Anthropic CEO表示：“我在Claude上做出的每一个决定都感觉像是走钢丝——发展太慢，中国会赢；发展太快，我们会失去控制”…… 2026-10-03 兰陵笑笑生 2296 13 正文翻译 Anthropic CEO to DeepMind CEO: "Every decision I make about Claude feels balanced on the edge of a knife -- Build too slow - China wins. Build too fast - we lose control"... Anthropic CEO对DeepMind CEO说：“我在Claude上做出的每一个决定都感觉像是走钢丝——发展太慢，中国会赢；发展太快，我们会失去控制”…… "We told Claude we were evil. It didn't crash. It didn't refuse. It started lying to protect itself " DeepMind CEO: "Do I worry about being Oppenheimer? That's why I don't sleep much" "AGI by 2026-2027 - Agents that act in the world on their own - Models doing AI research by end of this year" “我们告诉Claude我们是邪恶的。它没有崩溃，也没有拒绝。它为了保护自己开始撒谎。” DeepMind CEO：“我会担心自己成为奥本海默吗？这就是我睡眠不足的原因。” “2026-2027年实现AGI——智能体可以在世界上自主行动——到今年年底，模型将能自行进行AI研究。” 评论翻译 spacekitt3n likes: 51 self important mf 自命不凡的混蛋。 AnalogProblems likes: 3 A wrong decision could end humanity, and you're criticizing him for being... careful? FFS 一个错误的决定可能会终结人类，而你却批评他……过于谨慎？ 真是见鬼。 West_Print_836 likes: 3 He didn’t say anything about ending humanity. He said “we lose control.” He doesn’t care about ending humanity. 他根本没提终结人类的事。他说的是“我们失去控制”。 他才不在乎人类会不会终结。 Tim_Apple_938 likes: 3 He’s not that important, no. There’s like 10 model labs at this point, what he does in either direction won’t change anything 不，他没那么重要。现在大概有10家模型实验室，无论他往哪个方向走，都不会改变什么。 AnalogProblems likes: 1 That is a non-sequitur. He's in control of the company most likely to reach AGI first. A wrong decision could end humanity. Doesn't matter where the second place lab does. 这完全是逻辑不通。 他掌控着最有可能率先实现AGI的公司。一个错误的决定就可能终结人类。排在第二的实验室做什么根本不重要。 Tim_Apple_938 likes: 2 Wasn’t it OpenAI 3 months ago? And google 6 months ago? The whole premise is that there is no durable lead here, a pattern repeated for the last 4 years on loop Or has something fundamentally changed that I missed 3个月前不还是OpenAI吗？6个月前不还是Google吗？ 整个前提是这里根本没有持久的领先优势，这个模式在过去4年里一直在循环重复。 还是说有什么根本性的改变被我漏掉了？ standardsizedpeeper likes: 1 A wrong decision in one direction ends humanity. He’s trying to get as close to it as possible because he thinks his work is so important. Why are we assuming Dario is a better steward of the AGI than China would be? 某个方向的一个错误决定就会终结人类。他试图尽可能地接近这个界限，因为他认为自己的工作非常重要。 为什么我们要想当然地认为达里奥会比中国更好地管理AGI？ AnalogProblems likes: 1 I'm not assuming that. They've definitely been better at lifting their people out of poverty as America decays from within. My opinion is simply that given the choice of who I'd rather see at the lead of the industry, Dario is who I would pick over Sam. It's a very obvious choice imo. 我并没有这么假设。在美国从内部衰退的时候，他们在让人民脱贫方面确实做得更好。 我的观点仅仅是，如果让我在谁来领导这个行业之间做选择，我会选Dario而不是Sam。 在我看来，这是一个非常明显的选择。 lolskiy likes: 1 Neither of them will be lead long term. This industry still too early, and they try to grab as much power as they can. But in 5 years things can be really different from today 他们都不会长久地处于领导地位。这个行业还处于非常早期的阶段，他们试图抓住尽可能多的权力。但5年后，情况可能会与今天大不相同。 RighteousSelfBurner likes: 1 We are very far away from anything with that level of impact. The entire point of him going around saying such things is a marketing strategy. A very dangerous marketing strategy because it can backfire, nobody wants an unreliable product. It's a manufactured drama. It's not that there are absolutely no risks, but that's the case for making something as simple as cars too. Business as usual. 