凯文·鲁斯的新书《AGI编年史》为理解现代人工智能的三位最重要建筑师提供了罗塞塔石碑：达里奥·阿莫代伊、萨姆·奥尔特曼和德米斯·哈萨比斯。
Kevin Roose's new book, "The AGI Chronicles," provides a Rosetta Stone to understanding the three most important architects of modern AI: Dario Amodei, Sam Altman and Demis Hassabis.
这很重要：了解每个人的动机有助于解释他们的商业策略、公开姿态和冲突。
Why it matters: Knowing what motivates each man helps explain their business strategies, public posturing and conflicts.
以下是我根据鲁斯的书、采访和播客节目所做的提炼 — — 尤其是德里克·汤普森的《Plain English》，该节目深入探讨了这三个人及其动机。
What follows is my distillation of Roose's take based on his book, interviews and podcast appearances — especially Derek Thompson's Plain English, which explores the three men and their motivations.
你可以完全追溯达里奥对AI灾难的警告（和强迫性担忧）到……原子弹。他过去会把他最喜欢的书《原子弹的制造》发给每一位新入职的Anthropic员工。他的偶像是利奥·西拉德。西拉德构想了核链式反应，然后敦促 fellow 物理学家保密。他起草了爱因斯坦致罗斯福的信，这封信启动了曼哈顿计划，随后他又努力限制原子弹的使用。
You can fully trace Dario's warnings (and obsessive worry) about AI catastrophe to ... the atomic bomb. He used to hand every new Anthropic employee his favorite book, "The Making of the Atomic Bomb." His hero is Leo Szilard. Szilard conceived the nuclear chain reaction and then urged fellow physicists to keep it secret. He drafted the Einstein letter to FDR that launched the Manhattan Project, then fought to restrain the bomb's use.
达里奥曾将自己比作西拉德，认为一旦核武器的蓝图被人所知，美国除了建造或控制它之别无选择，否则就会输给纳粹德国。
Dario has likened himself to Szilard, who decided once the blueprint of nuclear weapons was known, America had no choice but to build or control it — or lose to Nazi Germany.
这样理解他：一个相信自己正在制造原子弹，并且必须在鲁莽之人之前完成这件事的人。目前尚不清楚“鲁莽之人”是指中国还是萨姆，而达里奥深深不信任后者。
Read him as: A man who believes he's building the bomb and must get there before the reckless do. It's unclear if "the reckless" is China or Sam, whom he deeply distrusts.
卖家：萨姆·奥尔特曼（OpenAI）在鲁斯的叙述中，萨姆拥有两种独特的天赋：他能够向任何人推销任何东西，以及他对无限规模的不可动摇的信念。
The seller: Sam Altman (OpenAI) Sam has two distinct gifts, in Roose's telling: his ability to sell anyone on anything and his unshakable belief in limitless scaling.
他能够弄清楚某人想要什么，然后给予对方，或表面上如此。无数前同事的评价是：他对不同的人说不同的话，让人琢磨不透 — — 他内心到底有什么，如果有的话。
He can figure out what someone wants and give it to them, or appear to. The rap from countless former colleagues: he tells different people different things, leaving people wondering what — if anything — sits at his core.
正是这种能力让他能够筹集到惊人的资金，将一个非营利组织转变为即将IPO的庞然大物，并且规模扩张速度几乎无人能及。但这也让达里奥、埃隆·马斯克以及许多前同事对他深感厌恶。
This has allowed him to raise insane amounts of money, turn a nonprofit into an IPO-bound behemoth and scale faster than almost anyone. It's also left Dario, Elon Musk and many former colleagues loathing him.
这样理解他：规模最大、速度最快、几乎不惜一切代价。这有利于速度。但目的却是个谜。
Read him as: The biggest, fastest, at almost any cost. Great for speed. But the purpose is the mystery.
科学家：德米斯·哈萨比斯（Google DeepMind）德米斯很早就把科学搞对了。他的AlphaFold攻克了蛋白质折叠难题，让他分享了2024年诺贝尔化学奖。（纪录片《思维游戏》必看。）但他对大语言模型判断失误。他字面意义上认为，现代AI将在由科学家和全球领导人治理的实验室中开发，只有在重大问题解决后才会向公众展示。
The scientist: Demis Hassabis (Google DeepMind) Demis got the science right — early. His AlphaFold cracked protein folding and won him a share of the 2024 Nobel Prize in Chemistry. ("The Thinking Game" documentary is a must-watch on this.) But he got LLMs wrong. He literally thought that modern AI would be developed in a lab governed by scientists and global leaders and shown to the public only when big problems were solved.
他深度怀疑萨姆和达里奥骑着缩放定律跑出的领先优势——并取得了巨大的早期领先。将DeepMind出售给谷歌让他感到窒息，所以德米斯正拼命试图夺回他的魔力与荣光。
He was deeply skeptical of the scaling laws Sam and Dario rode to power — and to substantial early leads. Selling DeepMind to Google suffocated him, so Demis is scrambling to recapture his magic and glory.
解读他：这三人中大概最痛苦的一个。他是被迫在一场不想参加的比赛中冲刺、在一条他怀疑的道路上、对抗两个拥有巨大先发优势的对手的科学家和纯粹主义者。
Read him as: Probably the most tortured of the three. He's a scientist and purist forced to sprint in a race he didn't want, on a path he doubted, against two men with a big head start.
如果您是首席执行官或首席执行官团队成员：请申请加入吉姆全新推出的每周Axios C-Suite高管通讯。
If you're a CEO or on a CEO's team: Ask to join Jim's new weekly Axios C-Suite newsletter.