CADE METZ, KAREN WEISE2025年9月17日Sam Altman, the chief executive of OpenAI, isn’t shy about how much his company plans to spend on its quest to build artificial intelligence.OpenAI的首席执行官萨姆·奥尔特曼在谈到公司计划在构建人工智能方面的投入规模时毫不避讳。“You should expect OpenAI to spend trillions of dollars on things like data center construction in the not-too-distant future,” Mr. Altman recently said, referring to the massive computing facilities that power the company’s A.I. technologies.“可以预见，OpenAI在不久的将来会在数据中心建设等方面投入数万亿美元，”奥尔特曼近期谈及支撑AI技术的大型计算设施时表示。“You should expect a bunch of economists to wring their hands and say ‘This is so crazy. It’s so reckless’ or whatever. And we’ll just be like: ‘You know what? Let us do our thing.’”“大可预料到，一帮经济学家会忧心忡忡，说‘这太疯狂了，太鲁莽了’之类的。我们只会这样回应：‘你知道吗？让我们按自己的方式来做事情吧。’”So, what exactly is that thing? As the tech industry spends and spends, turning farmland into data centers and A.I. researchers into some of the most highly paid workers in the country, it has struggled to explain what it is building and why it is spending so much money.那么，那个“事情”到底是什么？随着科技行业不断砸钱，将农田变成数据中心，把人工智能研究员变成全国收入最高的一群人，它却难以向外界阐明自己到底在打造什么，又为什么要花这么多钱。Are they building an A.I. system as smart as humans? A godlike machine that will change the world if it doesn’t destroy humanity first? Are they working on fancier versions of software they have been selling for decades? Is all this money going into a bold plan to create fake online friends and more effective ads? Or are they just afraid of missing out on what everyone else is doing?他们是在打造一个智力可与人类媲美的人工智能系统吗？还是创造一种若不先毁灭人类，就将改变世界的神一般的机器？他们是在开发那些已经卖了几十年的软件“豪华升级版”吗？这些巨额投入是为了实现一个打造虚拟网络朋友和精准广告的大胆计划？还是说，他们只是害怕错过别人都在做的事情？Here is a rundown of the visions, from the very plausible to the fantastical and why they’re pursuing those ideas:以下是这些愿景的梳理——从最现实的设想到天马行空的宏图，以及他们试图实现这些想法的原因：The Promise: A Better Search Engine承诺：一个更好的搜索引擎Chatbots work a lot like a search engine, except they generate plain English answers rather than a list of blue links. This can be a quicker, easier and more intuitive way of answering questions, though chatbots often get things wrong and even make stuff up.聊天机器人在某种程度上就像搜索引擎，不过它们提供的是通俗易懂的自然语言答案，而不是蓝色链接的列表。这可能是一种更快捷、更简单、更直观的回答问题方式，尽管聊天机器人经常会出错，甚至凭空编造内容。Why are they building this?背后的动机是什么？Google’s search engine is the tech industry’s most profitable business. If companies could provide a better way of searching for information, they could capture a market of billions of people.谷歌的搜索引擎是科技行业最为盈利的业务。如果有公司能够提供更优质的信息检索方式，就能抢占一个用户规模达数十亿人的市场。How close is it to reality?这项技术距离实现还有多远？Hundreds of millions of people are already using chatbots to gather information. Every month, more than 700 million use ChatGPT alone.已经有数亿人在使用聊天机器人获取信息。仅ChatGPT的月活用户就突破7亿。OpenAI的ChatGPT月活用户超过7亿。But turning a profit with this technology is a challenge. Operating a chatbot is significantly more expensive than serving up an ordinary website. And the technology does not necessarily lend itself to the tried-and-true way of making money on a search engine: digital advertising.但实现盈利仍是行业难题。运行一个聊天机器人比维护普通网站的成本高得多。而且，传统搜索引擎的数字广告盈利模式未必适用于该领域。