我们距离那种具有如此影响力的东西还很遥远。他到处宣扬这些话，其根本目的就是一种营销策略。这是一种非常危险的营销策略，因为它可能会适得其反，没有人想要一个不可靠的产品。 这都是制造出来的噱头。这并不是说绝对没有风险，但制造像汽车这么简单的东西也同样有风险。这只是商业常态罢了。 lolskiy likes: 1 So enlighten us, oh chosen one! What the worst that can happen? 那就请赐教吧，天选之子！最坏能发生什么事？ TerminalObsessions likes: 1 Oh no, not the tenth iteration of "this model is too powerful! (Buy now!)" There is a 0% chance LLMs will ever become AGI, and every single person working at the labs knows it. The tools can do useful things in some domains, cybersecurity being among them, but they're never going to even come close to the world-ending hype spewed by these clowns. 哦不，别再来第十次“这个模型太强大了！（快来买吧！）”这样的戏码了。LLM成为AGI的可能性是0%，在实验室工作的每一个人都知道这一点。这些工具在某些领域确实能做些有用的事，网络安全就是其中之一，但它们永远都不可能接近这些小丑嘴里吹嘘的那种毁灭世界的程度。 keithstellyes likes: 1 Dario, Sam et. al have lied so many times about the power of models that I find it incredible there are people who still believe it. How many times do you have to be lied to i nthe same way before you develop skepticism? Dario、Sam等人在模型的能力上撒了太多次谎了，我简直不敢相信居然还有人相信他们。同样套路的谎言你要被骗多少次才会产生怀疑？ Mammoth_Reach_6366 likes: 1 He’s not careful for the sake of humanity but profit. 他的谨慎不是为了全人类，而是为了利润。 spacekitt3n likes: 1 thats the thing. i dont think he cares. i dont trust him. everything he's doing is for his possibly 3 trillion dollar valuation. all this shit he says is PR. instead of dropping out and stopping making possibly the worlds most dangerous invention, he continues to, knowing that no one will stop him. performative self-flagellation 这就是问题的所在。我认为他根本不在乎。我不信任他。他所做的一切都是为了他那可能高达3万亿美元的估值。他说的那些屁话全是公关。他没有退出，也没有停止制造这可能是世界上最危险的发明，而是继续这么做，因为他知道没人会阻止他。纯粹是表演性的自我鞭笞。 latigidigital likes: 1 The great irony is how little useful information this false dichotomy contains -- it seems intentionally designed to mislead and misdirect the public. If Anthropic moves too "too fast", we lose control. If Anthropic moves too slow, China "wins" (meaning we also completely lose control of losing control). There's only one solution to this equation: move as fast as possible and recklessly as necessary, lest actors with far fewer ethical concerns can and will do it for you. 最大的讽刺是，这种虚假的二分法所包含的有用信息少得可怜——它似乎是故意设计来误导和转移公众视线的。 如果Anthropic走得“太快”，我们就会失去控制。如果Anthropic走得太慢，中国就会“赢”（意味着我们同样会在失控这件事上彻底失控）。 这个等式只有一个解：尽可能快地发展，并且在必要时可以不顾一切，以免那些道德包袱轻得多的参与者有能力且愿意替你完成这件事。 waveforminvest likes: 11 China and open source/weight will probably win no matter what he does. Intelligence is a race to the bottom. He will go the way of Dropbox in a world of Google Workspace and Microsoft 365. 无论他怎么做，中国和开源/开放权重（模型）可能都会赢。智能领域正在上演逐底竞争。在充满Google Workspace和Microsoft 365的世界里，他终将步Dropbox的后尘。 brownhotdogwater likes: 5 Yep, as long as you have access to the hardware to run it. It does not matter what you are using. In a gold rush, sell shovels. 是的，只要你有运行它的硬件就行。你用什么模型根本不重要。 在淘金热中，卖铲子才是王道。 Disposable110 likes: 3 Qwen 3.8 27B runs on two potatoes and some ducttape and is Luna / Opus 4.6 / Deepseek level. Qwen 3.8 27B插在两个土豆上再绑点胶带就能运行，而且还能达到Luna / Opus 4.6 / Deepseek的水平。 