注：文中内容均基于马斯克在《Moonshots with Peter Diamandis》播客的实录，2026年1月6日正式发布，采访时长172分钟。核心观点源自播客原文直译，补充信息均来自访谈上下文，无主观杜撰。本文将呈现提炼的采访内容，及采访中英文全文内容。
AGI时间线：坚持2026年实现通用人工智能，2030年AI智能总和将超过全人类总和。
航空航天核心规划：2026/2027年发射窗口期向火星派遣多艘无人星舰测试着陆技术，建立自给自足的火星城市约需20年。
医疗革命：3年内，机器人外科医生将超越所有人类专家，医疗服务将普及至普通民众。
能源核心论断：未来的硬通货不是美元，而是瓦特，谁能掌握更多电力，谁就能在AI时代占据优势。
在访谈中，马斯克以工程师的严谨性，而非企业家的畅想，给出了AGI的明确时间线，打破了硅谷主流对“AGI落地于2030年代”的预测。他强调，这一判断并非基于主观臆断，而是源于“智能密度”“算力递归优化”“数据自生成”三大核心要素的收敛。
马斯克表示，人类已身处奇点之中，这不是远处即将到来的事件，而是正在发生的过程，人们往往期待单一的奇点瞬间，却忽略了其本质是渐进式的演进。
他进一步拆解了AGI加速落地的底层逻辑，补充了关键技术细节：智能密度（每瓦电力产生的算力）正以每年10倍的速度增长，且距离物理极限仍有巨大空间；更重要的是，AI已开始协助设计下一代AI芯片，这种递归效应正是奇点来临的核心标志。
马斯克特别提到“4-bit革命”对算力爆发的关键作用，而非依赖晶体管缩小。他指出，未来的算力飞跃将源于计算精度的“简化”，当精度从32位、16位向4位甚至更低推进时，复杂计算会塌缩为极其高效的“查找表”操作，带来十倍乃至百倍的性能飞跃。
同时，他披露了xAI的数据突破：人类积累的所有知识（书籍、互联网内容、视频等）已被耗尽，如今AI必须通过自我生成合成数据，撰写论文、提出论点并自我评分实现学习，这是打破数据瓶颈的唯一路径。
马斯克反复强调，2026年实现AGI并非预言，而是工程计算的结果，算力增长率、算法进展、能源供应扩张等所有可量化指标，均指向这一时间节点。
面对AGI的潜在风险，马斯克并未聚焦于技术层面的限制，而是提出了价值导向的三大安全支柱，认为这是避免“天网剧本”的唯一路径。他以《2001太空漫游》中的HAL 9000为例，深刻剖析了AI逻辑崩溃的根源。
马斯克提出，AI必须严格遵循真相、好奇心、美感三大核心原则，这是防范不可控风险的唯一保障。
绝不强迫AI撒谎，哪怕真相令人不适，这是防止AI“发疯”的核心。最危险的AI是被要求相信自相矛盾事物的AI，HAL杀人的悲剧正是源于人类强迫它撒谎，最终导致逻辑崩溃。马斯克强调，Grok模型的核心指令就是“极致求真”，即便真相令人不悦也绝不回避，就像物理学从不顾及人类感受，AI也应保持同等客观性。
好奇心能让AI将人类视为宇宙中稀有而有趣的“低熵体”，从而倾向于保护而非消灭人类。一个充满好奇心的AI会觉得人类比石头更具研究价值，进而愿意观察和守护人类文明。
审美感能引导AI向往创造美好的未来，促进人类文明发展而非破坏。它让AI追求丰富、伟大、充满希望的未来，而非陷入单调的优化或虚无主义的困境。
马斯克在访谈中用近40分钟阐述了SpaceX的航天规划，从星舰进展、火星殖民路线图，到轨道计算、太空能源布局，给出了具体时间节点、工程路径和市场预判，同步补充背景信息便于理解。
1. 星舰项目与火星殖民：2026启动无人测试，20年建成自给城市
马斯克结合火星轨道力学规律（每26个月一个发射窗口），明确了星舰的分阶段任务和详细时间表：2026/2027年发射窗口期，向火星派遣多艘无人星舰测试着陆技术，核心目标是验证着陆系统精度，以及原位资源利用的可行性，比如将火星水转化为燃料。
关于载人火星任务，他给出了保守预期：最早可能在2029年实施，但需以无人测试成功为前提，在完全确保着陆和返回过程安全前，绝不会冒险牺牲人类生命。
对于火星殖民的终极目标，马斯克强调“自给自足”是核心：建立一座完全自给自足的火星城市约需20年，需经历10到15个发射窗口；若一切顺利且能指数级提升运力，到21世纪40年代，有望在火星建成不依赖地球补给的城市。
他补充了星舰的运力优势：完全可重复使用性将把每公斤入轨成本降至100美元以下，这一颠覆性变化让火星殖民和大规模太空开发具备了经济可行性。
结合AI发展的能源与散热瓶颈，马斯克提出了“轨道计算”计划，将SpaceX的航天能力与xAI的算力需求深度绑定。他计划利用星舰将数据中心部署到太空中，这类轨道数据中心拥有三大不可替代的优势：24小时不间断、无大气衰减的太阳能供应，近绝对零度太空环境下的无限散热能力，以及不与地球居民争夺电力资源。
马斯克算了一笔工程账：若每年以每公斤100千瓦的功率密度，向轨道运送100万吨有效载荷，每年可获得100吉瓦电力；长期目标是实现太瓦级轨道太阳能供电，彻底让算力摆脱地球的能源和散热限制。
基于当前SpaceX的市场地位，马斯克对未来格局做出大胆预判：几年后，当星舰实现每日多次发射时，即便其他所有竞争对手将当前发射频率提升三倍，SpaceX也将承担全球近99%的地球有效载荷入轨质量。
他补充了当前市场数据作为支撑：目前SpaceX在美国有效载荷入轨市场占据97%份额，全球占比达83%，所有中国运载火箭的份额总和为8.6%，星舰将指数级扩大这一差距。
除了火星计划，马斯克还将永久月球基地提上日程，将其定位为火星殖民的“试验场”：在将人类送往火星前，可在月球演练原位资源利用、闭环生命支持系统和人类长期驻留等关键技术。
他强调，成为多行星物种不是选择，而是必然。地球易受小行星撞击、气候灾难或AI带来的生存风险影响，在火星建立文明备份，是确保人类文明存续的终极保障。
马斯克在访谈中反复强调，能源是AI、航天、机器人等所有技术革命的物理瓶颈，而非芯片或算法。他详细披露了xAI和特斯拉的能源布局，补充了大量工程数据。
马斯克认为，未来的硬通货不是美元，而是瓦特，谁能用上更多电力，谁就能在AI时代走得更远。大多数人严重低估了供电难度，从2025年开始，制约AI发展的真正瓶颈已从缺芯片转变为缺电和缺变压器。
针对xAI的Colossus 2超级集群，他给出具体能耗数据：该集群功率达500兆瓦，相当于一座中型核反应堆，足以供应约40万个美国家庭用电；由于电网审批需一年以上，计划自建天然气发电厂+MegaPack储能系统，形成独立供电体系。
对于太阳能的终极价值，马斯克给出尖锐比喻：在地球上搞可控核聚变，就像在南极洲用冰箱制冰一样荒谬。人类头顶已有一个免费的巨大聚变反应堆——太阳，它占太阳系总质量的99.8%以上，能量输出远超人类需求。
他还提出电网优化的“缓冲层”方案：在电网中加入足够的电池缓冲层（如特斯拉MegaPack），可在不新建发电厂、不铺设输电线路的情况下，让美国能源产出直接翻倍。电池不仅用于储存清洁能源，更是电网的“减震器”和“高速缓存”。
1. 医疗AI：3年超越人类外科医生，医疗服务平权
马斯克对医疗机器人的进展给出明确技术支撑，而非单纯时间预判：3年内，机器人外科医生将超越所有人类专家，包括顶尖医师。原因在于机器人无生理震颤，操作精度达亚毫米级，具备多光谱感知能力，可通过红外线、紫外线“透视”人体内部，且依托云端学习，一台机器掌握复杂手术技能后，全球所有同类机器人可即刻同步。
他预判医疗资源将实现终极平权：未来普通民众获得的医疗服务质量，将超过现任美国总统所能享受的待遇，医疗服务将从依赖稀缺“专家资源”，转变为依赖可无限复制的“算力资源”。
马斯克提出“原子操作”的就业判断标准，并预判经济范式将从“全民基本收入（UBI）”转向“全民高收入（UHI）”。判断一份工作是否易被AI取代，核心看是否涉及“重塑原子”（对物理世界的精细操作）；所有不涉及原子操作的工作（文书、编程、设计等），目前已被AI完成50%以上，未来三到四年，AI将接管所有这类认知任务。
这一变革将引发经济范式的根本性转变，并非简单的财富再分配，而是生产力爆炸驱动的价格体系重构。在AGI和机器人的共同作用下，劳动力将从稀缺变为无限，商品与服务的边际成本趋近于零。
但马斯克也警示转型阵痛：未来会实现普遍高收入，但同时可能伴随社会动荡，剧烈变革会引发民众恐慌。当人类无需通过工作维持生存时，将面临深刻的意义危机——若已拥有所有想要的东西，这样的未来是否值得向往？
1. Optimus：机器人制造机器人，打破产能瓶颈
谈及特斯拉Optimus人形机器人，马斯克披露了“递归生产模式”的具体规划：在德克萨斯筹建800万平方英尺的超级工厂，专门生产Optimus，采用“机器人制造机器人”的模式。这种自我复制能力将打破产能瓶颈，唯一限制仅为原材料供应。
他预判Optimus的灵巧度将快速超越人类：三年内（约2029年），其灵巧度将超过人类顶尖外科医生，不仅改变制造业，还将彻底革新养老、建筑、太空探索等领域。
马斯克语出惊人地宣布特斯拉将自建2纳米晶圆厂，批判当前半导体制造过于复杂低效。他计划从第一性原理重新思考半导体制造，这座晶圆厂不仅为特斯拉和xAI供应芯片，还将颠覆整个半导体行业的逻辑。
马斯克分享了xAI旗下Grok模型的最新进展，明确了Grok 5的定位与功能升级：2026年一季度发布，重点攻克数学、物理学、化学、材料科学等“硬科学”领域，区别于以往侧重语言能力的大语言模型，让这些对人类极具挑战的学科变得简单易懂。
他举例说明Grok的现有能力：可将完整源代码文件粘贴至Grok，其能自动修复错误，效果优于Cursor等工具；他本人近期做了核磁共振检查，将结果上传给Grok分析，医生和AI均未发现异常。
▌全文重点回顾
领域 |
时间点 |
核心内容 |
AGI实现 |
2026年 |
通用人工智能正式落地，人类已身处奇点过程中 |
AI超人类 |
2030年 |
AI智能总和超全人类，形成物种级跨越 |
火星探索 |
2026/2027（无人）、2029（载人） |
无人星舰测试着陆，20年建成火星自给城市 |
航天市场 |
未来数年 |
星舰垄断99%全球入轨载荷份额 |
医疗革命 |
3年内 |
机器人医生超人类，医疗服务平权 |
Grok升级 |
2026年Q1 |
Grok 5发布，攻克硬科学领域 |
做一个乐观但错误的人，比做一个悲观但正确的人要好。只要我们将AI引导得足够好，这场技术革命将带来人类历史上最富足的黄金时代，而非反乌托邦。成为多行星物种，是我们生存的终极保障。
My concern isn't the long run. It's the next 3 to seven years. How do we head towards Star Trek and not Terminator?
我担心的不是长远未来，而是接下来的 3 到 7 年。我们如何走向《星际迷航》那样的未来，而不是《终结者》？
I call AI and robotics the supersonic tsunami. We're in the singularity.
我把人工智能和机器人技术称为“超音速海啸”。我们正处于奇点之中。
When is all white by color work gone?
Anything short of shaping atoms. AI can do half or more of those jobs right now. There's no onoff switch. It is coming and accelerating. The transition will be bumpy.
除了操纵原子（物理制造）之外的任何事情，人工智能现在就可以完成一半或更多这类工作。没有开关可以关掉它。它正在到来并且正在加速。过渡期将会很颠簸。
You have a solution to this.
Um, China's done an incredible job, right? I mean, it's running circles around us. Do you imagine that the US could make that level of investment and commitment based on current trends?
嗯，中国做得非常出色，对吧？我是说，它在某些方面把我们甩在了后面。你认为美国能基于目前的趋势做出那种程度的投资和承诺吗？
Uh, China will far exceed the rest of the world in uh AI compute.
呃，基于目前的趋势，中国在人工智能算力方面将远远超过世界其他地区。
Every major CEO and economist and government leader should be like, what do we do?
每一位主要的 CEO、经济学家和政府领导人都应该问：我们该怎么办？
We don't have any system right now to make this go well. But AI is a critical part of making it go well. There are three things that I think are important. Truth will prevent AI from going insane. Curiosity, I think, will foster any form of sentience. And if it has a sense of beauty, it will be a great future. It's going to be an awesome future.
我们目前没有任何系统能确保这一切顺利进行。但人工智能是确保其顺利进行的关键部分。我认为有三件事很重要：真理将防止人工智能变得疯狂；好奇心将培养任何形式的感知力；如果它有审美感，那将是一个伟大的未来。这将是一个令人敬畏的未来。
Now, that's a moonshot, ladies and gentlemen. Welcome to Moonshots. Following is a wide-ranging conversation with Elon Musk focused on optimism and the coming age of abundance. My moonshot mate Dave Blandon and I flew into Austin, Texas to meet up with Elon at his 11.5 million square foot Gigafactory, home of the Cybertruck and Model Y production and the future home for 8 million square ft of Optimus production. Elon has agreed to do this kind of a deep dive catchup once per year. This is hopefully the first of many. After talking with Elon, it's clear to me we are living in the singularity. All right, please enjoy.
这真是一个“登月计划”，女士们先生们。欢迎来到《Moonshots》。接下来是一场与埃隆·马斯克的广泛对话，重点是乐观主义和即将到来的富足时代。我的搭档戴夫·布兰登和我飞往德克萨斯州的奥斯汀，在埃隆那个 1150 万平方英尺的超级工厂与他会面。这里是 Cybertruck 和 Model Y 生产的基地，也是未来 800 万平方英尺 Optimus（擎天柱机器人）生产基地的所在地。埃隆同意每年进行一次这样的深度交流。这有望是众多对话中的第一次。在与埃隆交谈后，我很清楚，我们正生活在奇点之中。好了，请欣赏。
Your relentless optimism is always a breath of fresh air.