OpenAI sells a version of ChatGPT for $20 a month, and according to the company, it at least pays for the cost of its own delivery. But those subscribers account for fewer than 6 percent of the people who now use ChatGPT.OpenAI提供一个每月20美元的ChatGPT付费版本，据公司称，这至少能覆盖其自身的运营成本。但付费用户仅占用户总数的6%不到。The free version is still in the red, as OpenAI has not yet begun to experiment with ads. Google, on the other hand, generates $54 billion in ad revenue each quarter from its search engine, which is used by about 2 billion people each day.免费版本仍处于亏损状态，因为OpenAI还没有开始尝试投放广告。相比之下，谷歌的搜索引擎每天服务约20亿用户，每个季度可带来高达540亿美元的广告收入。(The New York Times has sued OpenAI and Microsoft, claiming copyright infringement of news content related to A.I. systems. The companies have denied those claims.)（《纽约时报》已起诉OpenAI和微软，指控它们在未经授权的情况下使用它的新闻内容训练人工智能系统。这两家公司对此予以否认。）The Promise: Tools That Make Office Workers More Productive (and Maybe Replace Them)承诺：提升办公效率（乃至取代人力）的工具The technology that drives ChatGPT does not just answer questions, it is a tool that can help people do their jobs. A.I. can generate computer programs, summarize documents and meetings, draft emails and even use other software applications, like spreadsheets and online calendars.驱动ChatGPT的技术不仅可以回答问题，更是提升工作效率的利器。人工智能可以生成计算机程序、总结文档和会议内容、起草电子邮件，甚至还能操作其他软件应用，比如电子表格和在线日历。Why are they building this?背后的动机是什么？Tech executives believe A.I. could transform the business world as it moves into law offices, hospitals, newsrooms and more. Companies like Microsoft and OpenAI are already pulling in ample revenue through the sale of A.I. systems that can generate computer programs.科技企业的高管们认为，随着人工智能进入律师事务所、医院、新闻编辑部等领域，它可能会重塑商业生态。微软和OpenAI等公司通过销售能够生成计算机程序的人工智能系统已经获得了可观收入。Amazon, Google, Meta, Microsoft and OpenAI plan to spend more than $325 billion combined on giant data centers this year. That is $100 billion more than the annual budget of Belgium. Over time, about 10 percent of the infrastructure will be used build A.I. technologies while 80 to 90 percent will be used to deliver these technologies to customers, according to Amazon’s chief executive, Andy Jassy.亚马逊、谷歌、Meta、微软和OpenAI计划今年在大型数据中心上的合计总投资将超过3250亿美元，比比利时一年的国家预算还多出1000亿美元。亚马逊首席执行官安迪·贾西表示，从长期来看，大约10%的基础设施将用于构建人工智能技术，而80%到90%将用于将这些技术交付给客户。亚马逊正在印第安纳州建设一个大型数据中心综合体，用于支持其与人工智能初创公司Anthropic的合作。How close is it to reality?这项技术距离实现还有多远？Many businesses are already kicking the tires on A.I. But massive rollouts across the U.S. economy are not yet happening. Unless companies like Amazon, Google and OpenAI continue to improve these technologies, adoption may be slower than expected.许多企业已经开始试用人工智能，但尚未在美国经济中实现大规模应用。除非像亚马逊、谷歌和OpenAI这样的公司能持续改进这些技术，否则人工智能的普及速度可能低于预期。Almost eight in 10 businesses have started to use generative A.I., but just as many have said it has “no significant bottom-line impact,” according to research from McKinsey & Company.根据麦肯锡公司的一项研究，近八成企业已经开始使用生成式人工智能，但同样比例的企业表示，它“未能带来显著的经济效益”。“The house of cards is going to start crumbling,” said Sasha Luccioni, a researcher at the artificial intelligence start-up Hugging Face. “The amount of money being spent is not proportionate to the money that’s coming in.”