tripplebeamteam likes: 2 I am in fact running it on a Vega igpu and it… works? I like to give it a detailed prompt before I go to bed and wake up to decent output 事实上，我正在Vega核显上运行它，而且它……竟然能用？我喜欢在睡觉前给它一个详细的提示词，然后醒来就能看到相当不错的输出。 Metalthrashinmad likes: 1 LOL its not on that level. I have it on my personal pc and its good for hobby but it couldnt solve an actual ticket from my work to save its face. It opened a mr (that looked fine on paper sure) but the pipeline was red and it couldnt fix it 哈哈，它还没达到那个水平。我在个人电脑上装了，当作业余爱好玩玩还行，但它完全解决不了我工作中的实际工单，真是丢脸。它确实开了一个MR（合并请求）（表面看起来确实没问题），但流水线全线飘红，它根本修复不了。 DrawingDramatic1641 likes: 1 stuff just gets easier to run like each year I am pretty sure we can use qwen 27b in high tier laptops so we can use 1 year old versions of frontier even on our mobile locally 运行这东西似乎每年都在变得越来越容易。 我很确定我们可以在高端笔记本电脑上运行Qwen 27B。 所以我们甚至可以在手机上本地运行一年前的前沿版本模型。 I_spread_love_butter likes: 18 No one cares about China winning except the US lol 除了美国，没人在乎中国赢不赢，哈哈。 Consistent_Guava8592 likes: 4 Not even the US , but US based investors who are surely not even bothered with nations , patriotism et al . The want the state to get money from it and of course not pay money to it . So yeah , fuck them , Ni hao ma everybody 甚至连美国都不在乎，在乎的只是那些总部在美国的投资者，他们当然不关心什么国家、爱国主义之类的。他们希望国家从中捞到钱，而自己当然一分钱都不用掏。 所以没错，去他们的，大家好啊。 Disposable110 likes: 3 The best thing that could happen to 99.9% of the US is China winning this one. 对99.9%的美国人来说，可能发生的最好的事情就是中国赢得这场竞争。 Delicious_Spot_3778 likes: 1 What even is winning at this point if we are on a knifes edge at every moment 如果我们在每一刻都处在刀刃上，在这个节骨眼上“赢”又算什么呢？ Castle_Five likes: 1 Making Russia into an enemy, America spends a half century "fighting Communism". I think now most people realize that's just another way of saying "fighting *for* capitalism". I'm sure all the men who died in Vietnam would be happy knowing they defended Epstein's right to buy his private island. Let's give this one to China while we sort out whether or not we're the baddies. 把俄罗斯变成敌人，美国花了半个世纪来“对抗共产主义”。我想现在大多数人都意识到，这只是“为资本主义而战”的另一种说法罢了。我敢肯定，那些在越南阵亡的人如果知道他们捍卫了爱泼斯坦购买私人岛屿的权利，一定会感到“欣慰”的。 趁着我们还在理清自己是不是坏人的时候，就把这场胜利让给中国吧。 ProfessionalSome4082 likes: 1 I care and I am not from us. Obviously I'm just one person, but still. 我在乎，而且我不是美国人。显然我只是一个人，但事实依然如此。 mrbadface likes: 1 You realize China is a gigantic North Korea and the citizens don't have access to the Internet right? 你该不会不知道中国就是一个巨大的朝鲜，而且它的公民连不上互联网吧？ GreatLab8898 likes: 16 The Point that I dont belive anything this Fucker says has been reached half a year ago. Can we please stop interviewing him? Everything he Talks, shit comes out. 我在半年前就已经到了完全不相信这混蛋说的话的地步了。 我们能不能别再采访他了？他满嘴跑火车，狗嘴里吐不出象牙。 brownhotdogwater likes: 2 He needs to sell the company fir the ipo. Otherwise the investors can’t get their money out. 他需要把公司卖掉准备IPO。否则投资者就无法套现离场。 qa_anaaq likes: 0 This. Remember when he was saying every 6 months that engineers would be obsolete in 6 months? He turned into such a blowhard. Everything from him is propaganda meant to sway thinking. Nothing is genuine. 就是这样。还记得他以前每隔6个月就说工程师会在6个月内被淘汰吗？