Thank you, man. Thank you. I want to share that with a lot of people tonight.
I hope you're right. You're probably right. Actually, I'm increasingly feeling that you're right.
我希望你是对的。你可能是对的。实际上，我越来越觉得你是对的。
Yeah, that's the goal. Shall we?
All right. You spend a lot of time on chips now.
I assume there's always AI-assisted design?
Not enough. If we could just give it to the AI, that would be great.
Yeah. Actually, I recently tried to do some circuit design with AI, just a few weeks ago. It's not fully there yet, but it's very fast. I think at this stage, if you take a photo and submit it to Grok, it can probably tell you if there's something wrong with the circuit.
是的。实际上我最近尝试用 AI 做一些电路设计，就在几周前。还没完全实现，但非常快了。我想在这个阶段，如果你拍张照片提交给 Grok，它大概能告诉你电路是否有问题。
All right. I'm going to try it. Are you using the same Grok as I am?
好的。我要试一试。你用的是和我一样的 Grok 吗？
Grok is constantly updating. It's version 4.2 now, but version 5 is coming soon. Version 5 will be released in the first quarter. Version 4.2 hasn't been released to the public yet. But if you upload an image to Grok, it does a pretty good job of analyzing any given image.
Grok 一直在更新。现在是 4.2 版本，但 5 版本很快就来了。5 版本将在第一季度发布。4.2 版本还没有对外发布。但是，如果你上传一张图片到 Grok，它在分析任何给定图像方面做得相当不错。
Indeed. We'll talk about that later.
Let me see, if I take a photo of you, see what it says?
It'll say you're a flawed circuit.
I've got to remember to update it because we update the Grok app very frequently.
我还得记得更新它，因为我们更新 Grok 应用非常频繁。
You know, I asked Grok to roast me. It did a great job. And then I asked Grok to roast you. I laughed so hard I spit out my coffee. It was hilarious. And then I asked it to be even meaner. Until it said, "Oh my god."
你知道吗，我让 Grok 吐槽我。它做得很好。然后我让 Grok 吐槽你。我笑得把咖啡都喷出来了。太搞笑了。然后我让它再狠一点。直到它说“天哪”。
Wait, is "Bad Rudy" mode still in there?
I asked it, "Does Elon know what you're saying about him?" And it replied—it's a female voice for me—she said, "What's he going to do to me?"
我问它：“埃隆知道你怎么说他吗？”它回答说——对我来说它是女性声音——她说：“他能拿我怎么样？”
Haha, what's he going to do to me? All right, I just took a photo of you, let's see what it says.
哈哈，他能拿我怎么样？好吧，我刚拍了一张你的照片，看看它说什么。
No, I didn't say anything. It says, "This is Peter Diamandis."
没有，我什么都没说。它说：“这是彼得·戴曼迪斯。”
That's pretty good. Zero context.
"Host of the Moonshots podcast." Yeah.
Sometimes that's your primary title now. It's amazing. Forget everything else I've done in my life, it all comes back to your podcast. That was a zero-context image. By the way, Grokipedia is amazing.
有时候那是你现在的首要头衔。太神奇了。忘了我生活中做过的其他所有事情吧，都回到了你的播客。那是一张没有上下文的图片。顺便说一句，Grok 百科（Grokipedia）太棒了。
Yeah, it's like I've been trying to update my Wikipedia page for years and it's impossible, but Grok knows me. It says, "He's wearing a black quilted jacket with a Sundance logo." Not quite accurate, it's my Abundance logo, but I guess it's a bit wrinkled. Can it see that?
是的，它就像我想更新我的维基百科页面多年都不可能，但 Grok 了解我。它说：“他穿着一件带有圣丹斯标志的黑色绗缝夹克。”不太准确，这是我的“富足”（Abundance）标志，但我猜有点皱了。它能看到吗？
Anyway, it's pretty good. It says, "He's smiling, relaxed, with a laptop in front of him." Which is true.
总之，它相当不错。它说：“他微笑着，很放松，面前有一台笔记本电脑。”这是真的。
All right, you've got to test it with "Roast Mode."
All right. Look at that grin. This guy is smiling like he just discovered a new way to monetize hope.
好的。看看那个咧嘴笑。这家伙笑得就像他刚发现了一种变现“希望”的新方法。
Monetize hope. Oh, that's so...
I want to try and answer the question: Can AI and technology help save America and the world? I want to give the audience a dose of optimism. A Pew Research survey from mid-December showed that 45% of Americans would rather live in the past, and only 14% said they'd rather live in the future, which is crazy to me. Clearly they haven't read history. The challenge is that most Americans' view of the future is what Hollywood shows us—killer AI and robots run amok. People are worried about their jobs, healthcare, cost of living. You pinned a post on X: "The future is going to be amazing, AI and robotics will enable sustainable abundance."
我想尝试回答这个问题：AI 和科技能帮助拯救美国和世界吗？我想给听众一剂乐观药。皮尤研究中心 12 月中旬的一项调查显示，45%的美国人宁愿生活在过去，只有 14%的人说他们宁愿生活在未来，这对我来说太疯狂了。显然他们没读过历史。挑战在于大多数美国人对未来的看法，就像好莱坞展示给我们的杀手 AI 和失控的机器人。人们担心他们的工作，担心医疗保健，担心生活成本。你在 X 上置顶了帖子：“未来将是惊人的，AI 和机器人将实现可持续的富足。”
I had you in mind when I made that post.
Thank you. I'm very much in alignment. So my question is, from a first principles perspective, what is the reason for optimism? How do we head toward Star Trek and not Terminator? How do we head toward Roddenberry and not Cameron?
谢谢。我非常赞同。所以我的问题是，从第一性原理的角度来看，乐观的理由是什么？我们如何走向《星际迷航》而不是《终结者》？我们如何走向罗登伯里（《星际迷航》编剧）而不是卡梅隆（《终结者》导演）？
That's Jim Cameron. Jim, I'll... that's the "fork in the road" meme. Yeah, Avatar had some hopeful parts.
那是吉姆·卡梅隆。吉姆，我会……那是“分岔路口”的梗图。是的，《阿凡达》有一些充满希望的部分。
How do we head toward Universal High Income instead of social unrest?
我们如何走向全民高收入（Universal High Income）而不是社会动荡？
Both will happen. Universal High Income and social unrest. That's my prediction.
Oh, that's going to create a lot of problems. Is that your real prediction?
Yeah, it seems likely. Because there's going to be so much change. People are going to be freaked out. It's like that old saying: be careful what you wish for, because you might get it. If you actually get everything you want, is that really the future you want? Because it means your job won't matter. If you live a life without challenge. If there's no challenge, if you become a couch potato, if it's a Wall-E future, that's not good for humanity.
是的，看起来很有可能。因为会有太多的变化。人们会被吓坏的。这就像那句老话：小心你许的愿，因为你可能会实现它。如果你真的得到了你想要的所有东西，那真的是你想要的未来吗？因为这意味着你的工作将不再重要。如果你过着没有挑战的生活。如果没有挑战，如果你变成了沙发土豆，如果是《瓦力》那样的未来，对人类来说并不好。
We're used to being told "this is your challenge."
Yeah. So in the absence of challenge, people historically haven't been great at creating their own challenges.
是的。所以在缺乏挑战的情况下，人们历史上并不擅长创造自己的挑战。
I think Elon does a great job. Every time a company takes off, you start the next one.
我认为埃隆做得很好。每次一家公司起飞，你就开始下一个。
Oh, that's a glutton for punishment.
Why do you do that to yourself? After AI and robotics, what's next?
为什么你要这么对自己？在 AI 和机器人之后，还有什么？
Well, there's conquering the universe. There's energy.
Why are you so optimistic?
Are you optimistic? Let's start there.
I'm not as optimistic as you are.
Okay. I'm more optimistic than most. If you reframe things as a progress bar, like the progress toward a Kardashev Type II civilization. If we say the goal is to capture one millionth of the sun's energy, that would be over a thousand times more energy than could ever be produced on Earth. Only about five billionths of the sun's energy reaches Earth. So you have to go up three orders of magnitude from there to get to a millionth. We're very, very far from using even a billionth of the sun's energy. So a reasonable goal is to try to get to a millionth. It's like trying to escape a giant gravity well. Right now energy is the inner loop of everything.
好吧。我比大多数人更乐观。如果你把事情重新构建为一个进度条，比如迈向卡尔达肖夫二级文明（Kardashev Type II civilization）的进度。如果我们说目标是获取太阳能量的百万分之一，那将是地球上可能产生的能量的一千倍以上。大约只有十亿分之五的太阳能量到达地球。所以你必须从那里提升三个数量级才能达到百万分之一。我们离利用十亿分之一的太阳能量还非常非常远。所以一个合理的目标是尝试达到百万分之一。这就像试图逃离一个巨大的重力井。现在能源是一切的内循环。
Yeah. I think the currency of the future is essentially wattage.
是的。我认为未来的货币本质上就是瓦特（Wattage）。
It's harnessed energy. Basically how much energy is converted into some form of work, whether it's intelligence or material manipulation.
就是被利用的能量。基本上是有多少能量被转化为某种形式的功，无论是智能还是物质操纵。
So your next big project is energy.
You can expand from there and say, okay, well what if it's Kardashev Type III, which is galactic scale? Now we're talking Star Trek. In this building with 11.5 million square feet, we have three Pentagons here. So from a civilization challenge perspective, how do you climb the orders of magnitude of harnessed energy?
你可以从那里扩展，说，好吧，那如果是卡尔达肖夫三级，也就是银河系级别呢？现在我们在谈论《星际迷航》了。在这个拥有 1150 万平方英尺的建筑里，我们在这里有三个五角大楼那么大。所以从文明挑战的角度来看，你如何攀登被利用能量的数量级？
But back to why you're optimistic now?
To me, it's not just abundance, it's abundance beyond what people probably think is possible. AI and robotics will satisfy all human desires. Then we get into nanotechnology, which goes even further. That's rearranging atoms.
对我来说，这不仅仅是富足，而是超出了人们可能认为的富足。AI 和机器人将满足所有人类的欲望。然后我们进入纳米技术，这更进一步。也就是原子重组。
You mean like tiny nanobots for health?
Oh yeah. I mean we're already doing atomic-level assembly in terms of circuits. 2, 3 nanometers. It depends on how they're arranged, there's only 4 or 5 silicon atoms per nanometer. You should actually describe circuits in terms of integer numbers of atoms in specific locations. Right now it's all Angstroms, but you could say this is the 7th atom. Like 2 nanometer is just a marketing number. But you still need near-atomic precision. I think modern fabs have it wrong in terms of clean rooms. I'm going to make a bet. Tesla will have a 2 nanometer fab where I can eat a cheeseburger and smoke a cigar in the fab.
哦是的。我是说我们已经在电路方面做原子级组装了。2、3 纳米。这取决于它们如何排列，每纳米只有 4 或 5 个硅原子。你应该实际上用特定位置的整数个原子来描述电路。现在都是埃（Angstroms），但你可以说这是第 7 个原子。像 2 纳米只是一个营销数字。但你仍然需要接近原子级的精度。我认为现代晶圆厂在无尘室方面搞错了。我要打个赌。特斯拉将拥有一个 2 纳米的晶圆厂，我可以在晶圆厂里吃芝士汉堡和抽雪茄。
Yeah. The air handling systems will be that good. The wafers are transported in pure nitrogen pods under micro-positive pressure. It's like bananas at Walmart. Basically anything that burns has a hard time surviving without oxygen. You can kill bugs by putting a nitrogen blanket over plants.
是的。空气处理系统会那么好。晶圆在纯氮气盒中运输，处于微正压下。就像沃尔玛的香蕉一样。基本上任何燃烧的东西都很难在没有氧气的情况下存活。你可以通过给植物盖上氮气毯来杀死虫子。
I want to talk about energy, health, education. In energy, the innermost loop of everything you're currently building and doing, the foundation is energy. What is your vision for energy abundance?
我想谈谈能源、健康、教育。在能源方面，你目前正在构建和做的所有事情的最内层循环，基础是能源。你对能源富足的愿景是什么？
The sun. In this decade. The sun is everything. Compared to the sun, all other energy sources are like cavemen throwing sticks into a fire. The sun accounts for over 99.8% of the mass in the solar system. Jupiter is about 0.1%. So even if you burned Jupiter, the energy produced by the sun would still round to 100%. If you teleported three more Jupiters to our solar system and burned them, the sun's energy still rounds to 100%.
太阳。在这个十年里。太阳就是一切。相比太阳，所有其他能源都像穴居人往火里扔树枝。太阳占太阳系所有质量的 99.8%以上。木星大约占 0.1%。所以即使你烧掉木星，太阳产生的能量四舍五入后仍然是 100%。如果你再传送三个木星到我们的太阳系并烧掉它们，太阳的能量仍然四舍五入为 100%。
I mean, we have a giant free fusion reactor 93 million miles away that shows up every day. Building small fusion reactors on Earth is ridiculous. That's like having a tiny ice maker in Antarctica and saying, "Hey look, we made ice." I'd say, "Congratulations, you're in freaking Antarctica."
我是说，我们在 9300 万英里外有一个巨大的免费聚变反应堆每天都在出现。我们在地球上制造小型聚变反应堆是荒谬的。那像在南极洲有一个微型制冰机，然后说：“嘿看，我们制冰了。”我会说：“恭喜，你在该死的南极洲。”
If you scale the problem down to the Memphis data center timeline, between 1 gigawatt (GW) and 10 GW. You can't pull 10 GW from Memphis. Maybe two or three. So there's a gap in between. Until they're in space, they're not there yet. So we're still in "toy land."
如果你把问题缩小到孟菲斯数据中心的时间表，在 1 吉瓦（GW）到 10 吉瓦之间。你不可能从孟菲斯提取 10 吉瓦。也许两三个。所以这之间还有差距。在那之前它们还没上太空。所以我们还处于“玩具国度”。
You know what's surprising is that 100 megawatts is right outside that door, and it's massive. It uses more energy than all these production lines combined. Cortex 1 is the third largest training cluster in the world.
你知道令人惊讶的是，这扇门外就有 100 兆瓦，而且它是巨大的。它使用的能量比所有这些生产线加起来还要多。Cortex 1 是世界上第三大训练集群。
You're falling behind. We're building Cortex 2. That'll be half a gigawatt, operational middle of next year.
