重要研究——正是这些个体的性格、事件和决策，导致了第一次世界大战的爆发。你会惊讶于这么多个体事件和人物命运交织在一起，最终导向了一个如此悲剧和可怕的结果。总之，强烈推荐这本书。
**Daniela Amodei:** Ooh, oh man. I don't know if I've discovered a new book, which maybe means I should get— I was like, "I like reading." Just kidding. I should probably go look more, but I was reminded of a favorite book there. Maybe that counts. There's a book called The Guns of August. I don't know if there's any World War I, like— I'm seeing some blank faces, so maybe not. It's a great book. If you're curious about World War I... I picked it up from the library and actually reread it because I read it, I think, maybe right after college or something. But I think it's a really important study of just, like, the individual people and personalities that resulted in and sort of led to the beginnings of what became World War I, and just how much, like, so many individual events and sort of people and personalities hinged on this, like, really ultimately sort of tragic and terrible thing that happened. So anyway, blanket shout-out: read that book. It's great.
**Interviewer：** 太好了。如果 Anthropic 最终用了一个不同的名字，会是什么？
**Interviewer:** Amazing. And if Anthropic had ended up with a different name, what would it be?
**Daniela Amodei：** 天哪，在想出 Anthropic 这个名字之前，我们经历了一些非常灾难性的想法。当时不知道为什么我们对鸟特别着迷。我记得有一个叫"Sparrow Systems"。完全不知道这个是怎么来的。但现在回想起来，早期的模型名字好像也都是鸟的名字。有 Bert，后来还有一个叫 Snuffleupagus——那不是鸟，但总之我们有这个"鸟"的执念。有一些非常糟糕的鸟名。幸运的是，更好的判断最终占了上风，我们把公司命名为 Anthropic。现在已经完全无法想象它叫别的名字了。
**Daniela Amodei:** Oh, man. Wow, we went through some, like, truly tragic ideas before we came up with Anthropic. I think for some reason we were, like, very into birds at the time. We're like Sparrow Systems, I think. No idea where that one came from. But I think actually now that I'm remembering, I think some of the, like, model names from the early days were all birds. So there was, like— Bert, which then became, like we had a Snuffleupagus. That's not a bird, but like then anyway, we had this kind of like bird thing. I don't know. There were some terrible bird names. Thankfully, better decisions prevailed and we named it Anthropic, which is now impossible to imagine as not our name. Imagine.
**Interviewer：** 最后一个，不是快速问答。你收到过的最好的建议是什么？
**Interviewer:** And finally, not a rapid-fire question. Best advice you ever received?
**Daniela Amodei：** 让我想想……我收到过的最好的建议。应该是在我们考虑离开的时候。现在回头看，人们会说"你们当然会离开然后创办 Anthropic"。但当时的感受完全不是这样的。我们觉得"这是一件很疯狂的事，也许我们应该留下来，也许我们能让现状好转。"
但我找了一个工作之外的朋友兼导师聊，她说："说实话，我觉得你不需要打这个电话给我。你其实已经知道正确答案是什么了。"我觉得这个道理是普遍适用的——当你面临"这对我的人生来说是不是正确的决定"的时刻，你往往其实已经知道答案了。这真的是很好的建议。
但我找了一个工作之外的朋友兼导师聊，她说："说实话，我觉得你不需要打这个电话给我。你其实已经知道正确答案是什么了。"我觉得这个道理是普遍适用的——当你面临"这对我的人生来说是不是正确的决定"的时刻，你往往其实已经知道答案了。这真的是很好的建议。
**Daniela Amodei:** Ooh. Drop the ant. No, I'm trying to think of... best advice I've ever received. I think probably, you know, I will say when we were thinking about leaving. And it was— it's now in retrospect— people are like, "Of course, you guys, you all left and you founded Anthropic." It didn't feel that way at the time. Yeah. We were like, "This is a really, you know, this is kind of a crazy thing to do. Maybe we should, maybe we should stay. Maybe we could make this work."
But we talked, I talked to one, you know, sort of friend and mentor outside of work and she was like, "Honestly, I don't think you really need to be on the phone with me. Like, you already know what the right answer is." And I think that, in general, like when you're in a moment of, "Is this the right thing for my life?" Often you actually know what the right answer is. And I think that was really good advice.
But we talked, I talked to one, you know, sort of friend and mentor outside of work and she was like, "Honestly, I don't think you really need to be on the phone with me. Like, you already know what the right answer is." And I think that, in general, like when you're in a moment of, "Is this the right thing for my life?" Often you actually know what the right answer is. And I think that was really good advice.
**Interviewer：** Daniela，今天的对话非常愉快。
**Interviewer:** Daniela, it's been a pleasure.
**Daniela Amodei：** 谢谢。
**Daniela Amodei:** Thank you.
**Interviewer：** 非常感谢你。
**Interviewer:** Thank you so much.
**Daniela Amodei：** 感谢你们邀请我来。
**Daniela Amodei:** Thanks for having me.
