
近日,OpenAI首席执行官山姆·奥特曼(Sam Altman)与美国总统唐纳德·特朗普(Donald Trump)以及软银集团(SoftBank)和甲骨文(Oracle)的领导人一道,大力宣传“星际之门项目”。该项目预计将在美国投资5000亿美元建设数据中心,以满足未来几年人工智能使用量大幅增长的预期。“星际之门项目”将获得OpenAI、软银集团、甲骨文和阿联酋人工智能投资者MGX总计1000亿美元的先期投资,奥特曼称其为“这个时代最重要的项目”。
无论你是否认同他的观点,“星际之门项目”都可以说是科技行业有史以来最大的一场豪赌。毕竟,除了令人咋舌的高昂成本和天文数字般的能源需求(可能与整个城市的电力需求相当),这笔巨额投资毫无回报保障。鉴于当今的人工智能技术尚处于通用技术的起步阶段,如何从这等规模的人工智能中获取盈利,无人能给出确切答案。而且,尽管OpenAI可能认为“星际之门项目”对于开发“惠及全人类”的通用人工智能至关重要,但事实是,关于通用人工智能,业界甚至尚未形成公认的定义(最常见的定义是能在某些关键任务上与人类相匹敌的人工智能)。即便达成了共识,宾夕法尼亚大学沃顿商学院(University of Pennsylvania’s Wharton School)管理学教授伊桑·莫利克(Ethan Mollick)也在X平台上指出,“对于大多数人而言,一个拥有通用人工智能的世界究竟会是何种模样,至今仍缺乏清晰的愿景”。对于那些认为通用人工智能即将到来的人来说,他写道:“5到10年后,日常生活会是什么样子?”
多年来,科技领域的其他高风险投资无论是从耗资之巨还是不确定性之高来看,皆难以与“星际之门项目”相提并论:二战期间为研制原子弹而开展的曼哈顿计划改变了历史。然而,为该项目提供支持的是政府,而非私营企业,而且该项目还有一个优势,那就是建立在众所周知的科学基础上。另一方面,人工智能领域的创新者则是在押注一个无人完全理解的结果。
另一个例子是科技公司在云计算基础设施上投入数千亿美元。与人工智能不同,进军云计算服务有明确的商业案例,而且资金投入已持续十多年之久。与此同时,Meta对元宇宙(即虚拟世界)的痴迷以500亿美元的失败告终。但这不过是首席执行官马克·扎克伯格(Mark Zuckerberg)短暂分心而已。
当然,还有互联网泡时期,其中有成功也有失败。但那是一场全行业的豪赌,不像“星际之门项目”那样风险集中。
当然,这些在人工智能领域进行最新一轮豪赌的科技公司无疑拥有雄厚的财力作为支撑。它们高达数万亿美元的估值以及来自投资者的空白支票,更不必提及来自州、地方和联邦政府的财政激励和补贴,都让这场豪赌变得容易一些。毕竟,它们的商业使命就是追逐科技领域最新、最前沿的技术。
尽管如此,“星际之门项目”的赌注还是空前巨大的,因为奥特曼和特朗普不仅将其视为一项技术飞跃,更将其视为国家的当务之急。他们将其描述为能巩固美国在人工智能领域领先中国地位的项目,承诺创造10万个新就业岗位,并极大地推动经济发展。特朗普甚至称其为美国“黄金时代”的开端,而甲骨文(Oracle)执行董事长拉里·埃里森(Larry Ellison)声称,该项目有望在癌症治疗方面带来突破。
但并非所有人都相信这种炒作。加里·马库斯(Gary Marcus)等批评人士认为,人工智能的变革潜力被极大地夸大了,他警告称,在大规模过度投资之后,美国经济或将面临严重后果。事实上,当“星际之门项目”在4月首次宣布时,马库斯称其为“史上第二糟糕的人工智能投资”——仅次于过去十年间投入数十亿美元但成果寥寥的自动驾驶汽车。另一些人,比如人工智能研究先驱约书亚·本吉奥(Yoshua Bengio),则持更为悲观的看法,他们认为,人工智能非但不会带来繁荣,反而会如此深刻地重塑世界,以至于对人类的生存构成威胁。
开源人工智能平台Hugging Face的政策研究员阿维吉特·戈什(Avijit Ghosh)从另一个角度强调了以下事实——像“星际之门项目”这样不受限制的资金注入,将权力集中在最富有的人手中,而将公众和独立研究人员排除在外。此外,他表示,所有对构建基础设施以推动通用人工智能发展的关注,都损害了那些“并未致力于构建通用人工智能(无论其确切含义究竟为何)”的人的利益。“我们把资源投入到这个定义尚且模糊的‘事物’上,却忽视了当下可以利用技术解决的真正危机。”