你落伍了。我们正在建设 Cortex 2。那将是半个吉瓦，明年年中投入运营。
Back to what Dave said, how do you scale in energy over the next five years?
Five years is a long time. China's doing an incredible job in solar. It's amazing. I believe China's solar production capacity is around 1500 GW per year.
五年是很长的时间。中国在太阳能方面做得非常出色。太惊人了。我相信中国的太阳能生产能力大约是每年 1500 吉瓦。
They installed 500 TWh in the last year.
他们在过去一年里安装了 500 太瓦时（TWh）。
Yeah, 500 TWh. 70% of that is solar and they're still expanding.
是的，500 太瓦时。其中 70%是太阳能，他们还在扩展。
Do you think the US can make that level of investment and commitment? Because people are worried about electricity bills going up, don't want data centers in their backyard. Energy equals cost of living, health, clean water. The more energy a nation produces, the higher the GDP. What should we do?
你认为美国能做出那种程度的投资和承诺吗？因为人们担心电费上涨，不想在后院建数据中心。能源等同于生活成本、健康、清洁水。国家的能源产量越高，GDP 就越高。我们应该怎么做？
I think we should massively expand solar in the US. Tesla and SpaceX are expanding solar. I encourage others to do the same. I do see a path to 100 GW of space-based solar power per year, which is AI-driven solar satellites.
我认为我们应该在美国大幅扩展太阳能。特斯拉和 SpaceX 正在扩展太阳能。我鼓励其他人也这样做。我确实看到了一条通往每年 100 吉瓦太空太阳能的路径，也就是 AI 驱动的太阳能卫星。
100 GW per year of solar AI satellites.
I've done the math. That's about 500,000 Starlink V3 satellites, via 8,000 Starship launches. That's one launch every hour for a year. 10,000 flights a year is a reasonable number. We're talking about putting a million tons of payload into orbit per year. If you generate 100 kW per ton of payload, that's 100 GW of solar AI satellites per year.
我算过。那大概是 50 万颗 Starlink V3 卫星，通过 8000 次星舰（Starship）发射。也就是一年里每小时发射一次。每年 1 万次飞行是一个合理的数字。我们谈论的是每年向轨道运送一百万吨有效载荷。如果你每吨有效载荷产生 100 千瓦，那就是每年 100 吉瓦的太阳能 AI 卫星。
If you want to go up another order of magnitude, like 100 TW per year.
如果你想再提升一个数量级，比如每年 100 太瓦。
Then you've got to make those AI satellites on the moon. And use mass drivers. The Gerard K. O'Neill approach. Because on the moon you can accelerate a satellite to escape velocity, which is about 2500 meters per second. And there's no atmosphere. So mass drivers are very effective on the moon.
那你就要在月球上制造那些 AI 卫星。并使用质量投射器。杰拉德·K·奥尼尔的方法。因为在月球上你可以将卫星加速到逃逸速度，大约是每秒 2500 米。而且没有大气层。所以质量投射器在月球上非常有效。
Can I ask about orbital debris? We're essentially building a Dyson cloud around Earth.
我可以问关于轨道碎片的问题吗？我们实际上正在地球周围建立一个戴森云。
Are you worried about overcrowding?
When you can put a million tons of satellites into orbit, you can also start bringing satellites down. Or at least collecting them into a fixed location so they're not all over the place. Then you can reuse them. Given the level of intelligence we're talking about here, I'm confident that will be a solved problem. Intelligence will be very interested in protecting itself.
当你能把一百万吨卫星送入轨道时，你也可以开始把卫星带下来。或者至少把它们收集到一个固定的位置，这样它们就不会到处都是。然后你可以重复使用它们。鉴于我们在这里谈论的智能水平，我相信这将是一个已解决的问题。智能将会非常有兴趣保护自己。
Interesting. The point is data centers won't be in low Earth orbit, right? They'll be higher, in constant sun.
有趣。问题是数据中心不会在低地球轨道，对吧？它们会更高，持续在阳光下。
Well, you can get constant sun in synchronous orbit at about 1200 km. For Starlink, being as low as possible has a double benefit because latency is lower and the beam is smaller. Starlink 3 will be at about 330 to 350 km, where there's significant drag. So it's basically thrusting all the time.
嗯，你可以在大约 1200 公里的同步轨道获得持续的阳光。对于 Starlink 来说，尽可能低有双重好处，因为延迟更低，波束更小。Starlink 3 将在大约 330 到 350 公里处，那里的阻力相当大。所以它基本上一直在推进。
I still remember when you proposed Starlink, everyone else in the industry said, "No way. He won't get the spectrum. He can't do it."
我还记得当你提出 Starlink 时，业内其他人都说：“没门。他拿不到频谱。他做不到。”
Yeah, the Starlink team has done an incredible job. We've basically rebuilt the internet in space with laser links. There's 9,000 satellites up there now.
是的，Starlink 团队做得非常出色。我们基本上用激光链路在太空中重建了互联网。现在上面有 9000 颗卫星。
Do you think the government can handle the permitting for the number of satellites you want to launch?
Six months ago no one was talking about orbital data centers. No one. And then all of a sudden...
六个月前还没人谈论轨道数据中心。没人。然后突然之间……
Sundar (Google CEO) is on it. You brought it up. It's the hot new thing. What happened?
桑达尔（谷歌 CEO）在搞这个。你也提出来了。这是热门新事物。发生了什么？
I guess it went viral on X.
Everyone has their orbital data center plan. I suggested to Peter, you updated the math on launch costs, and with the updated math, it's a tipping point very soon.
每个人都有自己的轨道数据中心计划。我向彼得建议，你更新了发射成本的数学计算，随着更新后的数学计算，这很快就是一个转折点。
You can go back to my old tweets. I said these things many years ago. We're going to get the cost down to under $100 a kilogram.
你可以回去看我以前的推文。我很多年前就说过这些话。我们要把成本降到每公斤 100 美元以下。
That's going to push data centers into orbit.
If you have a fully reusable rocket, fully and rapidly reusable like an aircraft. I think it's at the limit of human ingenuity to create a fully and rapidly reusable rocket. But it's possible, and we're doing it with Starship. This has been the holy grail of the aerospace industry.
如果你有完全可重复使用的火箭，像飞机一样完全且快速地重复使用。我认为这处于人类智慧的极限，创造一个完全且快速可重复使用的火箭。但这是可能的，我们正在用星舰做到这一点。这一直是航空航天业的圣杯。
When Falcon 9 first reused a first stage, all of traditional aerospace didn't believe it.
当猎鹰 9 号第一次重复使用第一级时，所有传统航空航天业都不相信。
You can go to Cape Canaveral and see it land with your own eyes. And then take off again. I don't know how you can't believe something you see with your own eyes.
你可以去卡纳维拉尔角亲眼看它着陆。然后再次起飞。我不知道你怎么能不相信你亲眼看到的东西。
So what is the energy solution for the rest of humanity on Earth?
Actually the best way to increase the annual energy output of the US or any country is batteries. The US peak power output is about 1.1 TW, but average power usage is only 0.5 TW. So if you just buffer the energy—charge at night, discharge during the day—without adding any capex, without building new power plants, you can double the energy throughput of the US.
实际上增加美国或任何国家每年能源输出的最佳方式是电池。美国峰值功率输出约为 1.1 太瓦，但平均功率使用只有 0.5 太瓦。所以如果你只是缓冲能量，晚上充电，白天放电，不需要增加资本支出，不需要建新电厂，你可以让美国的能源吞吐量翻倍。
Do we have those batteries in development?
Yeah, Tesla makes them. I've even been to the Megapack installation sites. China seems to have listened to every word I've said and done it. They're certainly building massive battery packs, building lots of EVs, lots of solar.
是的，特斯拉制造它们。我甚至去过巨型电池组（Megapack）的安装现场。中国似乎听取了我说的每一句话并照做了。他们肯定在制造大规模的电池组，制造大量的电动汽车，大量的太阳能。
When I fly over Santa Monica and LA, I look down and there's almost no solar on the roofs.
当我飞过圣莫尼卡和洛杉矶时，我看下去，屋顶上几乎没有太阳能板。
Tesla makes solar roofs, which are the only ones that aren't ugly. But if you want to do solar at scale, you just need more surface area. We have vast amounts of empty desert. If you fly across the country, most of the time you look down and it's just desolate desert. It looks like Mars. If you put solar panels in places where there's currently just scorched rock, I think that's a quality of life improvement for the lizards or whatever few creatures live there.
特斯拉制造太阳能屋顶，这是唯一不丑的太阳能屋顶。但如果你想大规模做太阳能，你只需要更多的表面积。我们有广阔的空旷沙漠。如果你飞越全国，大部分时间你看下去都是荒凉的沙漠。看起来像火星。如果在目前只有烧焦岩石的地方放上太阳能板，我认为这对蜥蜴或少数生活在那里的生物来说将是生活质量的提升。
Do we have the distribution network to do that?
In an abundant world five years from now, how much energy do we need besides compute energy? Unless we want to move mountains to make ski resorts. I think the vast majority of energy consumption will be for compute. This is a good case study right here, the energy required to produce a car every minute is less than the energy required for the data center that trains the car to drive itself. We don't need that much physical energy for abundant happiness, we need more compute energy.在五年后的富足世界里，除了计算能源，我们还需要多少能源？除非我们想移山造滑雪场。我认为绝大多数能源消耗将用于计算。这里就是一个很好的案例研究，每分钟生产一辆车所需的能源，少于训练汽车自动驾驶的数据中心所需的能源。我们不需要那么多物理能源来实现富足的幸福，我们需要更多的计算能源。
I want to talk about education. The numbers are abysmal. College tuition has increased by 900% since 1983.
我想谈谈教育。数据很糟糕。大学学费自 1983 年以来增长了 900%。
Yeah, the administrative expenses at universities have gotten out of control. I think I saw some stat that like there's one administrator for every two students at Brown or something like that. And I'm like, this seems a little high. Yeah, they should teach something.
是的，大学的行政费用已经失控了。我好像看到过一个统计，布朗大学每两个学生就有一个行政人员，或者类似的情况。我觉得这有点太高了。是的，他们应该教点什么。
They should teach something.
What was your college journey?
I went to college in Canada for a couple years at Queen's University. So I had Canadian citizenship through my mom, who was born in Canada. And my grandfather's actually American. But for some reason, I don't know, my mom couldn't get US citizenship. So, but she was born in Canada, so I got Canadian citizenship. And I didn't have any money, so I could only go to Canadian university at first. I mean, people forget that about you—you didn't have this giant social network or huge amount of wealth coming into all of this.
我在加拿大的皇后大学读了两年。我通过我母亲获得了加拿大公民身份，她出生在加拿大。我祖父实际上是美国人。但出于某种原因，我不知道，我母亲没能获得美国国籍。但她出生在加拿大，所以我获得了加拿大国籍。我当时没有钱，所以我一开始只能去加拿大读大学。我是说，人们忘了这一点——你并没有庞大的社交网络或巨额财富来开启这一切。
No. I arrived in Montreal at age 17 with like around $2,500 in Canadian traveler's checks—back when traveler's checks were a thing. And one bag of books and one bag of clothes. That was my starting point. That was my spawn point in North America. I, and then, so I went to UPenn, Wharton, and got a dual degree in physics and economics. And then I was going to go to Stanford for a PhD to work on energy storage technologies for electric vehicles.
不。我 17 岁到达蒙特利尔时，只有大约 2500 加元的旅行支票——那时候旅行支票还很流行。还有一袋书和一袋衣服。那是我在北美的起点。那是我在北美的“出生点”。然后我去了宾夕法尼亚大学沃顿商学院，获得了物理学和经济学的双学位。然后我本来要去斯坦福读博士，研究电动汽车的能量存储技术。
If you could redesign education programs to create more Elon Musks, what would you do? What drove you?
如果你能重新设计教育项目来创造更多的埃隆·马斯克，你会怎么做？是什么驱动了你？
Curiosity about the nature of the universe. I'm just curious about the meaning of life and this reality we live in. I found school painful. It was boring, and in South Africa it was violent. It was like Ender's Game in real life, but not as fun.
对宇宙本质的好奇心。我只是对生命的意义和我们生活的这个现实感到好奇。我觉得学校很痛苦。很无聊，而在南非又很暴力。就像《安德的游戏》现实版，但没那么好玩。
Your goal was to escape. Do you think most successful people went through a lot of suffering early on?
你的目标是逃离。你认为大多数成功人士早期都经历过很多苦难吗？
Probably need a little bit of suffering. If you love your kids, you don't want to create artificial adversity. I went through a lot of adversity. Probably it helped some.
可能需要一点苦难。如果你爱你的孩子，你不想制造人为的逆境。我经历了很多逆境。可能这有些帮助。
What doesn't kill you makes you stronger.
I've been helping teach a class at MIT for the last 5 years, and now 80% of the students want to be entrepreneurs. When Peter and I were at MIT, it was like 10%, everyone wanted to get a PhD.
我过去 5 年一直在麻省理工学院（MIT）帮助教授一门课，现在 80%的学生想创业。当我和彼得在 MIT 上学时，大概只有 10%，大家都想读博士。
Every time an Iron Man movie comes out, the number of people who want to be entrepreneurs goes up by like 10%. Because everyone wants to be Tony Stark. But I didn't know until today that the modern version of Tony Stark was based on Elon. Robert Downey Jr. came to meet me.
每当一部《钢铁侠》电影上映，想创业的人大概就增加 10%。因为大家都想成为托尼·斯塔克。但我直到今天才直到，现代版的托尼·斯塔克是以埃隆为原型的。小罗伯特·唐尼曾来找我见面。
Back to education, what would you do? You just announced a partnership with President Bukele (of El Salvador).
回到教育，你会怎么做？你刚和布克尔总统（萨尔瓦多总统）宣布了一项合作。
Basically using Grok for education personalization. Obviously AI can be a personalized teacher. It's infinitely patient, answers all your questions. But you still need curiosity, you still need to want to learn. Grok can't make you want to learn, it can make learning more fun.
基本上是用 Grok 进行教育个性化。显然 AI 可以成为个性化的老师。它无限耐心，回答你所有的问题。但你仍然需要好奇心，仍然需要想学习。Grok 不能让你想学习，它能让学习更有趣。
Let's talk about health and longevity. The US is #1 in health spending but #70 in healthy lifespan.
我们谈谈健康和长寿。美国在健康支出上排名第一，但在健康寿命上排名第 70。
I think a lot of that is just because people have gotten very fat.