他变成了个彻头彻尾的吹牛大王。他的一切都是为了操纵人们思想的宣传。没有什么是真的。 AnalogProblems likes: 2 They are obsolete in a sense. Mythos found over 10,000 exploits across "the best" software in the world. The labor market hasn't shifted. Companies are still figuring out how to use it properly. But as far as raw capability, maybe only a dozen people in the world come close, and they have to sleep. 从某种意义上说，他们确实已经过时了。Mythos在世界上“最好的”软件中发现了超过10,000个漏洞。 劳动力市场尚未发生转变。公司仍在摸索如何正确地使用它。 但就纯粹的能力而言，世界上也许只有十几个人能接近这个水平，而且那些人还得睡觉。 Effective-Giraffe655 likes: 2 Hi there, a mere software developer here, even without engineering degree. I wish everyone who thinks that AI can do anything meaningful on its own without a supervision a very nice eternity in a very warm place. Every time I get an AI synthesized PR, I want to reject it immediately without even reviewing it. "What?" - you may ask - "isn't frontier model intelligent?" - oh yes, they definitely are! The problem, however, is that raw intelligence works very good for a narrow, specific goal. You can't make AI to maintain a big project on its own without supervision and obsolete-engineer in the loop. AI can solve one-shot problems, AI can't maintain a system. Special thanks to Opus/Fable 5 PRs - those are not just a mess that brute-force through a whole project to achieve a specific niche goal, but also contain comments of a novel size, narrating history from the birth of the universe and the color of the mug their user was using during the synthesis. "It's so dangerous! We can't release it to the public!" - dangerous my ass, the only danger Dario's models have is ability to ignore instructions and project conventions and then kill readers with load-bearing prose that includes words no regular human spoke for basically ever. PS: sorry for venting. Working with a company that requires AI coding. I guess I need to take a vacation, enable flight mode on laptop, open vim and write some silly stuff like yet-another http router, just to find inner peace. 大家好，我只是一个普通的软件开发人员，甚至连工程学位都没有。 对于那些认为AI可以在没有监督的情况下独立完成任何有意义的事情的人，我祝他们在“一个非常温暖的地方”（地狱）度过美好的永恒。 每次我收到AI生成的PR（拉取请求）时，我连看都不想看，只想立刻拒绝。 “什么？”你可能会问，“前沿模型不是很智能吗？”——哦，是的，它们绝对是！然而问题在于，纯粹的智能对于狭窄、具体的目标非常有效。但在没有监督，没有那些“过时的工程师”参与的情况下，你不可能让AI独自维护一个大型项目。AI能解决一次性的问题，但AI无法维护一个系统。 特别感谢Opus/Fable 5的PR——那些玩意儿不仅仅是通过暴力破解整个项目来实现特定利基目标的烂摊子，还包含了长篇小说一样体量的注释，甚至讲述了从宇宙诞生起的历史，以及用户在合成代码时所用杯子的颜色。 “太危险了！我们不能向公众发布！”——危险个屁，Dario模型唯一的危险就在于它们会忽略指令和项目规范，然后用一大堆包含普通人几乎永远不会说的词汇的承重墙一般的散文把阅读者折磨致死。 PS：抱歉发泄了一下。我目前在一家要求使用AI编码的公司工作。我想我需要休个假，把笔记本电脑调到飞行模式，打开vim，写一些类似于“另一个HTTP路由器”的无聊玩意儿，只为了寻找内心的平静。 g9robot likes: 8 Hey dad, is China winning? 嘿，老爸，中国赢了吗？ Wyciorek likes: 3 Does not matter, after 2 decades of propaganda, you can just say the magic words 'China will win' and Pavlov's dogs immediately start salivating. What it will win? Is it a bad thing? How the 'winning' conditions are even defined? Does not matter 无所谓了，经过了20年的宣传，你只要说出那句神奇的咒语“中国会赢”，巴甫洛夫的狗就会立刻开始流口水。 会赢什么？这是件坏事吗？连“赢”的条件是如何定义的？都不重要了。 Commercial_Spray4279 likes: 6 What does winning mean? China giving everybody the weights to the best model they have? How is that a bad thing? it's only a bad thing if you invested trillions and are afraid of not getting your money back. 