At the end of last month, President Donald Trump summoned tech giants to Washington for a luncheon and held a ceremonial signing of a voluntary safety agreement.
上月末,唐纳德·特朗普总统将科技巨头们召集到华盛顿参加午餐会,并举行了一项自愿安全协议的仪式性签署。
He called the agreement "almost a constitution," centered on "superintelligence"—the administration's new favored term for AI.
他称该协议“几乎是一部宪法”,其核心围绕“超级智能”——这是本届政府对AI的新偏爱术语。
On Wednesday, in a crowded event space in San Francisco, safety challenges were not people's foremost concern.
周三,在旧金山一处拥挤的活动空间里,安全挑战并非人们最关心的问题。
Founders and funders said that the government's hands-off attitude would bring higher profits to companies building AI-driven defense and security systems.
创始人和资助者表示,政府的放手态度将为构建人工智能驱动的国防和安全系统的公司带来更高利润。
At a conference aimed at "investing in the frontier," panelists expressed appreciation for the government's bullish support of the industry.
在一场旨在“投资前沿”的会议上,小组成员对政府给予该行业的看涨支持表示赞赏。
This was one of the pop-up events organized by Andreessen Horowitz.
这是安德森·霍洛维茨组织的快闪活动之一。
The head of this venture capital firm is Trump's science and technology advisor.
这家风险投资公司的负责人是特朗普的科学与技术顾问。
"I'd like to begin with a figure that I believe truly underscores the investment opportunity in defense, and that is one trillion dollars.
“我想从一个我认为真正凸显国防投资机遇的数字开始,那就是1万亿美元。
For the first time, the US government is allocating one trillion dollars to defense spending, of which 13.5 billion dollars will go toward autonomous systems and AI," said Kyle Stanford, director of venture capital research at the market intelligence platform PitchBook.
美国政府首次将1万亿美元用于国防开支,其中135亿美元将投入自主系统和AI,”市场情报平台PitchBook的风险投资研究总监凯尔·斯坦福说。
"This clearly brings a massive amount of capital into defense tech spending, more than we have ever seen before."
“这显然为国防科技支出带来了大量资本,比我们以往见过的都多。”
A day later, three researchers at OpenAI's headquarters said that the ChatGPT maker had fired them, claiming that "fear and unclear rules are hindering AI safety work."
一天后,OpenAI总部的三名研究人员表示,这家ChatGPT制造商解雇了他们,并称“恐惧和不明确的规则阻碍了AI安全工作”。
In addition, David Robinson, the head of OpenAI's safety team, who only recently resigned, expressed concern about allowing advanced laboratories to regulate themselves.
此外,OpenAI安全团队负责人大卫·罗宾逊直到最近才辞职,他表示担心让先进实验室自我监管。
"This is no substitute for hard law, and it does give companies a lot of freedom," Robinson said.
“这不能替代硬性法律,而且它确实给了公司很多自由,”罗宾逊说。
"The time between model generations is measured in weeks... I think we face some very serious safety challenges in the near term."
“模型代际之间的时间以周计……我认为我们在近期面临一些非常严峻的安全挑战。”
Trump has staked his political capital on the success of the AI industry, calling its unfettered development the key to winning the AI race against China.
特朗普将其政治资本押注于AI行业的成功,称其无拘无束的发展是赢得对华AI竞赛的关键。
On the first day of his second term, the Republican president repealed a Biden-era executive order that was partly intended to guard against the risks AI systems pose to consumers, workers, and national security.
在第二个任期的第一天,这位共和党总统废除了拜登时代的一项行政命令,该命令部分旨在防范AI系统对消费者、工人和国家安全构成的风险。
After that, his administration shifted the focus of federal policy from regulation to rapid growth, including the removal of a previous requirement that AI developers share safety test results with the federal government.
此后,其政府将联邦政策的重心从监管转向快速增长,包括取消此前要求AI开发者与联邦政府分享安全测试结果的规定。
In December, Trump signed an executive order aimed at blocking state-level AI regulations, saying that a patchwork of rules could stifle the industry.
12月,特朗普签署了一项行政命令,旨在阻止州级AI法规,称拼凑的规则可能扼杀该行业。
Concerns about AI safety have largely centred on existential risks rather than present-day, real-world harms, and this became the focus of debate last month when a relatively unknown Anthropic engineer resigned and called for a pause in technological development to prevent "superhuman" systems from slipping out of control.
对AI安全的担忧主要集中在生存风险上,而非当前现实世界的危害,这在上个月成为了辩论的中心,当时一位相对不知名的Anthropic工程师辞职,并呼吁暂停技术开发,以防止“超人类”系统摆脱控制。
Subsequently, the chief executives of both Anthropic and OpenAI echoed calls to slow down AI development.