“这座纸牌屋很快就要开始倒塌了，”人工智能初创公司Hugging Face的研究员萨莎·卢西奥尼说道。“当前资金投入与商业回报完全不成比例。”The Promise: An Everything Assistant承诺：全能型数字助手Tech companies are also folding chatbot-like technology into a wide range of consumer products and services. A.I., they say, will operate like a digital assistant that pops up wherever it is needed.科技公司也正在将类似聊天机器人的技术整合进各种消费类产品和服务中。他们表示，人工智能将像一个数字助手一样，在你需要的任何地方随时出现。Meta is adding the technology into its smart glasses, letting people identify landmarks as they walk down the street and translate street signs when they visit foreign countries. Amazon sees A.I. as a way of improving everything from its shopping websites to its Alexa voice assistant.Meta正将这项技术应用到智能眼镜中，让人们在街上行走时能够识别地标建筑，出国时翻译路牌。亚马逊将人工智能视为革新其所有业务的关键——从购物网站到Alexa语音助手，皆可通过AI实现全面升级。Why Are They Building This?背后的动机是什么？If you start using a digital assistant, the company behind the bot has more ways of grabbing your attention and, ultimately, selling you things.如果你开始使用数字助手，背后的公司就有了更多方式吸引你的注意力，最终向你推销商品。So, these companies are adding A.I. technology into as many devices and online services as they can, hoping to control the way you use the internet.因此，这些公司正全力将人工智能技术植入各种设备和在线服务中，力图掌控用户的互联网使用方式。“Everything will be transformed with A.I.,” said Rohit Prasad, an Amazon senior vice president. “This is not a science project.”“一切都将被人工智能彻底改变，”亚马逊高级副总裁罗希特·普拉萨德表示。“这绝非科学实验。”How close is it to reality?这项技术距离实现还有多远？亚马逊已经在Alexa智能家居设备中植入了人工智能。 Graham Dickie/The New York TimesMeta已经在其雷朋智能眼镜中加入了人工智能技术。 Aaron Wojack for The New York TimesMeta’s A.I. glasses are still a niche product used by a few million people. Amazon’s Alexa is far more popular, but its audience is still small compared to all the computers and phones in the world.Meta的人工智能智能眼镜仍属于小众产品，仅有几百万用户。亚马逊的Alexa更受欢迎，但与全球所有电脑和手机相比，用户数量仍然有限。On its own, Alexa has been a money loser since it was first released more than a decade ago. It is used primarily to juice other products and services.自十多年前首次发布以来，Alexa本身一直处于亏损状态。它主要被用来推动其他产品和服务的发展。When Amazon rebooted Alexa with new A.I. technology, it gave the upgrade away for free to anyone who pays for its Prime membership program. A.I could make it more popular, but it is unlikely to turn into a moneymaker anytime soon.当亚马逊用新的人工智能技术重塑Alexa时，向所有Prime会员免费提供了升级版本。人工智能可能会提升它的受欢迎程度，但短期内仍难实现盈利突破。The Promise: A.I. Friends承诺：人工智能好友Meta and various start-ups, including Character.AI and Elon Musk’s xAI, are beginning to offer A.I. bots that provide a new kind of companionship. People can interact with these bots on social networks in much the same way they interact with friends.Meta以及包括Character.AI和埃隆·马斯克的xAI在内的多家初创公司开始提供一种全新的人工智能机器人陪伴服务。人们可以在社交网络上与这些机器人互动，几乎与真人社交无异。“The average person wants more connectivity, connection, than they have,” Mr. Zuckerberg said in a recent podcast interview.扎克伯格在最近一次播客采访中表示：“普通人渴望获得比他们现在拥有的更多的连接和互动。”Why are they building this?背后的动机是什么？Mr. Zuckerberg and Mr. Musk run social networks and could charge money for their virtual friends. Mr. Musk is offering his bots through a subscription services that costs $300 a month.