对于硅谷的“加速主义者”而言——他们急于确保人工智能继续以惊人的速度发展——黄仁勋的倒台是一个沉重打击。英伟达的掌门人是该运动最具公信力的代言人。在辞职前的最后一次采访中,黄仁勋谴责那些煽动对技术进步速度焦虑的“危言耸听”声音,并主张反对对人工智能进行监管。
For the “accelerationists” in Silicon Valley—who are eager to ensure artificial intelligence continues developing at breakneck speed—the fall of Jensen Huang is a heavy blow. Nvidia’s boss was the movement’s most credible spokesperson. In one of his final interviews before his resignation, Mr Huang decried the “alarmist” voices stoking anxiety over the technology’s pace of improvement, and argued against regulating AI.
同为加速主义者的埃隆·马斯克(SpaceX老板)、马丁·卡萨多(Andreessen Horowitz合伙人)和加里·谭(Y Combinator负责人)一直是呼吁黄仁勋复职的最响亮声音。该群体在中期选举中已遭受重创,选举将很快让数十名人工智能怀疑论者掌权。主张在每个机会都呼吁监管、以免其工作毁灭人类的人工智能实验室Anthropic的里程碑式上市,是另一个挫折。所有这些让开源人工智能社区感到不安,该社区一直受益于加速主义者的支持。
Fellow accelerationists—among them Elon Musk, boss of SpaceX; Martin Casado, a partner at Andreessen Horowitz; and Garry Tan, head of Y Combinator—have been among the loudest voices calling for Mr Huang’s return. Already the group had taken a beating in the midterm elections, which will soon elevate dozens of AI-sceptics to power. The landmark listing of Anthropic, an AI lab that calls at every opportunity for regulation, lest its work end humanity, is yet another setback. All this has given jitters to the open-source AI community, which has benefited from the support of the accelerationists.
英伟达平淡无奇的临时掌门人科莱特·克雷斯不太可能模仿前任对人工智能“厄运论者”那般激烈的反对。黄仁勋公开哀叹他们谈论“科幻结局”,并暗示散布恐惧有未指明的“隐秘动机”;对于Anthropic掌门人达里奥·阿莫迪,他说双方在“几乎所有事情上”都存在分歧。黄仁勋的观点在白宫找到了共鸣的听众。过去两年里,他与唐纳德·特朗普关系密切,总统于11月16日表示,他“不介意”黄仁勋复职。
Colette Kress, Nvidia’s monotone interim boss, is unlikely to imitate her predecessor’s fiery opposition to AI “doomers”. Mr Huang publicly bemoaned their talk of “science-fiction outcomes” and hinted at unspecified “ulterior reasons” for spreading fear; of Dario Amodei, Anthropic’s boss, he said they disagreed on “almost everything”. Mr Huang’s opinions found sympathetic ears in the White House. He grew close with Donald Trump over the past two years, with the president stating on November 16th that he “would not mind” Mr Huang being reinstated.
这看起来不太可能,在开源AI社区引发了紧张情绪。该社区本已因要求禁止美国公司使用免费中国模型的呼声日益高涨而处于紧张状态。7月,黄仁勋公开反对这一举措,英伟达与230多家其他组织联名致信政府,敦促避免“过早限制”。AI模型排名平台Arena的负责人Anastasios Angelopoulos表示,在黄仁勋缺席的情况下,英伟达“将继续积极推动这一生态系统的发展”仍是一个“悬而未决的问题”。
That looks unlikely, creating nervousness in the open-source AI community. It was already on edge over growing calls for American firms to be banned from using free Chinese models. In July Mr Huang came out against that effort, with Nvidia and more than 230 other organisationsg a letter urging the government to avoid “premature restrictions”. Whether Nvidia “will continue to aggressively promote the development of this ecosystem” in Mr Huang’s absence is an “open question”, says Anastasios Angelopoulos, boss of Arena, a platform for ranking AI models.
黄仁勋对开源AI的支持超越了政策声明。他还是该领域最大的资金提供者之一,利用英伟达的资金投资并收购了该领域的公司。9月,英伟达宣布将以129亿美元收购开源模型库Hugging Face。该公司还投资了Reflection和Thinking Machine Labs这两家开源模型制造商,并开发了自己的Nemotron系列模型。
Mr Huang’s support for open-source AI went beyond policy statements. He was also one of its biggest funders, using Nvidia’s cash to invest in and acquire companies in the field. In September Nvidia said it would buy Hugging Face, a library of open-source models, for $12.9bn. It has also invested in Reflection and Thinking Machine Labs, two makers of open-source models, and developed a line of its own, called Nemotron.
商业逻辑可能促使英伟达继续支持这种做法。在本周的财报电话会议上,克雷斯表示,公司依然“致力于培育一个充满活力的开放模型生态系统”。但它是否会保持相同的投资步伐则不那么明朗。克雷斯指出,公司正在“严格评估”未来的投资。
Commercial logic may lead Nvidia to continue its support for the approach. On this week’s earnings call, Ms Kress said that the company remained “committed to fostering a vibrant open-model ecosystem”. Whether it will continue investing at the same pace is less clear. Ms Kress noted that the company was “rigorously evaluating” future investments.
尽管如此,硅谷的其他人士会欢迎更稳健的步伐。提供AI驱动背景调查服务的Checkr负责人Daniel Yanisse表示,他和许多同行已准备好接受一定程度的放缓,理由是这将给技术更多时间展示真实价值,而无需承受加速主义者制造的压力——后者承诺乌托邦式的未来即将到来。“真正忙于经营公司、不整天泡在X上的人,并不真信那一套。”
Still, others in Silicon Valley would welcome a steadier pace. Daniel Yanisse, boss of Checkr, which offers AI-powered background checks, says he and many of his peers are ready for some deceleration, arguing that it would give more time for the technology to demonstrate real value without the pressure created by the accelerationists, who promise an imminent utopian future. People who are “actually busy running companies, not spending their day on X, don’t really believe in that”.