字号·· | 护眼
商业内幕

萨姆·奥尔特曼不喜欢埃隆·马斯克设想的那种人类沦为AI“快乐”宠物的未来Sam Altman doesn't like Elon Musk's vision of a future where humans are 'happy' pets to AI

点「原文对照」整页切到原文,或双击某段只看那段的原文。

OpenAI首席执行官萨姆·奥尔特曼表示,人类保持对AI的控制至关重要。OpenAI CEO Sam Altman said it is important humanity retain control over AI

OpenAI首席执行官萨姆·奥特曼与埃隆·马斯克的竞争已演变成“狗咬狗”的局面。周三,奥特曼抨击马斯克此前曾暗示,人工智能可能变得如此强大,以至于将人类降格为家庭宠物的角色。“我不想当拉布拉多犬。我不想让我的孩子当拉布拉多犬,”奥特曼在《纽约时报》Dealbook峰会的现场访谈中表示。“我希望人们能像我们过去那样,对未来拥有主导权。”

OpenAI CEO Sam Altman's rivalry with Elon Musk has gone to the dogs. On Wednesday, Altman knocked Musk for previously suggesting that AI could become so powerful that it would relegate humanity to the role of a house pet. "I don't want to be a Labrador. I don't want my kids to be Labradors," Altman said during an onstage interview at The New York Times Dealbook summit. "I want people to have the agency over the future that we have had."

7月,《经济学人》询问马斯克关于十多年前的一番言论:如果人工智能发展过快,会让人类“变得像宠物拉布拉多犬一样”。“拉布拉多犬是相当快乐的生物,这是我的体验,”马斯克在接受《经济学人》采访时一边笑着一边说道。

In July, The Economist asked Musk about a comment he made more than a decade ago that if AI advanced too quickly, it would make humans "like a pet Labrador." "Labradors are pretty happy creatures, that's my experience," Musk told The Economist while chuckling.

2015年,马斯克与天体物理学家尼尔·德格拉斯·泰森探讨了如果人工智能达到递归自我改进阶段会发生什么,即人工智能变得如此先进,以至于能在极少或无人类指导的情况下开始自我进步。人工智能领域的一些人表示,如果达到该阶段,人工智能将改进得如此之快,以至于可能迅速超越人类智能,并可能超越控制该技术的任何手段。“那就完蛋了,”马斯克重复了泰森的话。

In 2015, Musk and astrophysicist Neil deGrasse Tyson discussed what would happen if artificial intelligence reached the stage of recursive self-improvement, where AI becomes so advanced that it can begin advancing itself with minimal or no human guidance. Some in AI have said that if that stage is reached, AI will improve so rapidly that it could quickly surpass human intelligence and potentially any means of controlling the technology. "That's all she wrote," Musk said, repeating Tyson's words.

马斯克描述了他对递归自我改进的人工智能感到担忧,认为它“可能会重新编程自己变得更聪明,并极快地迭代,在数百万台计算机上每天24小时不间断地进行这种迭代”。马斯克当时谈论的内容(在ChatGPT发布近八年前)现在不那么理论化了。

Musk described being worried that a recursively self-improving AI "could reprogram itself to be smarter and iterate very quickly and do that 24 hours a day on millions of computers." What Musk talked about then (almost eight years before ChatGPT's release) is less theoretical now.

Anthropic首席执行官达里奥·阿莫代伊、奥特曼和马斯克都支持阿莫代伊9月份的呼吁,即放缓将模型推向自我改进的前沿人工智能研究,并利用这段时间寻找更有效的方法,确保模型保持在人类控制之下。奥特曼在最近几周多次表示,他希望确保行业不接受人类将控制权让渡给人工智能的任何情景。他表示,他不相信他的竞争对手也有同样的想法。

Anthropic CEO Dario Amodei, Altman, and Musk all endorsed Amodei's September call to slow frontier AI research that pushes models toward self-improvement and to use the time in between to find more effective ways to ensure the models remain under human control. Altman has repeatedly said in recent weeks that he wants to ensure the industry doesn't accept any scenario in which humans cede control to AI. He has said he is not convinced that his competitors feel the same way.

“我们都深信,人类需要掌控未来,”奥尔特曼周三在Dealbook活动上表示,OpenAI总裁格雷格·布罗克曼同台出席。“现在越来越多的声音说,要么将会、要么应该出现这种向AI的逐步移交。我们都认为这是个糟糕的主意。”

"We both believe very deeply that people need to stay in control of the future," Altman told Dealbook on Wednesday, while being joined onstage by OpenAI President Greg Brockman. "And there's increasing talk about, well, there is either going to be or there should be this sort of gradual handover to AI. And we both agree that that's a terrible idea."

虽然这番言论比他们以往的一些互怼要温和,但奥尔特曼和马斯克的恩怨由来已久,源于马斯克对OpenAI转变方向的愤怒。马斯克曾共同创办该AI实验室,并是早期资金支持者。今年5月,陪审团裁定马斯克就此事起诉奥尔特曼和OpenAI等待时间过长。马斯克有时曾表达警觉,担心人类本质上会被驯化。“我不喜欢变成家猫的想法,”马斯克2016年曾说。马斯克当时表示,防止这种结果的方法是将人脑连接到计算机。

While tamer than some of their back-and-forths, Altman and Musk have a long-running feud dating back to Musk's anger over OpenAI's shift in direction. Musk cofounded the AI lab and was an early financial backer. In May, a jury ruled that Musk waited too long to sue Altman and OpenAI over the matter. At times, Musk has expressed alarm about humanity essentially being domesticated. "I don't love the idea of being a house cat," Musk said in 2016. Musk said then that the way to prevent such an outcome was to hook human brains into computers.

阅读媒体原文 ↗