扎克伯格和马斯克经营着社交网络平台，具备对其虚拟朋友服务进行收费的能力。马斯克已推出月费高达300美元的聊天机器人订阅服务。Meta could also charge a subscription fee for virtual friends, much like OpenAI does for ChatGPT, though Meta has long preferred to boost ad revenue by keeping people on sites like Facebook, Instagram and WhatsApp. (The company is applying A.I. in this area, too. Meta recently found that people were almost 7 percent more likely to click on ads created with new A.I. techniques.)Meta虽也可像OpenAI对ChatGPT那样收取订阅费，但它一直更倾向于通过让用户留在Facebook、Instagram和WhatsApp等平台上来提升广告收入。（该公司也在这一领域应用人工智能。Meta最近发现，使用新人工智能技术制作的广告，点击率几乎提高了7%。）How close it to reality?这项技术距离实现还有多远？Some people already treat chatbots like friends. But A.I. companionship is starting to receive heavy criticism. These technologies can pull people away from human relationships and push them toward alarmingly delusional behavior.有的人已经把聊天机器人当作朋友看待，但人工智能的陪伴也开始受到严厉批评。这些技术可能会让人们脱离真实的人际关系，甚至导致令人担忧的妄想行为。It is still years away from becoming a viable market, and is just one of many scenarios that companies are exploring.该领域距离形成成熟市场尚有数年之久，仅是科技企业探索的众多方向之一。Some observers compare what tech executives are doing to moving pieces on a game board — trying to beat their rivals to the next big technology.一些观察人士将科技巨头的布局比作棋局博弈，竞相争夺下一代技术主导权。“So much power resides in so few people,” said David Cahn, a partner with the Silicon Valley venture capital firm Sequoia, “and they are playing a chess game that has implications for all of us.”“如此巨大的权力集中在少数人手中，”硅谷风投公司红杉资本的合伙人戴维·卡恩指出，“他们正在进行的这场棋局，将影响我们所有人的未来。”The Promise: Scientific Breakthroughs承诺：科学突破Anthropic的首席执行官达里奥·阿莫代伊写了一篇关于人工智能未来的万字长文。 Massimo Berruti for The New York TimesDario Amodei, the chief executive of Anthropic, one of OpenAI’s primary rivals, believes that in just a few years — perhaps as soon as next year — artificial intelligence will be like having a “country of geniuses in a data center” that can work together in solving the biggest scientific problems that our society faces.Anthropic（OpenAI的主要竞争对手之一）的首席执行官达里奥·阿莫代伊相信，人工智能在未来几年——最快可能就在明年——就会如同在“数据中心里构建了一个天才国度”，能够协同合作，解决人类社会面临的重大科学难题。Why are they building this?背后的动机是什么？Technologists like Mr. Amodei believe this kind of technology will change life as we know it. Last year, in a 14,000-word essay, he said that A.I. could eventually cure cancer, end poverty and even bring world peace.像阿莫代伊这样的技术专家相信，这类技术将彻底改变我们所知的生活。去年，他在一篇万字长文中表示，人工智能最终可能治愈癌症、消除贫困，甚至带来世界和平。He predicted that within a decade, A.I. would double the life span of the average person to 150 years.他预测，在十年内，人工智能技术将使普通人的平均寿命翻倍，达到150岁。How close is it to reality?这项技术距离实现还有多远？It is unclear how these technologies will be built — or if they are even possible.这类技术的实现路径尚不明确，甚至其可行性仍存争议。But James Manyika, Google’s senior vice president of research, labs, technology and society, said that as Google pursues loftier goals, it will create technologies that can be used right away. As an example, he points to AlphaFold, a system developed by Google that can help accelerate drug discovery in small but important ways and recently won the Nobel Prize for chemistry.