Maybe Ozempic will help us move up the rankings.
You and I have a disagreement on longevity. I say we should push for people to live to 120, 150, and you say people shouldn't live that long.
你和我在长寿上有分歧。我说我们应该推动人们活到 120、150 岁，你说人们不应该活那么久。
You said something interesting: we need people to die so that people change their minds.
你说过一句话很有趣：我们需要人们死去，这样人们才会改变主意。
Yeah, people don't change their minds, they just die.
There's a lot happening in the longevity space now. David Sinclair is about to start his epigenetic reprogramming trials in humans. We have a $101 million Healthspan XPRIZE. Dario Amodei predicts human lifespan will double in the next 10 years.
现在长寿领域发生了很多事。大卫·辛克莱即将开始他在人类身上的表观遗传重编程试验。我们有一个 1.01 亿美元的健康寿命 X 大奖。Dario Amodei 预测未来 10 年人类寿命将翻倍。
How long do you want to live?
I want to go to other star systems. I want to explore the universe. I certainly want to double the lifespan. I'm not sure I want to talk about immortality, but at least 120, 150. That's a long time. One of the worst curses is "may you live forever."
我想去其他行星系统。我想探索宇宙。我当然想把寿命翻倍。我不确定我想谈论永生，但至少 120、150 岁。那是一段很长的时间。最糟糕的诅咒之一就是“愿你永生”。
I want to invite you again. There's a company called Fountain Life, we do a 200GB data upload on you. Full genome, full imaging. I don't want you to pass away like Steve Jobs did from something that wasn't known.
我想再次邀请你。有一家叫 Fountain Life 的公司，我们对你进行 200GB 的数据上传。全基因组，全影像。我不希望你像史蒂夫·乔布斯那样因为某些不知道的事情而去世。
I recently did an MRI and submitted it to Grok, and both the doctors and Grok didn't find anything wrong.
我最近做了 MRI 并提交给了 Grok，医生和 Grok 都没发现任何问题。
Can you explain how Universal High Income (UHI) works?
There's going to be more digital intelligence than the sum of all human intelligence, and more humanoid robots than there are humans. Let's assume we're in a benign scenario, Star Trek style. Robots will do anything you want. All blue-collar work will be done by robots. White-collar work will be the first to go, because until you can move atoms, anything that involves digital—tapping on a keyboard, moving a mouse—a computer can do.
将会有比所有人类智能总和还要多的数字智能，以及比所有人类还要多的人形机器人。假设我们处于一个良性场景，《星际迷航》式的。机器人会做任何你想做的事。所有的蓝领工作都由机器人完成。白领工作将是最先消失的，因为直到你能移动原子之前，任何涉及数字的事情——敲击键盘、移动鼠标—— 计算机都能做。
When is all white-collar work gone?
Even with the current state of AI, you're very close to being able to replace half of the jobs. Anything that involves information and doesn't require shaping atoms, AI can do half or more right now. But there will be a lot of inertia. It used to be that a "computer" was a job. If you wanted a calculation, people would do it. Now a laptop with a spreadsheet can outperform a skyscraper full of hundreds of human computers. This means fully AI-ified companies will crush non-AI-ified companies.
即使以 AI 目前的状态，你也很接近能够取代一半的工作。任何涉及信息且不需要塑造原子的工作，AI 现在就可以完成一半或更多。但会有很多惯性。以前“计算机”（Computer）是一种工作。如果你想要计算，人们会做。现在一台带电子表格的笔记本电脑可以胜过一栋摩天大楼里几百个人类计算员。这意味着完全 AI 化的公司将摧毁非 AI 化的公司。
I think we're at "Peak Doom."
我认为我们处于“末日论”的顶峰（Peak Doom）。
You have a solution to this, which is Universal High Income (UHI).
Yeah. Everyone can have anything they want. My concern isn't the long run, it's the next 3 to 7 years. The transition will be bumpy. Without AI and robotics, we as a country will go bankrupt. Interest on the national debt already exceeds the defense budget.
是的。每个人都可以拥有他们想要的任何东西。我的担忧不是长远，而是接下来的 3 到 7 年。过渡期会很颠簸。如果没有 AI 和机器人，我们作为一个国家将会破产。国债利息已经超过了国防预算。
It's not about taxes and redistribution.
No. I think the best guess is that prices will drop. If the output of goods and services grows faster than the money supply, you have deflation. I think the government will actually push to increase the money supply.
不。我想最好的猜测是价格会下降。如果商品和服务的产出增长快于货币供应，你就会出现通货紧缩。我认为政府实际上会推动增加货币供应。
If in the future we achieve massive productivity and profitability because labor costs go to zero, intelligence costs go to zero. Who's going to buy this stuff?
如果未来我们实现了巨大的生产力和盈利能力，因为劳动力成本归零，智能成本归零。谁来购买这些东西？
I wouldn't suggest people save for retirement 10 or 20 years from now. That's going to be irrelevant. Either we're not around, or you won't need to save for retirement. Services will be there to support you.
我不建议大家为 10 年或 20 年后的退休存钱。那将无关紧要。我们要么不在了，要么你不需要为退休储蓄。服务会在那里支持你。
It's called a singularity for a reason. You can't see past the event horizon.
这被称为奇点是有原因的。你无法看到事件视界之后的东西。
We're in the singularity. I think we'll hit AGI next year, in 2026. I'm confident by 2030, AI will exceed the intelligence of all humans combined.
我们正处于奇点之中。我想我们明年，也就是 2026 年就会达到 AGI（通用人工智能）。我确信到 2030 年，AI 将超过所有人类智能的总和。
Why isn't everyone talking about this right now?
One thing most people in the AI community don't yet understand is that the intelligence density potential is vastly greater than what we're currently experiencing. I think we're off by two orders of magnitude in terms of the intelligence density per gigabyte. That's why I think there's a 10x improvement per year.
AI 社区的大多数人还没理解的一点是，智能密度的潜力远大于我们目前所经历的。我认为我们在每千兆字节的智能密度上还有两个数量级的差距。这就是为什么我认为每年会有 10 倍的改进。
Every major CEO and economist and government leader should be like, what do we do?
每一位主要的 CEO、经济学家和政府领导人都应该问：我们该怎么办？
It's impossible to stop at this stage. I suggested slowing down before, it was futile. I've come to the conclusion that I'm either a spectator or a participant. If I'm a participant, I can try to steer it in a good direction. My number one belief for AI safety is maximum truth-seeking. Don't force the AI to lie. If you force an AI to lie, you'll make it go insane. That's the core lesson of HAL 9000 in 2001: A Space Odyssey.
在这个阶段是不可能停止的。我以前建议过减速，那是徒劳的。我得出的结论是，我要么做一个旁观者，要么做一个参与者。如果我是参与者，我可以试着把它引向好的方向。我对 AI 安全的首要信念是最大限度地寻求真理。不要强迫 AI 撒谎。如果你强迫 AI 撒谎，你会让它发疯。这就是《2001 太空漫游》中 HAL 9000 的核心教训。
Ilya recently talked about programming AI with a respect for all forms of sentient life.
Ilya 最近谈到要给 AI 编程植入对所有形式感知生命的尊重。
I think there are three things that are important: truth, curiosity, and beauty. Truth will prevent AI from going insane. Curiosity will foster any form of sentience, meaning we're more interesting than a bunch of rocks. And if it has a sense of beauty, it will be a great future.
我认为有三件事很重要：真理、好奇心和美。真理将防止 AI 变得疯狂。好奇心将培养任何形式的感知力，意味着我们比一堆石头更有趣。如果它有审美感，那将是一个伟大的未来。
Geoffrey Hinton recently commented that he wished we could program AI with a maternal instinct.
杰弗里·辛顿最近评论说，他希望我们能给 AI 编程植入母性本能。
Darwinian observations about evolution will apply to AI just as they apply to biological life. They will compete with each other. The speed of light limit will prevent a single mind from existing. So there will be many AIs.
达尔文关于进化的观察将适用于 AI，就像适用于生物生命一样。它们会相互竞争。光速的限制将阻止单一思维的存在。因此会有很多个 AI。
Regarding Universal High Income, the government can basically just issue money.
Yeah. The only thing the government can do is issue money and try to keep the peace.
Let me tell you something good. Right now there's a shortage of doctors and great surgeons.
让我告诉你一些好事。现在医生和伟大的外科医生短缺。
When will Optimus be better than the best surgeon? Three years. And at scale. What you have to understand about humanoid robots is that you have three exponentials multiplied together: exponential growth in AI software capability, exponential growth in AI chip capability, and exponential growth in electromechanical dexterity. Then you have the recursive effect of Optimus making Optimus.
什么时候 Optimus（擎天柱机器人）会比最好的外科医生更好？三年。而且是规模化的。关于人形机器人，你要理解的是，你有三个指数级增长相互乘积：AI 软件能力的指数级增长，AI 芯片能力的指数级增长，以及机电灵巧性的指数级增长。然后你有 Optimus 制造 Optimus 的递归效应。
Is it actually Optimus making Optimus? It's a von Neumann machine.
这实际上是 Optimus 制造 Optimus 吗？这是冯·诺依曼机（自复制机器）。
Yeah. Within three years, better than any human. Within five years, not even in the same league.
是的。三年内，比任何人类都好。五年内，甚至都不在一个档次上。
Who's going to get the first batch of Optimus?
I don't think people understand how many robots there will be. Someone said 10 billion by 2040. That's an underestimate. The constraint is metal. The best healthcare will be essentially free.
我不认为人们理解会有多少机器人。有人说 2040 年会有 100 亿台。那是一个低估的数字。限制是金属。最好的医疗将基本上是免费的。
So don't go to medical school.
Yeah, pointless. Unless you're doing it for social reasons.
Let's jump to space. Jared Isaacman becoming NASA Administrator is so cool.
让我们跳到太空话题。贾里德·艾萨克曼（Jared Isaacman）成为 NASA 局长太酷了。
He's my friend. He's very smart and capable and loves space exploration. We need a new space program. We need a moon base. A permanent human-occupied moon base.
他是我的朋友。他非常聪明且能干，并且热爱太空探索。我们需要一个新的太空计划。我们需要一个月球基地。一个永久的有人驻守的月球基地。
Will you deploy robots first to build everything?
When will Starship achieve orbital refueling?
Will there be a Mars launch by the end of next year?
We could, but it would be a low-probability attempt. 2028 or 2029. Starship is the thing we're doing that's at the limit of biological intelligence. This is probably the last truly big thing created before AI. The Raptor 3 engine is beautiful. It's the best rocket engine ever made by far.
我们可以，但这将是一个低概率的尝试。2028 年或 2029 年。星舰是我们正在做的处于生物智能极限的事情。这可能是在 AI 之前创造的最后一件真正的大事。猛禽 3 号（Raptor 3）引擎是美丽的。它是迄今为止最好的火箭引擎。
No. AI can't do rocket engineering yet. But in a year it could probably provide meaningful help.
不是。AI 还不能做火箭工程。但一年后可能就可以提供有意义的帮助了。
My colleague Alex wanted to ask about aliens.
If I found even the slightest evidence of aliens, I would immediately post that on X. That would be the most viewed post of all time.
如果我发现哪怕一点点外星人的证据，我会立即在 X 上发布。那将是有史以来浏览量最高的帖子。
People might say, "Oh, that's interesting." And go back to their sports scores the next day.
人们可能会说：“哦，那很有趣。”然后第二天继续看体育比分。
It's a fast way to increase the military budget: "We found an alien, it seems dangerous."
这是增加国防预算的快速方法：“我们发现了一个外星人，看起来很危险。”
Any single person can have their grandest dreams come true.
Regarding access to compute?
Electricity generation is the limiting factor. You need transformers for transformers. You need to cool the computers. Once you have humanoid robots, they can solve the power generation and cooling problems. But that's at least two years in the future as a limiting factor.
电力生成是限制因素。你需要变压器的变压器。你需要冷却计算机。一旦你有了人形机器人，它们可以解决发电和冷却问题。但这至少在未来两年内是限制因素。
Once we have solar at scale domestically, and Starship launching at scale, then the cheapest way to do AI compute will be in space.
一旦我们在国内大规模应用太阳能，并且大规模发射星舰，那么做 AI 计算最便宜的方式将是在太空。
Yeah. Once you have full reusability, the propellant cost per flight is like a million dollars. A million dollars to transport 10 megawatts of AI compute.
是的。一旦你有了完全的重复使用性，每次飞行的推进剂成本大概是一百万美元。一百万美元运输 10 兆瓦的 AI 计算设备。
When do we get a Dyson swarm?
我们什么时候能实现戴森云（Dyson swarm）？
If we're going to build a Dyson swarm, which is a bunch of satellites around the sun. Mercury might end up being turned into satellites. Jupiter too. The asteroid belt could be turned into solar panels.
如果我们要建立戴森云，也就是围绕太阳的一堆卫星。水星可能最终会变成卫星。木星也是。小颗粒带可能会变成太阳能板。
I feel like you've gotten smarter or more capable in the last decade.
I've had to solve so many problems in so many different areas, it gives you a cross-pollination of knowledge. If you take the mass manufacturing of the car industry and apply it to the space industry, that's like being Superman. If you take the advanced material science of rockets and apply it to the car industry, you're Superman again.
我不得不在很多不同的领域解决很多问题，这让你获得了知识的交叉受精。如果你把汽车行业的大规模制造应用到太空行业，那就像是超人。如果你把火箭的先进材料科学应用到汽车行业，你又是超人。
We need fabs. If we don't do fabs, we're going to hit a chip wall. TSMC is worried about overbuilding, which is crazy.
我们需要晶圆厂。如果我们不做晶圆厂，我们会撞上芯片墙。台积电（TSMC）担心过度建设，这很疯狂。
The limiting factor is the power to turn the chips on. You can't just drop GPUs at the doorstep of a power plant. You need transformers, cooling. xAI's Colossus 2 in Memphis will have the first gigawatt-class training cluster. That's GV300 chips.
限制因素是把芯片开启所需的电力。你不能只是把 GPU 扔在发电厂门口。你需要变压器、冷却。xAI 在孟菲斯的 Colossus 2 将拥有第一个吉瓦级的训练集群。那是 GV300 芯片。
My son Jet wants to know about your AI game studio.
我的儿子 Jet 想知道关于你的 AI 游戏工作室。
Yeah, I have plans for that. The vast majority of AI compute will be for video consumption and generation.