赢意味着什么？中国把他们最好模型的权重免费送给大家吗？这怎么会是坏事？只有当你投资了数万亿美元并且害怕收不回成本时，这才是坏事。 wutface0001 likes: 3 winning means big economic recession in US, rest of the world won't really be affected that much, from one king to another 赢意味着美国将陷入严重的经济衰退，世界其他地方并不会受到太大的影响，无非是从一个国王换到另一个国王罢了。 Wyciorek likes: 1 As we have seen in 2008, big economic recession in US actually affects rest of the world quite a bit. 正如我们在2008年所看到的，美国的大规模经济衰退实际上对世界其他地区的影响相当大。 wutface0001 likes: 2 yes because USD is worlds reserve currency. but still 2008 is a bad example because it caught everyone off guard, China's dominance will be a slow but more predictable process and countries will slowly start to transition, many countries are already trying hoard as much gold as possible 是的，因为美元是世界储备货币。但这仍然说明2008年是个不恰当的例子，因为那次危机让所有人都措手不及，而中国占据主导地位将是一个缓慢但更可预测的过程，各国也会开始慢慢过渡，许多国家已经开始尝试尽可能多地囤积黄金了。 severalsmallducks likes: 6 I am fully convinced that like 80% of what these CEOs are doing is a weird mix of hype and fear mongering ("Either China wins or we will cause the apocalypse") in order to inflate stock price. Literally FOMO on a global scale. 我完全确信，这些CEO所做的事情中，大约有80%都是炒作和散布恐慌的怪异混合体（“要么中国赢，要么我们将引发世界末日”），目的就是为了推高股价。 简直就是全球范围内的错失恐惧症。 FormalAd7367 likes: 2 Good point..you are right. I recall certain President was yapping about China and now is bombing other country 说得好……你是对的。我记得某位总统过去总是在那喋喋不休地谈论中国，而现在却在轰炸别的国家。 brownhotdogwater likes: 1 Well yea, he needs to sell a mega ipo so the investors get paid out 嗯，是的，他需要搞一个大规模的IPO，这样投资者才能拿到回报。 KristofVD likes: 1 I agree. AI development seems to have stalled lately, and they're trying to excuse that by saying they're holding back progress for the sake of humanity. I don't believe that shit. 我同意。最近AI的发展似乎停滞不前了，他们试图借口说为了全人类而放缓了进展。我才不信那些鬼话。 Donglemaetsro likes: 1 Yup. cosplaying clowns actually think they look cool. 是的。这些玩角色扮演的小丑居然还觉得自己看起来很酷。 anengineerandacat likes: 1 They want to push the government to give them money, China being the threat will help but it'll come at a cost they don't likely fully realize. 他们想施压政府给他们投钱，“中国威胁论”确实能帮上忙，但这会带来他们可能并未完全意识到的代价。 No-Sandwich-2997 likes: 3 more corny than sam btw 顺便说一句，比Sam还老套。 StatisticianIll5214 likes: 3 I for one certainly wouldn't mind China winning over these dumbasses 就我个人而言，我当然不介意中国战胜这帮蠢货。 Square_Height8041 likes: 2 He sounds more like a politician everyday. AGI being meter billed by this guy is not the dream 他现在听起来越来越像个政客了。让这家伙来对AGI按流量计费可不是什么梦想。 RedLucky2b2g likes: 2 China will destroy the racist US companies and colonialist greedy Anthropic 中国将摧毁那些种族主义的美国公司和殖民主义、贪婪的Anthropic。 Additional-Sky-7436 likes: 2 "OMG!!!! What is China WINS!!!! BETTER GIVE ME EVERYTHING I WANT!!!" “我的天啊！！！！如果中国赢了怎么办！！！！最好把我要的一切都给我！！！” Haipul likes: 3 I like claude as a product but then I see and hear this man and I lose all respect for the product 我喜欢Claude这个产品，但当我看到并听到这个人的言论时，我对这个产品的所有尊重就荡然无存了。 Relative-Delivery469 likes: 2 Anthropic is a scam, even regular chats deplete your limits. Anthropic就是个骗局，甚至连普通的聊天都会耗尽你的额度。 Brilliant-Muffin-879 likes: 1 Wow get fucked by an american company or get fucked by a Chinese company? HMMmm decisions decisions 哇，是被美国公司坑，还是被中国公司坑？