考虑到这些批评意见,“星际之门项目”可以被视为一项“登月计划”般孤注一掷的实验,它不仅会在失败时产生重大影响,而且如果真的取得成功,也会带来严重后果。虽然OpenAI、谷歌(Google)和Meta等公司拥有采取如此大胆行动的财力,但这些风险可能并不符合公众的最佳利益。
如果考虑到美国与中国的竞争,或许“星际之门项目”所冒的风险是值得的。掌握最先进人工智能技术的国家将在经济实力和国防方面拥有巨大优势。
就在两天前,中国初创公司DeepSeek发布了一款全新的开源人工智能模型,这一举动引起了硅谷的警觉。该公司声称,其新模型在多项数学、编码和推理基准测试中的表现超越了OpenAI最先进的o1模型。
咨询公司Futurum Group的分析师迪翁·欣奇克利夫(Dion Hinchcliffe)表示,此次发布对OpenAI和人工智能行业其他公司来说是“真正的当头一棒”。他表示,中国能够研发出与OpenAI最顶尖模型相抗衡的前沿技术,这“令人担忧”。欣奇克利夫解释说:“这是一场真正的国际竞争。”
特朗普总统于周一就职后数小时内,就废除了拜登政府在人工智能监管方面所做的努力,这其中就包括了拜登于2023年颁布的有关人工智能的行政命令。特朗普的计划是尽可能减少人工智能开发过程中所面临的障碍,以期在亲商环境中加快人工智能创新。
但至关重要的是,至少要认识到这是一场高风险的博弈。“星际之门项目”与监管的放松相结合,对OpenAI、大型科技公司,甚至特朗普而言,都是一场可能带来巨大胜利的豪赌。在美国的竞争对手不断加大赌注的时代,这也可能被视为一场必要的较量。但批评人士指出,我们应该承认,我们所有人——其中许多人既对ChatGPT感到惊叹,又对终结者式的未来感到恐惧——可能都对即将发生的事情毫无准备。
Hugging Face 公司的戈什说:"我确实担心,很多人都在关注构建代理型人工智能,或是赋予人工智能模型驱动的系统某种程度的自主权。这会带来很多未知风险。”
公众对这些风险毫无准备。布伦戴奇今天在X平台上指出:“人工智能公司对于以必要的速度和规模发展,进而使社会能够做好应对准备,几乎毫无兴趣,因为它们正忙于相互竞争,并应对复杂的政治环境。”他说,记者、学者和公民社会“需要填补这一空白”。
我们可以将“星际之门项目”和其他大型人工智能项目视为大型科技公司最大的赌博,但无论我们是否愿意,这实则是一场我们所有人都孤注一掷的豪赌。也许是时候确保我们真正了解其中的利害关系了。(财富中文网)
译者:中慧言-王芳
近日,OpenAI首席执行官山姆·奥特曼(Sam Altman)与美国总统唐纳德·特朗普(Donald Trump)以及软银集团(SoftBank)和甲骨文(Oracle)的领导人一道,大力宣传“星际之门项目”。该项目预计将在美国投资5000亿美元建设数据中心,以满足未来几年人工智能使用量大幅增长的预期。“星际之门项目”将获得OpenAI、软银集团、甲骨文和阿联酋人工智能投资者MGX总计1000亿美元的先期投资,奥特曼称其为“这个时代最重要的项目”。
无论你是否认同他的观点,“星际之门项目”都可以说是科技行业有史以来最大的一场豪赌。毕竟,除了令人咋舌的高昂成本和天文数字般的能源需求(可能与整个城市的电力需求相当),这笔巨额投资毫无回报保障。鉴于当今的人工智能技术尚处于通用技术的起步阶段,如何从这等规模的人工智能中获取盈利,无人能给出确切答案。而且,尽管OpenAI可能认为“星际之门项目”对于开发“惠及全人类”的通用人工智能至关重要,但事实是,关于通用人工智能,业界甚至尚未形成公认的定义(最常见的定义是能在某些关键任务上与人类相匹敌的人工智能)。即便达成了共识,宾夕法尼亚大学沃顿商学院(University of Pennsylvania’s Wharton School)管理学教授伊桑·莫利克(Ethan Mollick)也在X平台上指出,“对于大多数人而言,一个拥有通用人工智能的世界究竟会是何种模样,至今仍缺乏清晰的愿景”。对于那些认为通用人工智能即将到来的人来说,他写道:“5到10年后,日常生活会是什么样子?”