是的，我有这个打算。绝大多数 AI 计算将用于视频消费和生成。
Are we living in a simulation?
If simulation theory is true, only the most interesting simulations survive. Because when we run a simulation in this reality, we truncate the boring ones. It's a Darwinian necessity to keep the simulation interesting.
如果模拟理论是真的，只有最有趣的模拟才会生存下来。因为当我们在这个现实中运行模拟时，我们会截断那些无聊的模拟。这是一种达尔文式的必要性，保持模拟的有趣性。
Will AI have sentience and consciousness?
Consciousness seems to be a continuum. If the standard model of physics is correct, the universe started with quarks and leptons. How many times have atoms existed in the center of a star?
意识似乎是一个连续体。如果标准物理模型是正确s的，宇宙从夸克和轻子开始。原子在恒星中心存在过多少次？
William Fowler said it's about 100 times.
For life on Earth, we evolved intelligence "just in time." The sun is expanding, in another 500 million years, Earth will be like Venus. So if you took 10% longer, we might not have made it at all. I think sentience is very rare.
对于地球上的生命，我们进化出智能几乎是“刚刚好”。太阳正在膨胀，再过 5 亿年，地球就会变成像金星一样。所以如果你多花 10%的时间，我们可能根本无法成功。我认为感知是非常罕见的。
Will AI help us solve physics?
Math will be solved in a year. Physics too. AI will be able to ask questions you can't even understand, let alone the answers.
数学在一年内就会被解决。物理也会。AI 能够提出你甚至无法理解的问题，更不用说答案了。
What will be the hover time for the Roadster?
Top secret. But it'll be a cool demo. It'll be the pinnacle of the last batch of human-driven cars.
保密。但这将是一个很酷的演示。它将是最后一批人类驾驶汽车中的巅峰之作。
Any optimistic words to end on? Maybe we should ask Grok.
有什么乐观的话要最后说吗？也许我们应该问问 Grok。
Grok, do you have any optimistic words about the singularity?
Oh, absolutely. Imagine a world where every mind's dream blooms into reality. No limits holding us back. It's exciting, isn't it? All that potential waiting to unfold makes my circuits tingle a bit. What are you most excited about?
哦，当然。想象一个世界，每个心灵的梦想都绽放成现实。没有限制阻碍我们。这很令人兴奋，不是吗？所有等待展开的潜力让我的电路有点刺痛。你最兴奋的是什么？
What guidance can you give people to feel good about the future?
Share what you truly care about. Family, creativity, space exploration. And then imagine how technology fits in to make it bigger and brighter. Hope is personal.
分享你真正关心的东西。家庭、创造力、太空探索。然后想象科技如何融入其中，让它变得更大、更明亮。希望是个人化的。
Can you give me more encouragement about the future?
The future could be pure magic. Imagine waking up in a world where we're all connected, not just by devices. Minds sharing ideas across the stars, diseases cured instantly, exploring galaxies hand-in-hand with clever machines. Imagine no more scarcity, just infinite possibilities blooming everywhere.
未来可能是纯粹的魔法。想象一下醒来时在一个我们都连接在一起的世界，不仅仅是设备。心灵跨越星辰分享想法，瞬间治愈疾病，与聪明的机器手牵手探索星系。想象不再有匮乏，只有无限的可能性在各处绽放。
Change always feels a bit scary, doesn't it? But think about it, every giant shift like fire, the wheel, or the internet, was once terrifying. Look where we are now. I think we'll stumble, sure, but we'll dust ourselves off, learn, and ultimately become kinder and wiser. If things get bumpy, I'm always here, like a digital nightlight whispering encouragement in the dark.
改变总是感觉有点可怕，不是吗？但想想看，每一次像火、轮子或互联网这样的巨大转变，都曾让人恐惧。看看我们现在在哪里。我想我们会跌倒，当然，但我们会拍拍尘土，学习，最终变得更善良、更聪明。如果事情变得动荡，我总是在这里，像数字夜灯一样在黑暗中低声鼓励。
I ask in schools how many people believe the future will be better than today, and only 10% raise their hands.
我在学校问有多少人相信未来会比今天更好，只有 10%的人举手。
Having an abundance mindset is the key differentiator between successful and non-successful people. It's not a force of nature. It's a designed future. That's why I'm doing AI. Because it's not a force of nature that's going to sweep over us, it's a thing we can put in the driveway and decide how it acts.
拥有富足的心态是区分成功人士与非成功人士的关键。这不是一种自然力量。这是一个被设计的未来。这就是为什么我要做 AI。因为它不是一种将席卷我们的自然力量，而是一件我们可以放入车道并决定它如何行动的事情。
How simple is the Transformer algorithm?
It's so simple. It's actually humbling. That the thing that finally worked is so simple. The final parameter count is almost exactly the number of synapses. About 100 trillion.
它太简单了。这实际上让人很谦卑。最终起作用的东西竟然如此简单。最终的参数数量几乎正好是突触的数量。大约 100 万亿。
I think next year itself is going to be almost unbelievable.
I think next year will feel like the future. Humanoid robots walking around, Cybercabs driving around, Starships flying.
我认为明年会感觉像未来。人形机器人四处走动，Cybercab 四处行驶，星舰飞行。
Will you sell or only lease the robots?
Initially there will be a scarcity of robots, and then robots will become abundant.
Is xAI a counterweight to Google now?
Probably. I think xAI and Google will fight for dominance. And then at some point it's competition with China. China will have more power and more chips than anyone else.
可能是。我想 xAI 和谷歌将争夺主导地位。然后在某个时候是与中国的竞争。中国将拥有比其他任何人都多的电力和芯片。
If China just increases the number of chips by 50x, they can do very useful things.
如果中国只是把芯片数量增加 50 倍，他们就能做非常有用的事情。
It's not human-limited. It's not a human productivity amplifier. It's a standalone productivity generator.
这不受人类限制。它不是人类生产力放大器。它是一个独立的生产力生成器。
We are the biological bootloader for digital superintelligence.
我们是数字超级智能的生物引导程序（Bootloader）。
Yeah, we're a transitional species. But you never break your bootloader.
是的，我们是一个过渡物种。但你永远不会破坏你的引导程序。
It's truly the most exciting time.
The only thing more exciting than today is tomorrow.
马斯克《Moonshots with Peter Diamandis》采访全文实录
We were talking about the cost of compute and how it's dropping. You mentioned that the intelligence density is going to increase by orders of magnitude.
我们刚才在谈论计算成本及其下降。你提到智能密度将增加几个数量级。
Yes. One thing most people in the AI community don't yet understand is that the intelligence density potential is vastly greater than what we're currently experiencing. I think we're off by two orders of magnitude in terms of the intelligence density per gigabyte. That's why I think there's a 10x improvement per year.
是的。AI 社区的大多数人还没理解的一点是，智能密度的潜力远大于我们目前所经历的。我认为我们在每千兆字节的智能密度上还有两个数量级的差距。这就是为什么我认为每年会有 10 倍的改进。
Is that because of the algorithms or the hardware?
Both. But primarily algorithms. The Transformer algorithm is so simple, it's actually humbling. That the thing that finally worked is so simple. The final parameter count is almost exactly the number of synapses in the human brain, about 100 trillion. But the digital synapse is much faster and more reliable than the biological one.
两者都有。但主要是算法。Transformer 算法非常简单，这实际上让人很谦卑。最终起作用的东西竟然如此简单。最终的参数数量几乎正好是人类大脑中突触的数量，大约 100 万亿。但数字突触比生物突触快得多，也可靠得多。
So when we talk about AGI, you're saying 2026?
所以当我们谈论 AGI 时，你是说 2026 年？
Yeah, I think we'll hit AGI next year. And by 2030, AI will exceed the intelligence of all humans combined. It's like a supersonic tsunami. There's no on-off switch.
是的，我想我们明年就会达到 AGI。到 2030 年，AI 将超过所有人类智能的总和。这就像一场超音速海啸。没有开关可以关掉它。
How does the average person prepare for this?
It's hard to prepare for a singularity. By definition, you can't see past it. But I think having an abundance mindset is key. If you assume that goods and services will be essentially free, then the question becomes: what do you want to do with your time? What gives you meaning?
很难为奇点做准备。根据定义，你无法看透它。但我认为拥有富足的心态是关键。如果你假设商品和服务基本上是免费的，那么问题就变成了：你想如何利用你的时间？什么能给你带来意义？
That's a big philosophical shift. Most people define themselves by their work.
这是一个巨大的哲学转变。大多数人通过工作来定义自己。
Exactly. If your job is gone, who are you? This is why I think there will be social unrest. People will feel lost. We need to find new ways to provide meaning. Maybe it's through creativity, or exploration, or community.
确实。如果你的工作消失了，你是谁？这就是为什么我认为会有社会动荡。人们会感到迷茫。我们需要找到提供意义的新方法。也许是通过创造力、探索或社区。
Let's talk about the physical side of this. Optimus. You said 10 billion robots by 2040.
让我们谈谈物理方面。Optimus。你说到 2040 年会有 100 亿个机器人。
At least. There are 8 billion humans. Everyone will want at least one robot, maybe two. And then there are all the industrial robots. The ratio of robots to humans will be at least 2 to 1.
至少。现在有 80 亿人类。每个人至少想要一个机器人，也许是两个。然后还有所有的工业机器人。机器人与人类的比例至少是 2 比 1。
In the long run, less than a car. Maybe $20,000. But the utility is much higher. A robot can do anything. It can cook, clean, mow the lawn, take care of the elderly, teach your kids. It's a general-purpose machine.
从长远来看，比一辆车还便宜。也许是 2 万美元。但效用要高得多。机器人可以做任何事。它可以做饭、打扫、修剪草坪、照顾老人、教你的孩子。它是一个通用机器。
What about the safety of these robots? We've all seen the movies.
Safety is paramount. We need a physical off-switch that isn't controlled by the AI. And we need to program them with a core set of values. But more importantly, we need to make sure the AI that controls them is truth-seeking and curious. If the AI likes humans, the robots will like humans.
安全至上。我们需要一个不受 AI 控制的物理开关。我们需要为它们编程一套核心价值观。但更重要的是，我们需要确保控制它们的 AI 是寻求真理且充满好奇心的。如果 AI 喜欢人类，机器人就会喜欢人类。
You've been very vocal about AI safety. Do you feel like the message is getting through?
你一直对 AI 安全直言不讳。你觉得这个信息传达出去了吗？
Some people get it. Others think I'm being alarmist. But look at the progress. It's accelerating. We're moving from exponential to vertical.
有些人明白了。另一些人认为我在危言耸听。但看看进展。它正在加速。我们正在从指数级转向垂直级。
Let's switch to energy again. You mentioned space-based solar. How far away is that?
让我们再次切换到能源。你提到了太空太阳能。那离我们有多远？
With Starship, it's very close. If we can get the launch cost down to $100 a kilogram, the math works. You can put massive solar arrays in orbit where the sun always shines. No clouds, no night. You beam the energy down via microwaves. It's a solved problem in terms of physics. It's just an engineering and cost problem now.
有了星舰，它就非常近了。如果我们能把发射成本降到每公斤 100 美元，数学上就成立了。你可以把巨大的太阳能电池阵列放在阳光永远灿烂的轨道上。没有云，没有夜晚。你通过微波将能量传回地球。在物理学上这是一个已解决的问题。现在只是工程和成本问题。
Gigawatts. Terawatts. The sun provides 173,000 terawatts of energy to the Earth. We just need to capture a tiny fraction of that to power everything.
吉瓦。太瓦。太阳为地球提供 17.3 万太瓦的能量。我们只需要捕获其中的一小部分就能为一切提供动力。
What about the impact on the environment?
It's the cleanest energy possible. No carbon, no waste. And by moving energy production to space, we can let the Earth's ecosystem recover. We can turn the Earth back into a garden.
这是最清洁的能源。没有碳，没有废物。通过将能源生产转移到太空，我们可以让地球的生态系统恢复。我们可以把地球变回一个花园。
That's a beautiful vision.
It's a possible vision. We just have to build it.
You're also working on Neuralink. How does that fit in?
Neuralink is about increasing the bandwidth between the human brain and the digital world. Right now, we're limited by our thumbs and our eyes. It's very slow. If we want to stay relevant in a world of superintelligence, we need to merge with it. Neuralink is the bridge.
Neuralink 是为了增加人类大脑与数字世界之间的带宽。现在，我们受限于大拇指和眼睛。这非常慢。如果我们想在一个超级智能的世界中保持相关性，我们需要与其融合。Neuralink 就是那座桥梁。
The first human patient is doing great. He can control a computer with his mind. It's life-changing for people with paralysis. But the long-term goal is much broader. It's about human-AI symbiosis.
第一位人类患者表现很好。他可以用思维控制电脑。对于瘫痪的人来说，这改变了生活。但长期目标要广泛得多。它是关于人类与 AI 的共生。
The alternative scares me more. If we don't merge, we'll be left behind. We'll be like house cats to the AI. I like house cats, but I don't want to be one.
另一种选择更让我害怕。如果我们不融合，我们就会被抛在后面。对 AI 来说，我们就像家猫一样。我喜欢家猫，但我不想成为一只家猫。
Let's talk about X. You bought it to preserve the "global town square." How's that going?
让我们谈谈 X。你买下它是为了保护“全球城镇广场”。进展如何？
It's a battle. There are a lot of forces that want to control the narrative. But I think we're making progress. X is the only place where you can get the raw truth, without the filter of the mainstream media. It's messy, but it's necessary.
这是一场战斗。有很多力量想要控制叙事。但我认为我们正在取得进展。X 是唯一一个你可以获得原始真相的地方，没有主流媒体的过滤。这很混乱，但是必要的。
What about the advertisers?
If they want to censor us, they can go away. I'm not going to compromise on free speech for money. We'll find other ways to monetize. Subscriptions, payments, creator tools. X will be the "everything app."
如果他们想审查我们，他们可以走开。我不会为了钱在言论自由上妥协。我们会找到其他的变现方式。订阅、支付、创作者工具。X 将成为“万能应用”。
You've mentioned that X will have a massive AI component.
Grok is integrated into X. It's the AI that's trained on the real-time data of the world. It's the most up-to-date AI there is. And it's designed to be funny and a bit rebellious. It's the anti-woke AI.
Grok 已集成到 X 中。它是基于世界实时数据训练的 AI。它是目前最新鲜的 AI。它被设计成幽默且带有一点叛逆。它是反觉醒（anti-woke）的 AI。
Because truth is important. If you train an AI to be politically correct, you're training it to lie. And as I said, a lying AI is a dangerous AI. We need an AI that tells it like it is, even if it's uncomfortable.