但谷歌负责研究、实验室、技术与社会事务的高级副总裁詹姆斯·曼尼卡表示，谷歌在追求更远大目标的同时，也会开发出能够立即投入使用的技术。例如，他提到了AlphaFold ——这是谷歌开发的一个系统，能够在一些虽小但关键的方面加速药物研发，最近其开发者获得了诺贝尔化学奖。A Google spinoff called Isomorphic Labs aims to make money by helping drug companies use this kind of technology.谷歌旗下的Isomorphic Labs正致力于通过协助制药企业运用此类技术实现盈利。The Promise: A.I. That’s as Smart as a Human, or Smarter承诺：媲美乃至超越人类的人工智能Executives like Mr. Zuckerberg and Demis Hassabis, the head of Google’s DeepMind research lab, say their companies are pursuing artificial general intelligence, or A.G.I., shorthand for a machine that can match the powers of the human brain, or an even more powerful technology called superintelligence.像扎克伯格和谷歌DeepMind研究实验室负责人德米斯·哈萨比斯这样的高管表示，他们的公司正致力于研发通用人工智能(AGI)，即一种能够匹敌人类大脑能力的机器，甚至更强大的技术——超级智能(superintelligence)。Why are they building this?背后的动机是什么？Many technologists are determined to chase the largest goal they can imagine: superintelligence. Technologists have pursued this dream since the 1950s.许多技术专家决心追逐他们所能想象的最大目标：超级智能。自20世纪50年代以来，技术专家们就一直在追求这个梦想。How close is it to reality?这项技术距离实现还有多远？Terms like A.G.I. and superintelligence are hard to pin down. Scientists cannot even agree on a way of defining human intelligence.像通用人工智能和超级智能这样的概念其实难以精确定义。科学家们甚至无法就“人类智能”的定义方式达成共识。But a machine that truly matches the powers of the human brain is many years away, perhaps decades or more.真正媲美人脑的机器仍遥不可及，或许还需数十年乃至更久。No one has articulated how companies will make money from this kind of technology. As tech companies spend hundreds of billions on new data centers, they are taking a leap of faith.目前还没有人能够清晰阐释此类技术的盈利路径。当科技公司斥资数千亿美元建设新的数据中心时，他们其实是在“赌一把”。This leap is fueled by the same brew that often drives the moguls of Silicon Valley, said Oren Etzioni, the founding chief executive of the Allen Institute for AI: greed, ego and fear of being unseated by an unexpected breakthrough. “If I had to give a one-word answer,” said Mr. Etzioni, “it would be FOMO.”艾伦人工智能研究所创始首席执行官奥伦·埃齐奥尼表示，这种“赌博”背后，是硅谷大佬们常见的驱动力混合体：贪婪、自负，以及对被突如其来的技术突破取代的恐惧。“如果我只能用一个词来形容，”埃齐奥尼说，“那就是FOMO（错失恐惧症）。”艾伦人工智能研究所创始首席执行官奥伦·埃齐奥尼认为，许多科技公司担心错过人工智能的机遇，即使他们不太清楚自己到底会错过什么。Fear Of Missing Out does not come cheap. Mr. Altman said that as he and his rivals chase these lofty goals, some investors may be overspending. Researchers might develop ways of building A.I. using far less hardware. People might not want the A.I. technologies these companies are building. The rapid improvement of A.I. technologies over the last several years might slow or even hit a wall. The entire economy might shift for unrelated reasons.FOMO的代价不菲。阿尔特曼表示，当他和竞争对手追逐这些远大目标时，一些投资者可能已经投入过度。研究人员或许能找到硬件需求大幅降低的AI开发方式；人们也未必真正想要这些公司正在打造的人工智能技术。过去几年中人工智能技术的飞速发展可能会放缓，甚至陷入停滞。而整个经济格局也可能因其他无关的因素而发生转变。“Some of our competitors will flame out and some will do really well, and that’s just how capitalism works,” Mr. Altman said. “I do suspect that someone is going to lose a phenomenal amount of money.”“我们的一些竞争对手会烧光资金、失败出局，另一些则会非常成功——这就是资本主义的运作方式。”阿尔特曼说。“我猜测，有人最终会亏掉多得惊人的钱。”Cade Metz撰写有关人工智能、无人驾驶汽车、机器人、虚拟现实和其他技术新兴领域的新闻。Karen Weise常驻西雅图，为时报撰写科技新闻。她主要关注美国两大科技公司亚马逊和微软的新闻。翻译：Ziyu Qing点击查看本文英文版。