多年来,科技领域的其他高风险投资无论是从耗资之巨还是不确定性之高来看,皆难以与“星际之门项目”相提并论:二战期间为研制原子弹而开展的曼哈顿计划改变了历史。然而,为该项目提供支持的是政府,而非私营企业,而且该项目还有一个优势,那就是建立在众所周知的科学基础上。另一方面,人工智能领域的创新者则是在押注一个无人完全理解的结果。
另一个例子是科技公司在云计算基础设施上投入数千亿美元。与人工智能不同,进军云计算服务有明确的商业案例,而且资金投入已持续十多年之久。与此同时,Meta对元宇宙(即虚拟世界)的痴迷以500亿美元的失败告终。但这不过是首席执行官马克·扎克伯格(Mark Zuckerberg)短暂分心而已。
当然,还有互联网泡时期,其中有成功也有失败。但那是一场全行业的豪赌,不像“星际之门项目”那样风险集中。
当然,这些在人工智能领域进行最新一轮豪赌的科技公司无疑拥有雄厚的财力作为支撑。它们高达数万亿美元的估值以及来自投资者的空白支票,更不必提及来自州、地方和联邦政府的财政激励和补贴,都让这场豪赌变得容易一些。毕竟,它们的商业使命就是追逐科技领域最新、最前沿的技术。
尽管如此,“星际之门项目”的赌注还是空前巨大的,因为奥特曼和特朗普不仅将其视为一项技术飞跃,更将其视为国家的当务之急。他们将其描述为能巩固美国在人工智能领域领先中国地位的项目,承诺创造10万个新就业岗位,并极大地推动经济发展。特朗普甚至称其为美国“黄金时代”的开端,而甲骨文(Oracle)执行董事长拉里·埃里森(Larry Ellison)声称,该项目有望在癌症治疗方面带来突破。
但并非所有人都相信这种炒作。加里·马库斯(Gary Marcus)等批评人士认为,人工智能的变革潜力被极大地夸大了,他警告称,在大规模过度投资之后,美国经济或将面临严重后果。事实上,当“星际之门项目”在4月首次宣布时,马库斯称其为“史上第二糟糕的人工智能投资”——仅次于过去十年间投入数十亿美元但成果寥寥的自动驾驶汽车。另一些人,比如人工智能研究先驱约书亚·本吉奥(Yoshua Bengio),则持更为悲观的看法,他们认为,人工智能非但不会带来繁荣,反而会如此深刻地重塑世界,以至于对人类的生存构成威胁。
开源人工智能平台Hugging Face的政策研究员阿维吉特·戈什(Avijit Ghosh)从另一个角度强调了以下事实——像“星际之门项目”这样不受限制的资金注入,将权力集中在最富有的人手中,而将公众和独立研究人员排除在外。此外,他表示,所有对构建基础设施以推动通用人工智能发展的关注,都损害了那些“并未致力于构建通用人工智能(无论其确切含义究竟为何)”的人的利益。“我们把资源投入到这个定义尚且模糊的‘事物’上,却忽视了当下可以利用技术解决的真正危机。”
考虑到这些批评意见,“星际之门项目”可以被视为一项“登月计划”般孤注一掷的实验,它不仅会在失败时产生重大影响,而且如果真的取得成功,也会带来严重后果。虽然OpenAI、谷歌(Google)和Meta等公司拥有采取如此大胆行动的财力,但这些风险可能并不符合公众的最佳利益。
如果考虑到美国与中国的竞争,或许“星际之门项目”所冒的风险是值得的。掌握最先进人工智能技术的国家将在经济实力和国防方面拥有巨大优势。
就在两天前,中国初创公司DeepSeek发布了一款全新的开源人工智能模型,这一举动引起了硅谷的警觉。该公司声称,其新模型在多项数学、编码和推理基准测试中的表现超越了OpenAI最先进的o1模型。
咨询公司Futurum Group的分析师迪翁·欣奇克利夫(Dion Hinchcliffe)表示,此次发布对OpenAI和人工智能行业其他公司来说是“真正的当头一棒”。他表示,中国能够研发出与OpenAI最顶尖模型相抗衡的前沿技术,这“令人担忧”。欣奇克利夫解释说:“这是一场真正的国际竞争。”
特朗普总统于周一就职后数小时内,就废除了拜登政府在人工智能监管方面所做的努力,这其中就包括了拜登于2023年颁布的有关人工智能的行政命令。特朗普的计划是尽可能减少人工智能开发过程中所面临的障碍,以期在亲商环境中加快人工智能创新。
但至关重要的是,至少要认识到这是一场高风险的博弈。“星际之门项目”与监管的放松相结合,对OpenAI、大型科技公司,甚至特朗普而言,都是一场可能带来巨大胜利的豪赌。在美国的竞争对手不断加大赌注的时代,这也可能被视为一场必要的较量。但批评人士指出,我们应该承认,我们所有人——其中许多人既对ChatGPT感到惊叹,又对终结者式的未来感到恐惧——可能都对即将发生的事情毫无准备。
Hugging Face 公司的戈什说:"我确实担心,很多人都在关注构建代理型人工智能,或是赋予人工智能模型驱动的系统某种程度的自主权。这会带来很多未知风险。”
公众对这些风险毫无准备。布伦戴奇今天在X平台上指出:“人工智能公司对于以必要的速度和规模发展,进而使社会能够做好应对准备,几乎毫无兴趣,因为它们正忙于相互竞争,并应对复杂的政治环境。”他说,记者、学者和公民社会“需要填补这一空白”。
我们可以将“星际之门项目”和其他大型人工智能项目视为大型科技公司最大的赌博,但无论我们是否愿意,这实则是一场我们所有人都孤注一掷的豪赌。也许是时候确保我们真正了解其中的利害关系了。(财富中文网)
译者:中慧言-王芳