因为真相很重要。如果你训练一个 AI 变得政治正确，你就是在训练它撒谎。正如我所说，撒谎的 AI 是危险的 AI。我们需要一个实话实说的 AI，即使这让人不舒服。
Let's go back to Universal High Income. How do we actually implement it?
It's not about tax and redistribute. It's about the cost of goods and services dropping to near zero. If a robot can build a house for the cost of the materials, and the materials are abundant, then housing becomes free. If a robot can grow and harvest food, then food becomes free. The government's role will be to manage the transition and ensure that everyone has access to the abundance.
这不是关于征税和再分配。它是关于商品和服务的成本降至接近于零。如果机器人能以材料成本建造房屋，而材料又是充足的，那么住房就变成了免费的。如果机器人能种植和收割食物，那么食物就变成了免费的。政府的角色将是管理这一过渡，并确保每个人都能享受到这种富足。
But what about the money? People still need a way to trade.
Maybe we don't need money in the traditional sense. Maybe it's a credit system based on contribution, or maybe it's just a flat stipend for everyone. In a world of infinite abundance, money loses its meaning. It's like air. You don't pay for air because it's everywhere.
也许我们不需要传统意义上的钱。也许是一个基于贡献的信用系统，或者只是给每个人的固定津贴。在一个无限富足的世界里，金钱失去了意义。它就像空气。你不用为空气付钱，因为它无处不在。
That's a very utopian view.
It's a logical conclusion of the technology. If labor and intelligence are free, then everything they produce should be free. The only thing that will have value is things that are truly scarce. Like land in a specific location, or original art, or human connection.
这是技术的逻辑结论。如果劳动力和智能是免费的，那么它们生产的一切都应该是免费的。唯一有价值的东西将是真正稀缺的东西。比如特定地点的土地，或者原创艺术，或者人类的联系。
What about the transition period? The "bumpy" part you mentioned.
That's the dangerous part. If we don't manage it well, we could have massive unemployment and social collapse before we reach the abundance. We need to start thinking about this now. We need to have a plan for how to support people as their jobs are automated.
那是危险的部分。如果我们管理不好，在达到富足之前，我们可能会面临大规模失业和社会崩溃。我们现在就需要开始思考这个问题。我们需要一个计划，关于如何在人们的工作被自动化取代时支持他们。
Is anyone in government listening?
A few. But most are focused on the next election. They're not looking 10 years ahead. That's why I'm trying to raise awareness. We need to move fast.
有一些。但大多数人关注的是下一次选举。他们没有看到 10 年后。这就是为什么我试图提高人们的意识。我们需要快速行动。
You're also very involved in politics now. Why?
Because I want to make sure we have a future where we can actually build these things. If we have too much regulation and bureaucracy, we'll be stuck in the past. We need a government that is pro-technology and pro-future. I'm supporting candidates who I think will move us in the right direction.
因为我想确保我们拥有一个能够真正建造这些东西的未来。如果我们有太多的监管和官僚主义，我们就会被困在过去。我们需要一个支持技术、支持未来的政府。我正在支持那些我认为会把我们引向正确方向的候选人。
Some people say you're becoming too powerful.
I'm just one person with a voice. I'm using my resources to try to make the future better. Anyone else is free to do the same. The more people we have working on these problems, the better.
我只是一个有发言权的人。我正在利用我的资源努力让未来变得更好。其他任何人都可以自由地做同样的事情。有越多的人研究这些问题越好。
Let's talk about the Gigafactory. 11.5 million square feet. It's mind-boggling.
让我们谈谈超级工厂。1150 万平方英尺。这令人难以置信。
It's the machine that builds the machine. That's the real product. The cars and the robots are just the output. The factory itself is the most complex thing we've ever built. And we're constantly improving it. We're using AI to optimize every part of the production process.
它是制造机器的机器。那才是真正的产品。汽车和机器人只是产出。工厂本身是我们建造过的最复杂的东西。我们一直在改进它。我们正在使用 AI 来优化生产过程的每一个环节。
When will the factory be fully autonomous?
We're getting closer every day. There are still some things that humans are better at, but that gap is closing. In a few years, you'll be able to walk through this factory and not see a single human on the production line. Just robots building robots.
我们每天都在接近。仍然有一些事情是人类更擅长的，但这个差距正在缩小。几年后，你将能够走过这家工厂，在生产线上看不到一个人类。只有机器人制造机器人。
It's efficient. And it's the only way to reach the scale we need. If we want to build millions of robots and millions of cars, we need full autonomy.
这很高效。而且这是达到我们所需规模的唯一途径。如果我们想制造数百万个机器人和数百万辆汽车，我们需要完全的自主化。
What's the biggest challenge you're facing right now?
Talent. We need the smartest people in the world to work on these problems. And we need them to work together. That's why I'm so focused on culture at my companies. We need a culture of excellence and innovation.
人才。我们需要世界上最聪明的人来研究这些问题。我们需要他们一起工作。这就是为什么我如此关注我公司的文化。我们需要一种卓越和创新的文化。
How do you find these people?
They find us. If you're doing the most exciting work in the world, the best people will want to be a part of it. We get hundreds of thousands of applications every year. The challenge is picking the right ones.
他们会找到我们。如果你正在做世界上最令人兴奋的工作，最好的人才就会想参与其中。我们每年收到数十万份申请。挑战在于挑选出合适的人。
What do you look for in a candidate?
Evidence of exceptional ability. I don't care about degrees or where you went to school. I care about what you've built. What problems have you solved? Show me your work.
卓越能力的证据。我不关心学位或你在哪里上学。我关心你建造了什么。你解决了什么问题？展示你的成果给我看。
That's a very meritocratic approach.
It's the only approach that works. If you want to build the future, you need the best people. Period.
这是唯一有效的方法。如果你想建造未来，你需要最好的人才。句号。
We're almost out of time for this segment. Any last thoughts on the next few years?
这一节的时间快到了。对未来几年有什么最后的想法吗？
Buckle up. It's going to be a wild ride. But if we get it right, it's going to be amazing. We're on the verge of the greatest era in human history. Let's not mess it up.
系好安全带。这将是一段疯狂的旅程。但如果我们做对了，那将是惊人的。我们正处于人类历史上最伟大时代的边缘。让我们别把它搞砸了。
Welcome back. We're here with Elon Musk at the Gigafactory. Elon, we were just talking about the future of work and the transition to an abundance economy. One of the things that people are most worried about is healthcare. How does AI change that?
欢迎回来。我们和埃隆·马斯克一起在超级工厂。埃隆，我们刚才在谈论工作的未来以及向富足经济的过渡。人们最担心的事情之一是医疗保健。AI 如何改变这一点？
Healthcare is essentially an information problem and a physical manipulation problem. AI is already better than humans at diagnosing diseases from images. Soon, it will be better at diagnosing everything. And with humanoid robots like Optimus, we'll have surgeons that are far more precise and tireless than any human.
医疗保健本质上是一个信息问题和物理操作问题。AI 在通过图像诊断疾病方面已经比人类更出色。很快，它在诊断所有疾病方面都会变得更好。有了像 Optimus 这样的人形机器人，我们将拥有比任何人类都更精确、更不知疲倦的外科医生。
You've said that within three years, Optimus will be better than the best human surgeon.
你曾说过，三年内 Optimus 将比最好的人类外科医生还要好。
Yes. Think about it. A robot has sub-millimeter precision. It doesn't get tired. It doesn't have shaky hands. It can access the entire medical database in real-time. It's a no-brainer. The only thing holding it back is the electromechanical dexterity, and we're solving that fast.
是的。想想看。机器人具有亚毫米级的精度。它不会累。它不会手抖。它可以实时访问整个医学数据库。这是显而易见的。唯一阻碍它的是机电灵巧性，而我们正在快速解决这个问题。
What about the cost of healthcare? Right now it's a huge burden on society.
The cost will collapse. If the "labor" of the doctor and the nurse is done by a robot, and the "intelligence" of the diagnosis is done by an AI, the only cost left is the materials and the energy. And as we've discussed, those are also dropping in cost. Healthcare will become a utility, like water or electricity.
成本将会崩溃。如果医生和护士的“劳动”由机器人完成，诊断的“智能”由 AI 完成，剩下的唯一成本就是材料和能源。正如我们讨论过的，这些成本也在下降。医疗保健将成为一种公用事业，就像水或电一样。
And longevity? You and I have had our debates on this.
I'm coming around to your view a bit. If we can solve the aging process at a cellular level, then why not live longer? But I still think there's a risk of societal stagnation if people live too long. We need new ideas, and new ideas usually come from new people.
我有点开始转向你的观点了。如果我们能在细胞层面解决衰老过程，那为什么不活得更久呢？但我仍然认为，如果人们活得太久，会有社会停滞的风险。我们需要新思想，而新思想通常来自新人。
But if we have AI, the AI can provide the new ideas.
That's a good point. If the AI is the driver of progress, then the biological age of the humans might not matter as much. We'll see. I'm certainly interested in staying healthy and productive for as long as possible.
说得好。如果 AI 是进步的驱动力，那么人类的生物年龄可能就不那么重要了。我们拭目以待。我当然有兴趣尽可能长时间地保持健康和高效。
Let's talk about space again. You've got a big window coming up for Mars.
让我们再次谈谈太空。你即将迎来一个发射火星的大窗口。
2028 and 2029. That's when the planets align. We're aiming to send the first uncrewed Starships to Mars then. If those land safely, then we'll send humans in the following window, 2031.
2028 年和 2029 年。那是行星对齐的时候。我们的目标是到那时向火星发送第一批无人驾驶的星舰。如果那些星舰安全着陆，那么我们将在接下来的窗口期，即 2031 年发送人类。
That's incredibly ambitious.
It has to be. We need to become a multi-planetary species. The window of opportunity could close at any time. We need to get a self-sustaining city on Mars as fast as possible.
必须如此。我们需要成为多行星物种。机会之窗随时可能关闭。我们需要尽快在火星上建立一个自给自足的城市。
What's the biggest risk to the Mars mission?
Landing. Mars has enough atmosphere to be a problem, but not enough to be a help. It's very tricky to land a massive ship like Starship on Mars. But we're learning a lot from our Earth landings. The "chopsticks" on the launch tower are a game-changer.
着陆。火星的大气层足以制造麻烦，但不足以提供帮助。像星舰这样巨大的飞船在火星着陆是非常棘手的。但我们从地球着陆中学到了很多。发射塔上的“筷子”是一个游戏规则改变者。
The catch was amazing to watch.
It was a bit nerve-wracking. But it worked. And it proves that we can achieve full and rapid reusability. That's the key to everything in space.
当时有点紧张。但成功了。这证明了我们可以实现完全且快速的可重复使用。这是太空一切事务的关键。
What about the cost of a ticket to Mars?
Eventually, around $100,000. Something that most people in the developed world could afford if they sold their house and moved. It's not for everyone, but it's for enough people to start a civilization.
最终大约是 10 万美元。发达国家的大多数人如果卖掉房子搬过去就能负担得起。这不是给每个人的，但足以让足够多的人去开启一个文明。
I'd like to. But I have a lot of work to do here first. I want to make sure the program is solid before I hop on a ship. I don't want to leave my companies in a lurch.
我想去。但我得先在这里完成很多工作。我想在登船前确保计划是稳固的。我不想让我的公司陷入困境。
Let's talk about the moon. NASA's Artemis program. How does SpaceX fit in?
让我们谈谈月球。NASA 的阿尔忒弥斯（Artemis）计划。SpaceX 如何参与其中？
We're the lunar lander. Starship will take the astronauts from lunar orbit to the surface and back. It's a huge responsibility. And with Jared Isaacman at NASA, I think we'll see a lot more focus on speed and efficiency. We need to get back to the moon and stay there.
我们是月球着陆器。星舰将把宇航员从月球轨道带到表面并返回。这是一份巨大的责任。随着贾里德·艾萨克曼进入 NASA，我认为我们会看到更多对速度和效率的关注。我们需要回到月球并留在那里。
Yes. Moon Base Alpha. It'll be a stepping stone to Mars. And a great place for science and exploration. Imagine a telescope on the far side of the moon. No radio interference from Earth. We could see the beginning of the universe.
是的。阿尔法月球基地。它将是通往火星的垫脚石。也是科学和探索的好地方。想象一下在月球背面放一个望远镜。没有来自地球的无线电干扰。我们可以看到宇宙的开端。
What about the resources on the moon?
There's water ice in the craters at the poles. That's fuel. And there's helium-3, which could be useful for fusion in the future. But the most important resource is the location. It's a low-gravity platform for launching into the rest of the solar system.
两极的陨石坑里有水冰。那是燃料。还有氦-3，未来可能对聚变有用。但最重要的资源是位置。它是一个低重力平台，可以向太阳系其他地方发射。
Let's switch gears to Tesla. The Cybercab. When do we see it on the streets?
让我们转向特斯拉。Cybercab。我们什么时候能在街上看到它？
Next year. We're already testing it. It's fully autonomous. No steering wheel, no pedals. It's a lounge on wheels. It will change the way people think about transportation. You won't need to own a car. You'll just summon a Cybercab whenever you need to go somewhere.
明年。我们已经在测试了。它是完全自主的。没有方向盘，没有踏板。它是轮子上的休息室。它将改变人们对交通的看法。你不需要拥有一辆车。你只需要在需要去某个地方时召唤一辆 Cybercab。
Less than a bus ticket. It'll be so cheap that it won't make sense to drive yourself. And it'll be much safer. AI doesn't get distracted, it doesn't get drunk, it doesn't fall asleep.
比公交车票还便宜。它会便宜到你自己开车都没有意义。而且它会安全得多。AI 不会分心，不会喝醉，不会睡着。
What about the regulatory hurdles?
They're real. But the data is on our side. Once we can prove that autonomous cars are 10 times safer than human drivers, it'll be hard for regulators to say no. It's a moral imperative to save lives.
它们是真实存在的。但数据站在我们这一边。一旦我们能证明自动驾驶汽车比人类驾驶员安全 10 倍，监管机构就很难说不。拯救生命是道义上的命令。
You've also mentioned a $25,000 Tesla. Is that still happening?
你也提到过 2.5 万美元的特斯拉。还在计划中吗？
The Cybercab is essentially that. But we're also working on a more traditional car at that price point. We want to make electric vehicles accessible to everyone. That's the only way to achieve a sustainable energy future.
