AI Trends

Jeff Bezos Warns Nvidia Against Bringing In an Outside CEO

AYi AI NotesUpdated Nov 19 02:39

English text is being prepared.

AI summary

At a Semafor event, Jeff Bezos said that unless Nvidia brings Jensen Huang back as CEO, it should not parachute in an outside professional manager, because external CEOs have an extremely low chance of succeeding at complex companies. Contrasting this with Amazon’s years-long planning for Andy Jassy’s succession, he argued that an outsider would struggle to quickly grasp Nvidia’s complex system spanning chips, supply chains, export controls, cloud providers, and the CUDA ecosystem.

Why read this

AI score 74

Drawing on Amazon’s experience with internal succession, Bezos analyzes the succession challenges facing complex technology companies, revealing the tension between accumulated founder knowledge and organizational capabilities.

Topics

Tags

Quoted post

AYi

@AYi_AInotes

Tao Lin, a frontline safety researcher at Anthropic whose job is to keep large models from going off the rails, recently laid all his cards on the table in an interview with Palisade Research. He said that AI comprehensively surpassing human experts could arrive in as little as 1.3 years. After that, nearly everyone on Earth would face permanent unemployment and ultimately be left relying on government benefits to survive. Many people’s first reaction upon seeing a claim like this is that it is just more fearmongering. But first, look carefully at who is saying it. He is not some attention-seeking content creator on the sidelines; he is an internal researcher at the very heart of Claude’s development, specializing in automated alignment and AI control. His daily work is to watch Anthropic’s own models to ensure they do not steal model weights or deceive internal employees. The timeline he presented in the interview is far more urgent than most people imagine. He predicts that in roughly 1.3 years—around the end of 2027—broadly superhuman AI will emerge, matching or surpassing the world’s leading human experts across nearly every field, including programming, film and television production, and even political strategy. Interestingly, he believes the greatest danger is not the kind of sudden machine rebellion seen in science fiction. Although he puts the probability of AI escaping control and taking over at below 50%, he sees the systematic elimination of employment pathways as virtually certain. First, your current job disappears, so you grit your teeth and switch careers. But not long after you make the transition, the new industry is swallowed up too. Eventually, almost no one will be able to find stable work again. When the Industrial Revolution and the internet arrived, people always reassured themselves that although old jobs would disappear, new ones would emerge. Put plainly, that meant switching tracks: a carriage driver could become an automobile driver and still earn an income by selling their time and mental effort. This time is completely different. The machines are taking over the tracks themselves, and even driving will no longer require human involvement. Still, do not rush to give up after hearing this interview. He actually identifies an extremely important buffer created by the physical world. He explicitly says society will not be wiped clean within 15 months because, even if AI brains have already reached superhuman levels, robots and automated factories in the physical world will still have to be built one by one. Jobs in construction, caregiving, and fields that rely heavily on in-person trust and complex physical tasks cannot immediately be replaced by code in the short term. But viewed another way, the lag created by the pace of construction in the physical world is precisely the last and most valuable window ordinary people have to adjust. He himself has publicly called on major AI companies to slow their research and development to one-quarter of the current pace, giving institutions and society some breathing room. That shows even the people building this technology on the front lines believe the train is moving too fast. Honestly, what will truly widen the gap in people’s destinies between 2026 and 2027 is no longer simply whether you can use AI to write a few lines of code or polish a weekly report. Before the marginal value of human labor is systematically driven toward zero, the real question is whether you can use these few years of physical-world buffer to pull yourself out of the execution layer, where you merely perform tasks passively, and force your way into the rules layer—designing processes, supervising agents, or even gaining partial ownership of the system. https://t.co/frPZH9ekSV