随后,Anthropic和OpenAI的首席执行官也附和了放缓AI开发的呼吁。
To some critics, their public pronouncements appeared designed to court investors ahead of a public listing and the midterm elections, when the political tide could turn and many voters oppose data centres.
对一些批评者来说,他们的公开宣言似乎旨在公开上市和中期选举前讨好投资者,届时政治风向可能转变,许多选民反对数据中心。
"The only control or 'guardrail' AI needs is a strong and smart (high IQ!) president, and the United States has exactly that, and then some!" Trump responded on Truth Social, while warning voters not to "kill the golden goose" — a term he has used to refer to AI and data centers.
“AI唯一需要的控制或‘护栏’,不过是一位强大而聪明(高智商!)的总统,而美国恰恰拥有这一点,并且绰绰有余!”特朗普在Truth Social上回应道,同时警告选民不要“杀死金鹅”——他曾用这个词指代AI和数据中心。
The document has been jointly signed by representatives from OpenAI, Anthropic, Meta, Google, SpaceXAI, and Nvidia.
该文件已由OpenAI、Anthropic、Meta、谷歌、SpaceXAI以及英伟达的代表共同签署。
Following Anthropic's new disclosure, it added that the Trump administration expects the company to be fully transparent, stating that notification and remediation are "not optional.
在Anthropic的新披露之后,它补充说,特朗普政府期望该公司完全透明,称通知和补救“不是可选项。
This is a critical national security obligation."
这是一项关键的国家安全义务。”
Under Democratic President Joe Biden, leading AI companies also voluntarily committed to promoting the "safe, secure, and transparent development" of their models, and to that end established specific guidelines for testing and transparency reporting.
在民主党总统乔·拜登的领导下,领先的人工智能公司也自愿承诺,促进其模型的“安全、可靠和透明开发”,并为此制定了测试和透明度报告的具体指南。
Sarah Myers West, a former senior AI adviser at the Federal Trade Commission, said the current agreement lacks these elements.
前联邦贸易委员会高级人工智能顾问莎拉·迈尔斯·韦斯特表示,现行协议缺乏这些内容。
"On the whole, this self-regulatory approach has consistently done only one thing, and that is to fail.
“总的来说,这种自我监管方法始终只做一件事,那就是失败。
It has failed in any meaningful way to ensure that these companies even honestly keep their own house in order, let alone that they comply with existing laws," said Myers West, who is now co-executive director of the nonprofit AI Now, based in New York.
它未能以任何有意义的方式确保这些公司哪怕只是诚实地维持自身秩序,更不用说确保它们遵守现行法律,”迈尔斯·韦斯特说,她现任非营利组织AI Now的联合执行主任,该组织位于纽约。
"All they get from the government is leniency."
“他们从政府那里得到的只有宽限。”
After having lunch with Trump last month, the FTC said it had launched an investigation into OpenAI, Anthropic and other companies, on the grounds that their technology could pose a danger to consumers.
在上个月与特朗普共进午餐后,FTC表示已对OpenAI、Anthropic及其他公司展开调查,原因是其技术可能对消费者构成危险。
The FTC did not immediately respond to a request for comment.
FTC未立即回应置评请求。
Trump recently appointed Director of National Intelligence Jay Clayton and Federal Trade Commission Chairman Andrew Ferguson to his newly formed "Superintelligence Task Force," a move the president says will sustain America's dominance in artificial intelligence development.
特朗普近日任命国家情报总监杰伊·克莱顿和联邦贸易委员会主席安德鲁·弗格森加入其新组建的“超级智能部队”,总统称此举将维持美国在人工智能开发中的主导地位。
In a post on Thursday, he stated that the White House would regard anyone who does not use his preferred terminology to refer to AI as an "enemy!"
他周四发帖称,白宫将把任何不使用他偏爱的术语来指代AI的人视为“敌人!”
Graham Steele, a former Assistant Secretary at the U.S. Treasury Department, said that even if AI companies set up internal assessment teams or hire external evaluators, such an arrangement could leave them vulnerable to undue pressure on their work.
前美国财政部助理部长格雷厄姆·斯蒂尔表示,即便AI公司设立内部评估团队或聘请外部评估者,这种结构也可能使其容易受到对其工作的不当压力。
"We live in a democratic society.
“我们生活在民主社会。
Are they truly fit to hold power and make decisions that carry broad structural, social, and political consequences?" Steele said.
他们是否真的适合掌权,去做出那些会带来广泛结构性、社会和政治后果的决定?”斯蒂尔说。