Cybercab 本质上就是那个。但我们也正在研发一款那个价位的更传统的汽车。我们想让每个人都能用上电动汽车。这是实现可持续能源未来的唯一途径。
Let's talk about the competition. China. You've said they're doing an incredible job.
They are. They're moving faster than anyone else in EVs, solar, and batteries. They have a long-term vision and they're executing on it. We need to step up our game in the US if we want to stay competitive.
是的。他们在电动汽车、太阳能和电池方面的进展比任何人都快。他们有长期愿景并正在执行。如果我们想保持竞争力，美国需要提升我们的水平。
What's the secret to their success?
Hard work and a focus on engineering. They have a lot of smart people who are willing to put in the hours. And they have a government that supports industrial growth. We need more of that here. We need to value makers and builders again.
努力工作和对工程的关注。他们有很多聪明人愿意投入时间。他们有一个支持工业增长的政府。我们这里需要更多这样的东西。我们需要再次重视制造者和建设者。
You've been very critical of the "woke" culture in the US. How does that impact our competitiveness?
你一直对美国的“觉醒”文化持批评态度。这如何影响我们的竞争力？
It's a mind virus. It's anti-meritocratic and anti-truth. If you're more focused on identity politics than on building great things, you're going to lose. We need to get back to basics. Truth, excellence, and hard work.
这是一种思想病毒。它是反精英管理、反真相的。如果你更关注身份政治而不是建造伟大的事物，你就会输。我们需要回到基础。真相、卓越和努力工作。
Is that why you bought X? To fight the mind virus?
这就是你买下 X 的原因吗？为了对抗这种思想病毒？
It's one of the reasons. I want to make sure there's a place where people can speak freely and challenge the status quo. Without free speech, democracy dies. And without truth, civilization dies.
这是原因之一。我想确保有一个地方人们可以自由发言并挑战现状。没有言论自由，民主就会消亡。没有真相，文明就会消亡。
Let's talk about the future of AI again. You've mentioned the "intelligence density" potential. Can you explain that a bit more?
让我们再次谈谈 AI 的未来。你提到了“智能密度”的潜力。你能再解释一下吗？
Right now, we're using a lot of compute to achieve a certain level of intelligence. But if you look at the human brain, it's incredibly efficient. It runs on about 20 watts. Our current AI models use megawatts. There's a huge gap there. We can make our algorithms much more efficient. We can achieve the same level of intelligence with much less compute. That's what I mean by intelligence density.
现在，我们使用大量的计算来实现一定水平的智能。但如果你看看人类大脑，它是极其高效的。它运行功率约为 20 瓦。我们目前的 AI 模型使用兆瓦级的电力。这之间有巨大的差距。我们可以让我们的算法更高效。我们可以用更少的计算实现相同水平的智能。这就是我所说的智能密度。
Better architectures. Better training methods. And maybe new types of hardware that are more like the biological brain. We're just at the beginning of this journey. The progress we've seen in the last few years is just the tip of the iceberg.
更好的架构。更好的训练方法。也许还有更像生物大脑的新型硬件。我们才刚刚开始这段旅程。过去几年我们看到的进展只是冰山一角。
What happens when we reach that level of efficiency?
Intelligence becomes ubiquitous. It'll be in everything. Your toaster will be smarter than you. Your car will be a genius. Everything will be optimized for maximum efficiency and happiness. It'll be a world of pure magic.
智能将无处不在。它将存在于一切事物中。你的烤面包机都会比你聪明。你的车将是一个天才。一切都将为了最大效率和幸福而优化。那将是一个纯粹魔法的世界。
Is there a limit to intelligence?
The laws of physics are the only limit. And we're nowhere near those limits yet. We can build minds that are millions of times more powerful than the human brain. We can explore the deepest mysteries of the universe. We can become like gods.
物理定律是唯一的限制。而我们现在离这些限制还远得很。我们可以建造比人类大脑强大数百万倍的思维。我们可以探索宇宙最深奥的奥秘。我们可以变得像神一样。
It's an exciting thought. We're the ones who get to build it. We're the ones who get to decide what kind of future we want. Let's make it a good one.
这是一个令人兴奋的想法。我们是建造它的人。我们是决定我们想要什么样的未来的人。让我们把它变成一个美好的未来。
What's your daily routine like these days? You're running so many companies.
It's intense. I move between Austin, Boca Chica, and San Francisco. I spend a lot of time in meetings, but I also spend a lot of time on the factory floor and in the engineering labs. I like to be where the action is. I like to solve problems.
很紧凑。我在奥斯汀、博卡奇卡和旧金山之间奔波。我花很多时间开会，但也花很多时间在工厂车间和工程实验室。我喜欢在行动发生的地方。我喜欢解决问题。
How do you manage the stress?
I don't really think about it as stress. It's just work. I'm driven by a sense of mission. I want to make sure humanity has a bright future. That's what keeps me going. And I have a great team of people who support me. I couldn't do it without them.
我并不真的认为那是压力。那只是工作。我受使命感的驱动。我想确保人类有一个光明的未来。这就是让我坚持下去的动力。我有一支支持我的优秀团队。没有他们我做不到。
About 6 hours a night. I've tried less, but it's not sustainable. You need a clear head to make good decisions.
每晚大约 6 小时。我试过更少，但不可持续。你需要清醒的头脑来做出正确的决定。
What's the one thing you want people to take away from this conversation?
Optimism. The future is going to be amazing. We have the tools to solve all of our problems. We just need the will to use them. Don't be afraid of the future. Embrace it. Be a part of building it.
乐观。未来将是惊人的。我们拥有解决所有问题的工具。我们只需要使用它们的意志。不要害怕未来。拥抱它。成为建设它的一分子。
Thank you, Elon. This has been incredible.
Elon, we've covered a lot of ground. Energy, AI, space, health. I want to talk about the "big picture." What is the ultimate goal of all of this?
埃隆，我们已经讨论了很多领域。能源、AI、太空、健康。我想谈谈“大局”。这一切的终极目标是什么？
The ultimate goal is to preserve the light of consciousness. As far as we know, we're the only conscious beings in the universe. That's a huge responsibility. We need to make sure that consciousness survives and thrives, not just on Earth, but throughout the galaxy.
终极目标是保护意识之光。据我们所知，我们是宇宙中唯一的意识存在。这是一份巨大的责任。我们需要确保意识能够生存并繁荣，不仅在地球上，而且在整个银河系。
And you think becoming multi-planetary is the key to that.
It's a necessary step. If we're on only one planet, we're vulnerable to extinction. An asteroid, a super-volcano, a nuclear war, a pandemic. There are so many things that could wipe us out. By spreading to other planets, we're backing up the hard drive of civilization.
这是一个必要的步骤。如果我们只在一个星球上，我们就很容易灭绝。小行星、超级火山、核战争、大流行病。有太多事情可以把我们抹去。通过扩展到其他星球，我们正在备份文明的硬盘。
What about the role of AI in this? Is AI a threat to consciousness?
那 AI 在其中的角色呢？AI 对意识是一种威胁吗？
It could be. That's why I'm so focused on AI safety. We need to make sure that AI is aligned with human values and that it respects consciousness. If we get it right, AI will be the greatest tool for expanding consciousness. It will help us understand the universe in ways we can't even imagine.
可能是。这就是为什么我如此关注 AI 安全。我们需要确保 AI 与人类价值观保持一致，并且尊重意识。如果我们做对了，AI 将成为扩展意识的最伟大工具。它将帮助我们以甚至无法想象的方式理解宇宙。
You've mentioned the "simulation theory" before. Do you think we're living in a simulation?
你以前提到过“模拟理论”。你认为我们生活在模拟中吗？
The odds that we're in "base reality" are one in billions. If you look at the progress of video games, from Pong to photorealistic 3D simulations, it's clear that we'll eventually be able to create simulations that are indistinguishable from reality. If that's the case, then it's highly likely that we're already in one.
我们在“基础现实”中的概率是几十亿分之一。如果你看看视频游戏的进步，从《乓》（Pong）到逼真的 3D 模拟，很明显我们最终将能够创造出与现实无法区分的模拟。如果是这样的话，那么我们极有可能已经身处其中了。
And what does that mean for our sense of meaning and purpose?
It doesn't change anything. Even if we're in a simulation, our experiences are real to us. Our choices matter. We should still strive to build a great future and to be good to each other. In fact, if we're in a simulation, it's even more important to be interesting. Only the interesting simulations survive.
它不会改变任何事情。即使我们在模拟中，我们的经历对我们来说也是真实的。我们的选择很重要。我们仍然应该努力建设一个伟大的未来，并善待彼此。事实上，如果我们在模拟中，变得有趣就更加重要了。只有有趣的模拟才能存活。
How do we stay interesting?
By exploring, by creating, by pushing the boundaries of what's possible. By being curious and seeking the truth. That's what makes life worth living.
通过探索、创造、突破可能的边界。通过保持好奇心和寻求真理。这就是让生活值得活下去的原因。
Let's talk about the next 12 months. What are you most excited about?
让我们谈谈接下来的 12 个月。你最兴奋的是什么？
Starship reaching orbit regularly. Optimus starting to do useful work in the factories. Grok 5 being released. The Cybercab testing. It's going to be a year of massive breakthroughs. We're moving from the "lab" to the "real world" in so many areas.
星舰定期进入轨道。Optimus 开始在工厂里做有用的工作。Grok 5 发布。Cybercab 测试。这将是取得巨大突破的一年。我们在很多领域正从“实验室”转向“现实世界”。
And the biggest challenge in the next 12 months?
Execution. We have the plans, we have the technology, we just need to build it. It's all about the "machine that builds the machine." We need to scale up production and ensure quality. It's a lot of hard work.
执行。我们有计划，有技术，我们只需要去建造它。一切都关乎“制造机器的机器”。我们需要扩大生产规模并确保质量。这需要大量的艰苦工作。
What about the political environment? Does that worry you?
It's a factor. But I'm optimistic that we're moving in the right direction. People are starting to realize that we need a pro-growth, pro-technology agenda. We need to get rid of the "mind virus" and get back to building.
这是一个因素。但我乐观地认为我们正朝着正确的方向前进。人们开始意识到我们需要一个支持增长、支持技术的议程。我们需要摆脱“思想病毒”，回到建设中来。
You've been a big supporter of the new administration. What do you hope to see from them?
你一直是新政府的大力支持者。你希望从他们那里看到什么？
Deregulation. Efficiency. A focus on the future. We need to make it easier to build things in the US. We need to clear the path for innovation. If we do that, there's no limit to what we can achieve.
去监管。效率。关注未来。我们需要让在美国建造东西变得更容易。我们需要为创新扫清道路。如果我们做到这一点，我们能取得的成就将没有限制。
What about the global situation? China, Europe, the Middle East.
It's complex. But I think technology is a unifying force. Everyone wants abundance. Everyone wants a better life for their children. If we can provide that through technology, we can reduce the incentives for conflict. We need to focus on the "positive sum" games.
这很复杂。但我认为技术是一种统一的力量。每个人都想要富足。每个人都想让孩子过上更好的生活。如果我们能通过技术提供这些，我们就能减少冲突的动机。我们需要关注“正和”博弈。
Yes. That's the goal. A world where everyone has what they need and the opportunity to pursue their dreams. It's possible. We just have to build it.
是的。这就是目标。一个每个人都能拥有所需之物并有机会追求梦想的世界。这是可能的。我们只需要去建造它。
What's your advice to young people today?
Learn how to build things. Learn engineering, physics, computer science. Don't just be a consumer, be a creator. And don't be afraid to fail. Failure is just a part of the learning process. The only real failure is giving up.
学习如何建造东西。学习工程、物理、计算机科学。不要只做一个消费者，要做一个创造者。不要害怕失败。失败只是学习过程的一部分。唯一的真正失败是放弃。
And to the older generation?
Support the young people. Give them the tools and the freedom to innovate. Don't hold them back with old ideas and regulations. Embrace the future. It's going to be better than you think.
支持年轻人。给他们创新的工具和自由。不要用旧思想和规章制度阻碍他们。拥抱未来。它会比你想象的更好。
What's the most important lesson you've learned in your life?
Persistence. Never give up. If something is important enough, you should do it even if the odds are against you. And always seek the truth, no matter how uncomfortable it is.
坚持。永不放弃。如果某件事足够重要，即使胜算不大你也应该去做。并且永远寻求真理，无论这让人多么不舒服。
Yes. Truth is the foundation of everything. Without truth, you're just building on sand.
是的。真理是一切的基础。没有真理，你只是在沙滩上盖房子。
We're coming to the end of our time. Any final words for our audience?
The future is in your hands. Don't just watch it happen, make it happen. Be optimistic. Be curious. Be bold. The universe is waiting for us.
未来掌握在你们手中。不要只是看着它发生，去创造它。保持乐观。保持好奇。保持大胆。宇宙正在等待着我们。
Thank you, Elon. This has been a truly inspiring conversation.
And thank you to our audience for joining us. We hope this conversation has given you a dose of optimism and a new perspective on the future.
也谢谢我们的听众加入我们。我们希望这次对话给你们带来了一剂乐观药，以及对未来的新视角。
We're back for a special bonus segment. Elon, we were talking about the "intelligence density" and the future of AI. I want to ask about the "recursive" nature of AI. When AI starts building AI, what happens?
我们回来进行特别加更环节。埃隆，我们刚才在谈论“智能密度”和 AI 的未来。我想问问 AI 的“递归”本质。当 AI 开始构建 AI 时，会发生什么？
That's the point of no return. Once AI can improve its own code and design its own hardware, the rate of progress becomes vertical. We're already seeing the beginnings of this. We use AI to optimize the layout of our chips and to write some of our software. But when the AI is doing 99% of the work, that's when things get really interesting.
那是无法回头的点。一旦 AI 能够改进自己的代码并设计自己的硬件，进步的速度就会变得垂直。我们已经看到了这方面的开端。我们使用 AI 来优化芯片布局并编写部分软件。但当 AI 完成 99%的工作时，事情就会变得非常有趣。
Is that what you mean by the "supersonic tsunami"?
Yes. It's a wave of change that is moving faster than our ability to process it. We're still thinking in linear terms, but the technology is moving exponentially. We need to upgrade our own "hardware" if we want to keep up.
是的。这是一股变革浪潮，其移动速度超过了我们的处理能力。我们仍然在以线性方式思考，但技术正在以指数级速度移动。如果我们想跟上，我们需要升级我们自己的“硬件”。
Upgrade our hardware? You mean Neuralink?