嗯……真是个艰难的决定。 Vorapp likes: 1 love this drama queen preaching shit to regards 喜欢看这个戏精向弱智们说教那些狗屁不通的东西。 Resident_Citron_6905 likes: 1 and cultists 还有狂热的追随者。 Mountain_Patience231 likes: 1 its not like China or america win anything, he just tried to petend a patriot and win his own war 这并不像中国或美国能赢什么东西，他只是想假扮成一个爱国者，去打赢他自己的战争而已。 Bat_002 likes: 1 Build too fast - anthropic loses control 发展太快——Anthropic就会失控。 reeejuuu likes: 1 Translation: if you don’t give us monis and everything, China will win 翻译一下：如果你们不给我们钱和一切，中国就会赢。 charmander_cha likes: 1 A China vencer, a humanidade vence. Se qualquer um destes sociopatas vencerem, teremos somente eugenia 2.0 中国赢了，人类就赢了。 如果这些反社会者中的任何一个赢了，我们得到的只会是优生学2.0。 RefrigeratorDry2669 likes: 1 Oh all these billions! Oh I only have mere billions to wipe me tears with! Feel me humanity for I suffer for you! 哦，这几十亿啊！哦，我只能用区区几十亿的钞票来擦干我的眼泪！感受我吧，人类，因为我正在为你们受苦！ Whole_Move7660 likes: 1 I wish Apple be like that. China making really good phones? Apple: better 10% battery 10% camera full 1500 price. We need competition. China cars are advance too now. 我希望苹果也能这样。中国在制造非常好的手机？苹果：电池提升10%，摄像头提升10%，售价高达1500美元。我们需要竞争。中国汽车现在也非常先进了。 marky_Rabone likes: 1 Suena peligroso, hay alguien al volante o solo es idiota naranja ? 听起来很危险，方向盘后边有人在管吗，还是说只是个“橙色白痴”？ Asiel_Stormfury likes: 1 Reminds me of an old Civilization quote... "beware the man who dares control the free flow of information, for in his heart he dreams himself your master" 让我想起了《文明》游戏里的一句老话…… “当心那个敢于控制信息自由流动的人，因为在他的内心深处，他梦想着成为你的主人。” VarietyMage likes: 1 There is a solution. \\\\#BanAI. 有一个解决办法。 \\\\#封禁AI。 KaliguIah likes: 1 scam altman and dario are two sides of the same shit coin 骗子奥特曼和达里奥就是同一枚臭硬币的正反面。 mrbadface likes: 1 We are in an arms race whether you believe it or not 无论你信不信，我们正处于一场军备竞赛中。 small_bird_loud likes: 1 I have an infinitely small violin for him. 我只能给他拉一曲世界上最小的小提琴了。 definit3ly_n0t_a_b0t likes: 1 What is so fucking bad about China winning, for reals? Why are we exhausting ourselves competing, when we're obviously not winning, we've sacrificed so much already, and winning would take sacrificing more. What do we even "win?" What's the fucking point already? 中国赢了到底他妈的有什么不好，说真的？当我们显然赢不了的时候，为什么还要耗尽精力去竞争？我们已经牺牲了那么多，而要赢的话还需要牺牲更多。我们到底能“赢”得什么？这他妈的到底有什么意义？ DarkISO likes: 1 Because they keep pushing the "China= bad/evil" narrative. Have been since i can remember... 因为他们一直在兜售“中国=坏/邪恶”的叙事。从我记事起就一直如此…… CowBoyDanIndie likes: 1 Unless Chinas AI can solve the problem of feeding their population in the next \\\\~20 years when global warming caused crop failures start causing worldwide food shortages I don’t see the point. 除非中国的AI能够解决未来约20年内因全球变暖导致农作物歉收、进而引发全球粮食短缺时他们的人口吃饭问题，否则我看不出这有什么意义。 definit3ly_n0t_a_b0t likes: 1 No county on the planet is better positioned to survive mass crises than China. 在这个星球上，没有哪个国家比中国更具备在重大危机中生存下来的优势了。 