OpenAI CEO Sam Altman joined President Donald Trump and leaders of SoftBank and Oracle yesterday to tout Stargate, a $500 billion plan to build data centers in the U.S. to power the expected soaring use of AI in the coming years. Altman called Stargate, which will get an up-front investment of $100 billion from OpenAI, SoftBank, Oracle, and the Emirati AI investor MGX, the “most important project of this era.”
Whether or not you agree with him, Stargate is arguably the tech industry’s biggest gamble ever. After all, in addition to the eye-popping price tag and the astronomical energy needs (possibly rivaling the electricity demands of entire cities), the massive investment has zero guarantee of return. Given that today’s AI is a generalized technology in its infancy, no one knows how to make money from it at such an enormous scale. And further, while OpenAI may believe that Stargate is “critical” to developing artificial general intelligence (AGI) that will “benefit all of humanity,” the truth is there is not even an agreed-upon definition of AGI (the most common definition is AI that’s equal to humans at certain critical tasks). And even if there was consensus, Ethan Mollick, a professor of management at the University of Pennsylvania’s Wharton School, pointed out on X that there is “still no articulated vision of what a world with AGI looks like for most people.” For those who believe AGI is coming soon, he wrote, “what does daily life look like 5-10 years later?”
Other high-stakes tech bets over the years have not been as costly, nor as wholly uncertain: The Manhattan Project, for developing an atomic bomb during World War II, changed history. However, it was the government, not private business that backed that project, which also had the advantage of being based on well-understood science. AI innovators, on the other hand, are gambling on an outcome that no one fully understands.
Another example is the tens of billions of dollars that tech companies have spent on cloud computing infrastructure. Unlike AI, the push into cloud had a clear business case and the money was invested over more than a decade. Meanwhile, Meta’s obsession with the metaverse, or virtual worlds, was a $50 billion flop. But hey, that strategy was just CEO Mark Zuckerberg’s brief distraction.