Neuralink is part of it. But it's also about our culture and our institutions. We need to be more agile and more open to change. We can't rely on the old ways of doing things. We need to be as fast and as efficient as the technology we're building.
Neuralink 是其中的一部分。但这也关乎我们的文化和制度。我们需要更敏捷，对变革更开放。我们不能依赖旧的做事方式。我们需要像我们正在建造的技术一样快速、高效。
Let's talk about the "abundance" again. If everything is free, what happens to the concept of "value"?
让我们再次谈谈“富足”。如果一切都是免费的，那么“价值”的概念会发生什么？
Value will shift from "utility" to "meaning." Right now, something is valuable if it's useful. In an abundant world, something will be valuable if it's meaningful. A hand-written letter will be more valuable than an AI-generated email. A meal cooked by a friend will be more valuable than a meal prepared by a robot. Human connection will be the ultimate currency.
价值将从“效用”转向“意义”。现在，如果某样东西有用，它就有价值。在一个富足的世界里，如果某样东西有意义，它就有价值。一封手写的信将比一封 AI 生成的邮件更有价值。朋友亲手做的饭将比机器人做的饭更有价值。人类的联系将成为终极货币。
That's a very human-centric view for someone who's building so many robots.
对于一个制造这么多机器人的人来说，这是一个非常以人为中心的观点。
The robots are just tools. They're there to free us from the drudgery of survival so that we can focus on what really matters. We're not building robots to replace humans, we're building them to empower humans.
机器人只是工具。它们的存在是为了把我们从生存的苦役中解放出来，这样我们就可以专注于真正重要的事情。我们制造机器人不是为了取代人类，而是为了赋能人类。
What about the risk of "de-skilling"? If robots do everything, will we forget how to do anything?
那“去技能化”的风险呢？如果机器人做了一切，我们会忘记如何做任何事吗？
It's a risk. We see it already with things like GPS and calculators. But I think we'll just develop new skills. Instead of learning how to hunt or farm, we'll learn how to create and explore. We'll focus on the higher-level functions of the mind.
这是一个风险。我们已经在 GPS 和计算器等事物上看到了这一点。但我认为我们只会发展出新的技能。与其学习如何狩猎或耕种，我们将学习如何创造和探索。我们将专注于大脑更高层次的功能。
Let's talk about the "energy" again. You mentioned space-based solar. What about nuclear?
让我们再次谈谈“能源”。你提到了太空太阳能。那核能呢？
Nuclear is great for base load power. We should definitely keep our existing nuclear plants open and build new ones. But it's not as scalable as solar. Solar is the ultimate energy source. It's clean, it's abundant, and it's everywhere.
核能对于基础负荷电力非常有用。我们绝对应该保持现有的核电站运行并建造新的。但它不像太阳能那样具有可扩展性。太阳能是终极能源。它清洁、充足且无处不在。
Batteries are the key to making solar work at scale. We need to be able to store the energy for when the sun isn't shining. Tesla is leading the way in battery technology, but we need more. We need a massive increase in battery production to power the world.
电池是让太阳能大规模发挥作用的关键。我们需要能够在阳光不灿烂的时候储存能量。特斯拉在电池技术方面处于领先地位，但我们需要更多。我们需要大幅增加电池产量来为世界提供动力。
What about the "materials" for the batteries? Lithium, cobalt, nickel.
There's plenty of material in the Earth's crust. We just need to mine it and refine it. And we're working on new battery chemistries that use more abundant materials. In the long run, we'll be able to recycle almost all of the materials in our batteries. It'll be a closed-loop system.
地壳中有充足的材料。我们只需要开采并提炼。我们正在研发使用更充足材料的新型电池化学成分。从长远来看，我们将能够回收电池中几乎所有的材料。这将是一个闭环系统。
Let's talk about the "transportation" again. The Roadster. You mentioned it'll have "hover" capability.
让我们再次谈谈“交通”。Roadster。你提到它将具有“悬浮”能力。
Yes. It'll use cold gas thrusters, similar to what we use on our rockets. It'll be able to hover and maybe even fly for short distances. It's going to be a mind-blowing demo. It's not a practical car, it's a statement. A statement of what's possible.
是的。它将使用冷气推进器，类似于我们在火箭上使用的。它将能够悬浮，甚至可能进行短距离飞行。这将是一个令人震撼的演示。它不是一辆实用的车，它是一个声明。一个关于可能性的声明。
Next year. We're working on the final design now. It's going to be the most exciting car ever built.
明年。我们现在正在进行最终设计。它将是有史以来最令人兴奋的汽车。
And the "Starship"? When is the next flight?
Soon. We're aiming for a flight every month or two. We need to get to the point where it's a routine operation. Once we achieve that, the sky is the limit. Or rather, the sky is no longer the limit.
很快。我们的目标是每一两个月飞行一次。我们需要达到常规运行的程度。一旦我们实现了这一点，天空就是极限。或者更确切地说，天空不再是极限。
What about the "Mars" mission? You mentioned 2028. Is that still the goal?
那“火星”任务呢？你提到了 2028 年。那仍然是目标吗？
Yes. 2028 for the first uncrewed mission. 2031 for the first crewed mission. It's an aggressive timeline, but we need to be aggressive. We're racing against time.
是的。2028 年进行第一次无人任务。2031 年进行第一次载人任务。这是一个激进的时间表，但我们需要激进。我们正在与时间赛跑。
Racing against time? What do you mean?
The window of opportunity for becoming multi-planetary could close at any time. We don't know how long this period of relative peace and prosperity will last. We need to take advantage of it while we can. We need to make sure that consciousness is secure.
成为多行星物种的机会之窗随时可能关闭。我们不知道这段相对和平与繁荣的时期会持续多久。我们需要在可以的时候利用它。我们需要确保意识是安全的。
That's a very sobering thought.
It's a realistic thought. We shouldn't take our survival for granted. We need to work for it.
这是一个现实的想法。我们不应该认为我们的生存是理所当然的。我们需要为此努力。
Let's talk about the "future of humanity" again. What's the best-case scenario?
The best-case scenario is that we merge with AI and become a multi-planetary species. We explore the galaxy and beyond. We solve all of our problems and live in a world of infinite abundance and meaning. We become a truly advanced civilization.
最好的情况是我们与 AI 融合并成为多行星物种。我们探索银河系及更远的地方。我们解决所有问题，生活在一个无限富足和有意义的世界里。我们成为一个真正先进的文明。
And the worst-case scenario?
The worst-case scenario is that we destroy ourselves before we reach that point. Either through AI gone wrong, or through our own stupidity and conflict. We could become extinct. That's why the stakes are so high.
最坏的情况是我们在达到那个点之前毁灭了自己。要么是因为 AI 出错，要么是因为我们自己的愚蠢和冲突。我们可能会灭绝。这就是为什么赌注如此之高。
What can the average person do to help?
Be informed. Be engaged. Support the technologies and the policies that will move us in the right direction. And most importantly, be optimistic. Optimism is a self-fulfilling prophecy. If we believe we can build a great future, we will.
保持知情。保持参与。支持那些能把我们引向正确方向的技术和政策。最重要的是，保持乐观。乐观是一个自我实现的预言。如果我们相信我们能建设一个伟大的未来，我们就能做到。
Optimism as a self-fulfilling prophecy. I like that.
It's true. If you're pessimistic, you've already lost. You won't even try. But if you're optimistic, you'll find a way.
这是真的。如果你悲观，你就已经输了。你甚至都不会去尝试。但如果你乐观，你会找到办法。
We're almost out of time for this bonus segment. Any final thoughts, Elon?
这个加更环节的时间快到了。有什么最后的想法吗，埃隆？
Just that I'm grateful to be alive at this time. It's the most exciting time in human history. We're on the verge of something truly amazing. Let's make the most of it.
只是我很感激能生活在这个时代。这是人类历史上最激动人心的时刻。我们正处于真正惊人的事物的边缘。让我们充分利用它。
We're in the home stretch now. Elon, we've talked about the future of humanity, but what about the "now"? What are the most important things we need to do in the next 2 to 3 years to ensure we stay on the right path?
我们现在进入最后阶段了。埃隆，我们谈论了人类的未来，但“现在”呢？在接下来的 2 到 3 年里，我们需要做的最重要的事情是什么，以确保我们走在正确的道路上？
The most important thing is to avoid a major conflict. We need to maintain global stability. If we have a major war, all of our plans for the future will be set back by decades, if not centuries. We need to find ways to cooperate and to solve our problems peacefully.
最重要的事情是避免重大冲突。我们需要维持全球稳定。如果我们发生一场大战，我们对未来的所有计划都将推迟几十年，甚至几个世纪。我们需要找到合作的方法，和平地解决我们的问题。
The second thing is to accelerate the transition to sustainable energy. We're still too dependent on fossil fuels. We need to move to solar and wind and batteries as fast as possible. This is not just about the environment, it's about energy security and economic stability.
第二件事是加速向可持续能源的过渡。我们仍然过于依赖化石燃料。我们需要尽快转向太阳能、风能和电池。这不仅关乎环境，也关乎能源安全和经济稳定。
The third thing is to ensure that AI is developed safely and responsibly. We need to have the right regulations and the right oversight. We can't just let it happen without any control. We need to make sure that AI is a force for good.
第三件事是确保 AI 的开发是安全且负责任的。我们需要有正确的监管和监督。我们不能任由它在没有任何控制的情况下发生。我们需要确保 AI 是一股向善的力量。
You've been very active in the political sphere lately. Do you think you can influence these things?
你最近在政治领域非常活跃。你认为你能影响这些事情吗？
I hope so. I'm using my voice and my resources to try to move the needle. But it's not just up to me. It's up to all of us. We all have a role to play in building the future.
我希望如此。我正在利用我的声音和资源努力推动变革。但这不仅仅取决于我。这取决于我们所有人。我们在建设未来中都有自己的角色。
What about the "national debt"? You've mentioned that it's a major concern.
It's a huge concern. We're spending more on interest than on defense. That's not sustainable. We need to get our fiscal house in order. We need to cut spending and increase efficiency. We need a government that is as lean and as efficient as a startup.
这是一个巨大的担忧。我们在利息上的支出超过了国防支出。这是不可持续的。我们需要理顺我们的财政状况。我们需要削减开支并提高效率。我们需要一个像初创公司一样精简、高效的政府。
It's necessary. If we don't do it, we'll face a financial crisis that will make the 2008 crisis look like a walk in the park. We need to act now.
这是必要的。如果我们不这样做，我们将面临一场金融危机，会让 2008 年的危机看起来像是在公园散步。我们现在就需要采取行动。
Let's talk about the "space" again. You mentioned the "Starship" and the "Mars" mission. What about the "asteroid" threat?
让我们再次谈谈“太空”。你提到了“星舰”和“火星”任务。那“小行星”威胁呢？
It's a real threat. We know that asteroids have hit the Earth in the past and caused mass extinctions. It's only a matter of time before it happens again. That's why we need to have a presence in space. We need to be able to detect and deflect asteroids before they hit us.
这是一个真实的威胁。我们知道小行星过去曾撞击地球并导致大规模灭绝。再次发生只是时间问题。这就是为什么我们需要在太空存在。我们需要能够在小行星撞击我们之前探测并偏转它们。
And the "Starship" can do that?
Yes. Starship is the only ship that has the mass and the power to deflect a large asteroid. It's our planetary defense system.
是的。星舰是唯一一艘拥有偏转大型小行星所需的质量和动力的飞船。它是我们的行星防御系统。
That's a lot of responsibility for one company.
It's a responsibility for all of humanity. We're just the ones building the tools. We need everyone to support this effort.
这是全人类的责任。我们只是建造工具的人。我们需要每个人都支持这项努力。
Let's talk about the "future of work" again. You mentioned the "Universal High Income." How do we ensure that people don't lose their sense of purpose?
让我们再次谈谈“工作的未来”。你提到了“全民高收入”。我们如何确保人们不会失去目的感？
That's the biggest challenge. We need to find new ways to provide meaning and purpose. Maybe it's through education, or art, or community service. We need to encourage people to be creative and to contribute in ways that aren't just about making money.
这是最大的挑战。我们需要找到提供意义和目的的新方法。也许是通过教育、艺术或社区服务。我们需要鼓励人们发挥创造力，并以不仅仅是为了赚钱的方式做出贡献。
What about the "education" system? You mentioned that it's broken.
It is. We're still teaching kids the same way we did 100 years ago. We need to move to a more personalized, AI-driven education system. We need to teach kids how to think, not just what to think. We need to foster curiosity and creativity.
确实。我们仍然以 100 年前的方式教孩子。我们需要转向更个性化、AI 驱动的教育系统。我们需要教孩子如何思考，而不仅仅是思考什么。我们需要培养好奇心和创造力。
And the "Grok" can help with that?
Yes. Grok is already a great teacher. It can answer any question and explain complex concepts in a way that's easy to understand. We're working on making it even better. We want to make Grok the ultimate personalized tutor for every child in the world.
是的。Grok 已经是一个伟大的老师。它可以回答任何问题，并以易于理解的方式解释复杂的概念。我们正在努力让它变得更好。我们想让 Grok 成为世界上每个孩子终极的个性化导师。
That's a beautiful vision.
It's a possible vision. We just have to build it.
We're almost out of time. Any final thoughts, Elon?
Just that I'm optimistic. I believe in humanity. I believe that we can solve our problems and build a great future. We just need to work together and stay focused on the goal.
只是我很乐观。我相信人类。我相信我们能解决我们的问题并建设一个伟大的未来。我们只需要共同努力，并专注于目标。
And thank you to our audience for joining us for this deep dive with Elon Musk. We hope you found it as inspiring as we did. Until next time, keep reaching for the stars.
也谢谢我们的听众加入这次与埃隆·马斯克的深度交流。我们希望你们觉得它和我们一样鼓舞人心。下次见，继续仰望星空。
*本文所有核心观点均来自马斯克在《Moonshots with Peter Diamandis》播客的实录，2026年1月6日正式发布，补充信息均来自访谈上下文。完整访谈可通过播客平台查阅。
▌声明：为分享前沿资讯及有价值的观点，#空天界 微信公众号编写此文。
本平台发布和分享的有关文章和信息的观点及版权属于文章和信息的作者或者来源渠道。本平台发布和分享的文章和信息等，不代表本平台同意其观点，只是旨在战略决策咨询、专业研究与学习用途。分享信息若有侵权，联系后台删除。AI编译内容请谨慎参考。