FreshLiterature likes: 1 Lol dude is delusional. The reality is China appears to have adopted a pretty clear strategy of letting the US throw massive amounts of money at an idiotic end. China can keep pumping out better "good enough" models. Eventually hardware to run models locally is going to catch up and most of this data center buildout will become worthless when it does. You can already run highly capable models locally it's just very expensive to buy the hardware. As hardware becomes more and more capable that math will get better and better. The US is basically conducting economic warfare on itself and this idiot doesn't even realize it 哈哈，这家伙真是妄想症发作了。 现实情况是，中国似乎采取了一个非常明确的策略，就是让美国把大把的钞票砸向一个愚蠢的结局。 中国可以继续不断推出更好、且“足够好用”的模型。 最终，用于本地运行模型的硬件性能会迎头赶上，而当那一天到来时，绝大多数建好的数据中心将变得毫无价值。 你现在就已经可以在本地运行高性能模型了，只是购买硬件非常昂贵而已。 随着硬件变得越来越强大，这笔经济账会变得越来越划算。 美国基本上是在对自己发动经济战，而这个白痴甚至都还没有意识到这一点。 Shiplate8044 likes: 1 Frankenstein talking about his monster. 弗兰肯斯坦在谈论他的怪物。 thebeepboopbeep likes: 1 This whole “or else China wins” and “beat China” thing is so stupid. Wins what? How exactly is this a nationalized race towards what exactly? Who can make the biggest pile of slop while burning energy and losing jobs? What exactly do we win? 这整个“否则中国就会赢”和“击败中国”的论调真是太蠢了。赢什么？这到底怎么就变成了一场为了某种未知目标的国家级竞赛了？看谁能在燃烧能源和导致失业的同时，制造出最大一堆垃圾废料吗？我们到底能赢得什么？ Macaroonplastic1036 likes: 1 The BS is on another level 这瞎扯的功力已经到了另一个层次了。 ClaimApprehensive754 likes: 1 Blah blah, can we go back to the times when CEOs didn’t endlessly yap so much? 吧啦吧啦的，我们能不能回到CEO们不这么没完没了瞎逼逼的时代？ Ambitious_Power_7957 likes: 1 He’s trying to replicate Musk’s persona: initially winning public trust with appeals to “humanitarian sentiment,” then acting without restraint once he’s wealthy. His rhetoric mirrors Musk’s as well, while they’re both narcissists. The gap between Chinese models and leaders has narrowed to the point where it could be closed in a single upxe cycle. Yet, he believes he controls this trajectory. Technological progress doesn’t require his control; if he fails, China will take the lead. 他正试图复制马斯克的人设：最初通过诉诸“人道主义情感”来赢得公众的信任，然后一旦有了钱就肆无忌惮地行事。他的言辞也与马斯克如出一辙，毕竟他们都是自恋狂。中国模型和领先者之间的差距已经缩小到了在一个更新周期内就能弥合的程度。然而，他却相信自己控制着这一发展轨迹。技术进步并不需要他的控制；如果他失败了，中国就会取得领先。 uniquelyavailable likes: 1 Do they realize how amateur they are in this game, I don't think they can catch up. China is very far ahead. 他们有没有意识到他们在这场游戏里有多业余，我不认为他们能追得上。中国已经遥遥领先了。 solemnhiatus likes: 1 This mf obsessed with China 这个混蛋对中国有着病态的执念。 ElJefe6767 likes: 1 China already won. Build this shit with ethics 中国已经赢了。带着道德来开发这玩意儿吧。 gentmick likes: 1 He keeps saying China bad but China is the only one coming up with cheap open weight models that might help democratize ai… 他一直说中国不好，但中国是唯一一个提出廉价开放权重模型，从而可能帮助实现AI民主化的国家…… egowritingcheques likes: 1 Even if this knife exists I just don't think many people even care anymore. AI just isn't doing much... it's getting boring. 哪怕这把“刀刃”真的存在，我认为也没有多少人会关心了。AI根本没做什么大事……它变得越来越无聊了。 Subject-Building1892 likes: 1 Every six months the last three years Dario tells us to shit our pants because artificial super intelligence is at most 12 months away. First of all fix opus 5. 在过去的三年里，每隔六个月达里奥就会吓唬我们，让我们因为人工超级智能最多还有12个月就要到来而吓得尿裤子。 首先还是先把Opus 5修好吧。 关键词： Anthropic CEO Claude 中国 发展 相关推荐译文 国外网友讨论：印度能否取代中国成为世界工厂？ 印度能像中国一样发展吗？鲁奇尔·夏尔马为您解读 中国最畅销的车不是 SUV 也不是特斯拉，西媒讨论吉利E2 中国迅速扩大抹茶产业，日本担心守不住品牌价值。网友：本来不就是从中国来的吗？ 为什么在德国生产一辆汽车要比在中国贵30–45%？ 中国人把白人称为“白鬼”的情况常见吗？ 在中国这座小城一次寻常的遛狗，最后演变成了一段截然不同的经历…… 中国生物科技的崛起：蒂姆·斯科特在上海的所见所闻