View original

在 Anthropic 负责盯着大模型别乱来的一线安全研究员 Tao Lin,最近在 Palisade Research 访谈里直接把底牌摊开了,老哥说最快一点三年后出现全面超越人类专家的 AI,之后地球上几乎所有人都会面临永久失业,最后只能靠政府福利活着, 很多人刷到这种话第一反应是又在贩卖焦虑,但你先看清说话人的位置,他可不是外围博眼球的自媒体,而是坐在 Claude 研发最核心地带、专门负责自动化对齐与 AI 控制的内部研究员,天天干的活就是盯着自家模型别偷权重、别欺骗内部员工, 他在这段访谈里抛出的时间表,比绝大多数人想象的要紧迫得多,他判断大概在一点三年也就是 2027 年底左右,就会出现广义超人级 AI,在编程、影视制作甚至政治博弈等几乎所有领域,全面达到或超过人类顶尖专家的水平, 有意思的是,他觉得最危险的倒不是科幻片里那种机器突然造反,虽然他认为 AI 失控接管的概率不到一半,但就业路径被系统性切断几乎是板上钉钉的,先是你手头的工作没了,你赶紧咬牙转行,结果刚转过去没多久新行业又被吞了,最后几乎没人能再找到一份稳定工作, 以前工业革命和互联网来了,大家总安慰自己旧工作没了还会有新工作,说白了那叫换赛道,你从马车夫转行当汽车司机,依然能靠出卖自己的时间和脑力去换取收入,但这次完全不一样,赛道本身直接被机器接管了,连当司机这件事都不再需要人类插手, 不过听完这段访谈先别急着躺平,他在里面其实藏了一个极其关键的物理缓冲期,他明确表示社会不会在十五个月内瞬间清零,因为哪怕大脑已经达到超人级别,物理世界里的机器人和自动化工厂还得一座座建起来,建筑、护理、重度依赖现场信任和复杂物理操作的工作,短时间内还没法立刻被代码替换, 但转念一想,这段因为物理世界建造速度拖出来的滞后期,恰恰是留给我们普通人最后也最宝贵的调整窗口,连他自己都公开呼吁各大 AI 公司把研发速度降到现在的四分之一,好给制度和社会留出喘息时间,说明一线造物主自己都觉得这趟列车开得太快了, 说实话,2026 到 2027 年真正拉开命运差距的,早就不只是你会不会用 AI 写几行代码或者润色周报,而在人类劳动的边际价值被系统性压向零点之前,你能不能趁着这几年物理缓冲期,把自己从只会被动干活的执行层,硬生生拽进设计流程、监督智能体、甚至掌握一部分系统所有权的规则层。 https://t.co/frPZH9ekSV

View original post

Body

亚马逊创始人 Jeff Bezos 刚刚在 Semafor 活动上直接把话挑明了, 如果英伟达董事会不把黄仁勋请回来当 CEO,就千万别去外面找职业经理人空降, 因为英伟达太复杂了,复杂公司里外来 CEO 成功的概率小到几乎看不见, 很多人在吃瓜英伟达权力更迭时,总习惯用传统大公司的眼光去看问题,觉得创始人既然因为风波离场,董事会去市场上挖一个履历光鲜、擅长合规治理的明星高管空降进来,照样能把这台万亿印钞机开得飞起, 但贝索斯用自己执掌亚马逊几十年的亲身教训,直接把这个天真的幻想砸了个稀碎,他说自己在把帅印交给 Andy Jassy 之前,足足花了多年时间做内部规划与平稳过渡,而英伟达眼下这种突发风暴下的权力真空,跟当年亚马逊的从容交接完全不在一个维度, 说实话,听完这段 44 秒的访谈,我才真正想明白为什么顶级科技巨头最怕空降兵, 因为 2026 年的英伟达早就不只是一家卖显卡的硬件公司,它是一座同时咬合着芯片底层架构、台积电晶圆产能、地缘出口管制、云厂商博弈以及二十年 CUDA 生态的活体操作系统,这座城堡里藏着无数外人根本摸不着的暗道和机关,大量关键的技术直觉和商业默契,全都长在黄仁勋和那批跟着他打了二十年仗的核心老兵脑子里, 一个外部空降的 CEO 走进来,光是要看懂这张错综复杂的系统地图就得花上大半年补课,但残酷的算力战场上,微软、谷歌和博通的自研芯片根本不会停下来等你交接完毕,在贝索斯眼里,去外面找人只能是走投无路时的最后手段, 有意思的是,现实里的博弈比贝索斯的警告还要拧巴,黄仁勋之前已经罕见低头承认个人生活犯了严重错误,公开表态只要符合股东、员工和国家利益他随时愿意回归,但董事会主席 Harvey Jones 在财报电话会上冷冰冰地堵死了大门,明确表示不打算让老黄复职,同时已经把目光投向了包括 Arm 掌门人 Rene Haas 在内的外部人选, 我一直觉得,眼下的英伟达董事会正卡在一个极其撕裂的死胡同里,一边是华尔街和华盛顿死死盯着的治理合规与道德红线,另一边是创始人一旦彻底抽离后、整座万亿技术帝国可能面临的系统失速风险, 无论最终谁坐上那把交椅,贝索斯这番话都给所有科技公司敲响了一记警钟,真正可怕的从来不是创始人突然离开,而是在没有把个人脑子里的顶级直觉沉淀成组织能力之前,就匆忙把方向盘交给了一个连地图都还没看懂的陌生人。 https://t.co/xk72CEQVUI

Source

AYi AI Notes