And, of course, there was the dot-com boom, which had a mix of success and failures. But it was an industry-wide bet that did not have the concentrated risk of Stargate.
Of course, the tech companies making this latest giant gamble on AI can certainly afford it. Their trillion-dollar valuations and what are practically blank checks from investors, not to mention financial incentives and subsidies from state, local, and federal government, make rolling the dice a bit easier. And their business mission, after all, is going after the latest and the greatest in tech.
Still, the stakes with Stargate are exceptionally high, as both Altman and Trump frame it not just as a technological leap, but as a national imperative. They present it as a project that will solidify U.S. leadership over China in AI, promising 100,000 new jobs and a major economic boost. Trump has even called it the dawn of a “golden age” for America, while Oracle executive chairman Larry Ellison claims it could lead to breakthroughs in treating cancer.
But not everyone is buying the hype. Critics like Gary Marcus argue that AI’s transformative potential is vastly overstated, warning that the U.S. economy will be left holding the bag after a massive overinvestment. In fact, when Stargate was first announced in April, Marcus said it was “the second worst AI investment in history”—after the billions of dollars plowed into self-driving cars over the past decade with little to show for it. Others, like pioneering AI researcher Yoshua Bengio, take an even darker view, believing that far from ushering in prosperity, AI could reshape the world so profoundly that it threatens humanity itself.
Avijit Ghosh, a policy researcher at open source AI platform Hugging Face, emphasizes a different angle—the fact that unrestricted funding like that going towards Stargate concentrates power in the hands of the wealthiest, while excluding the public and independent researchers. In addition, all the attention to building infrastructure to boost AGI harms people who are not “‘building AGI, whatever that means,’” he said. “We are pouring resources into this ‘thing’ that is nebulously defined at best, at the expense of real crises that can be solved with technology at the very present.”
With those criticisms in mind, Stargate can be seen as a moonshot, make-or-break experiment that will not only have significant impact if it fails, but severe consequences if it actually succeeds. While companies like OpenAI, Google, and Meta can afford to make these power moves, the risks may not be in the public’s best interest.
Or maybe the risks of Stargate are worth it, if you consider the U.S. rivalry with China. The country with the best AI has an enormous advantage when it comes to economic power and national defense. If China ends up with the most advanced AI systems, the U.S. could be in danger.
Just two days ago, a Chinese startup, DeepSeek, set off alarm bells by releasing a new open-source AI model that has Silicon Valley buzzing. The company claims its new model beats OpenAI’s most sophisticated o1 model on several math, coding, and reasoning benchmarks.
The release is a “real shot across the bow” to OpenAI and the rest of the AI industry, said Dion Hinchcliffe, an analyst with the Futurum Group, a consulting firm. China’s ability to develop a frontier-level model that competes with the best from OpenAI, he said, is “concerning.” “There’s a real international competition,” Hinchcliffe explained.
Within hours of taking office on Monday, President Trump dismantled the Biden Administration’s efforts to tackle AI regulation, including Biden’s 2023 executive order on AI. Trump’s plan is to reduce as many barriers as possible to developing AI, thereby speeding up AI innovation in a business-friendly environment.
But it’s important to at least recognize the high-stakes game at play here. Stargate, combined with reduced regulation, is a gambit that could deliver huge wins for OpenAI, Big Tech, and possibly Trump. It may also be remembered as a necessary play in an era where America’s rivals are escalating the stakes. But we should acknowledge that all of us—many of whom both marvel at ChatGPT and fear a Terminator-style future—may be woefully unprepared for what’s about to unfold, critics say.
“I do worry that a lot of focus is going into building agentic AI, or giving some level of autonomy to AI model-powered systems,” said Hugging Face’s Ghosh. “That brings forth a lot of unknown risks.”
The public is unprepared for any of those risks. Brundage pointed out on X today that “AI companies have little interest in preparing society, at the speed/scale that’s needed, since they are busy trying to beat each other and navigate a complex political environment.” Journalists, academics, and civil society, he said, “need to fill the gap.”
We can look at Stargate and other massive AI projects as Big Tech’s biggest gamble, but it’s a bet that all of us are all-in on—whether we like it or not. Maybe it’s time to make sure we really understand the stakes.