内容来源：https://www.theverge.com/ai-artificial-intelligence/926383/mira-murati-sam-altman-musk-trial-ouster

内容总结：

独家报道：OpenAI高层政变内幕曝光，关键证人证词揭示权力游戏真相

2023年感恩节前一周，人工智能行业上演了一场堪比肥皂剧的戏剧性事件：ChatGPT制造商OpenAI的首席执行官萨姆·奥尔特曼（Sam Altman）被董事会突然解职，理由是他“在与董事会的沟通中并非一贯坦诚”。如今，通过“马斯克诉奥尔特曼”一案中的证人证词和庭审证据，公众首次得以窥见那个动荡周末的幕后真相，而焦点人物正是当时的首席技术官米拉·穆拉蒂（Mira Murati）。

穆拉蒂的证词揭开奥尔特曼被罢免内幕

前OpenAI首席技术官穆拉蒂手握大量证据，但这些证据却让她的立场变得扑朔迷离。

这场权力斗争曾每小时都在变化，且很大程度上公开进行。董事会发布了一封措辞模糊的博客文章宣布解职奥尔特曼，立即在社交媒体上引发无数阴谋论。有消息称，导火索是奥尔特曼涉嫌在OpenAI安全流程、自身在创业基金中的持股比例以及ChatGPT等工具发布问题上反复撒谎或隐瞒。

其他OpenAI高管和AI行业领袖纷纷公开支持奥尔特曼。数百名员工发起线上声援活动，许多员工发布“OpenAI离不开员工”的口号。流言四起，无数旁观者屏息等待最新消息。

在这一过程中，一个看似不起眼的人物却无处不在：OpenAI首席技术官米拉·穆拉蒂。起初，她被任命为临时CEO，但很快又将位置让给外部人士埃米特·希尔（Emmett Shear）。几天后，奥尔特曼重回公司领导岗位，而当初联手罢免他的董事会成员几乎全部出局。

穆拉蒂公开支持奥尔特曼复职，并在线发文支持他重返岗位。但随着时间的推移，有报道称她在奥尔特曼被罢免事件中扮演了关键角色。据一些说法，她几乎是最早在内部引发对奥尔特曼担忧的人，并将大量信息——包括截图、短信记录以及奥尔特曼在Y Combinator期间管理不善的指控——传递给了联合创始人伊利娅·苏茨克维（Ilya Sutskever）。苏茨克维随后以一份52页的备忘录将担忧提交给董事会。前董事会成员海伦·托纳（Helen Toner）在证词中表示，穆拉蒂和苏茨克维的担忧“实质性地推进了”董事会自身的判断，涉及奥尔特曼的“欺骗模式”、“抵制”董事会监督、“操纵”董事会流程以及管理问题。

2023年11月16日，OpenAI四名董事会成员一致签署文件，终止奥尔特曼的雇佣关系，并任命穆拉蒂为新的临时CEO。

然而，尽管穆拉蒂在奥尔特曼被罢免前发挥了重要作用，她几乎立刻转向支持奥尔特曼。在14小时内交换的78条短信中，穆拉蒂与奥尔特曼详细讨论了复职可能性。奥尔特曼透露，董事会成员试图达成某种安排，但微软CEO萨提亚·纳德拉（Satya Nadella）和他本人均表示“这行不通”。凌晨2点30分，奥尔特曼询问“方向是好是坏？”，穆拉蒂回答“方向非常糟糕”。她还告知奥尔特曼，董事会“铁了心要他离开”，并透露新CEO人选是“Twitch的陌生人”（指埃米特·希尔）。

在11月17日至20日期间，穆拉蒂与站在奥尔特曼一边的纳德拉多次短信沟通，纳德拉甚至提出将所有OpenAI员工招聘至微软。穆拉蒂曾表示“不会把自己的名字签在”董事会声明上。几天内，超过750名OpenAI员工签署联名信，威胁辞职并加入奥尔特曼领导的新微软子公司。联名信的第一个签名者正是穆拉蒂本人。

值得注意的是，联名信特别指出，董事会“在做出最初决定的两天内，违背公司最佳利益再次更换了临时CEO穆拉蒂”。但根据托纳的证词，穆拉蒂曾告诉董事会，除非董事会能“合法化”这一决定，否则她不愿担任临时CEO。托纳评价道：“她在等待风向，却没有意识到自己就是那个风向。”托纳还称，穆拉蒂在奥尔特曼被罢免后“极不支持”且“异常被动”，“她似乎完全没兴趣告诉团队，自己与我们的对话是解雇奥尔特曼的重要因素。”

在78条短信中，奥尔特曼询问是否该向董事会发送员工联名信，穆拉蒂表示“没用”，因为董事会“不在乎所有人辞职”，只是不想让奥尔特曼“染指AGI”。11月20日清晨，穆拉蒂向微软CTO凯文·斯科特（Kevin Scott）透露“董事会快要辞职了”，斯科特回应“这次是真的？”穆拉蒂答“看来是的，伊利娅签了我们的请愿书。”

大约一年前，穆拉蒂曾撰写一份投诉奥尔特曼管理风格的清单，其中批评道“项目、人员、目标方面的持续恐慌造成了混乱和流失”，“我们谈论专注，但实际执行却是‘什么都做、做快’，因为不断承受优先级调整和人员变动的压力”。她还要求奥尔特曼直接与她沟通：“我不想从别人那里得知……这对解决公司重要问题来说是个错失的机会，也损害了公司领导力。”她指出，“做用户想做的事不是OpenAI的基因”，公司首要目标是创收1亿美元，而奥尔特曼的态度是“不管怎么达到，必须达到”。

在庭审中，穆拉蒂坚持自己对奥尔特曼的批评，称其担忧“完全与管理相关”，“我有一份极其困难的工作，在一个高度复杂的组织里，我请求萨姆清晰领导，不要削弱我完成任务的能力。”

穆拉蒂虽未亲自出庭，但她的证词及披露的文件内容成为庭审中最引人注目的部分之一。

中文翻译：

2023年感恩节前的那一周，堪称人工智能行业最戏剧性的一出大戏。OpenAI首席执行官山姆·奥特曼突然被这家ChatGPT制造商扫地出门。原因是什么？据称是因为奥特曼“在与董事会的沟通中并非始终保持坦诚”。如今，通过马斯克诉奥特曼案中的证人证词和庭审证据，公众首次得以窥见那个戏剧性周末的幕后细节，其中大部分内容都围绕着前首席技术官米拉·穆拉蒂展开。

米拉·穆拉蒂的证词揭开了山姆·奥特曼被解职的内幕。这位前OpenAI首席技术官手握证据，但这些证据反而让她的故事更加扑朔迷离。

这个权力博弈的过山车——似乎每小时都在变化——之所以特殊，很大程度上是因为它发生在众目睽睽之下。董事会宣布解雇奥特曼的博文极为含糊，发布在OpenAI官网后，立刻在X平台上引发了一连串的阴谋论。（结果发现，据称导火索是奥特曼的一连串谎言或隐瞒，涉及OpenAI的安全流程、他在OpenAI创业基金中的持股，以及ChatGPT等某些工具或功能的发布。）其他OpenAI高管和AI行业领袖纷纷发表公开声明支持奥特曼。数百名OpenAI员工发起了一场线上声援活动，他们用发一颗红心来表示支持奥特曼复职，许多人还写下了“没有员工，OpenAI什么都不是”的口号。谣言四起，无数旁观者屏息以待，盼望着哪怕一丁点的新消息。（当时我正在巴塔哥尼亚背包旅行，身边没有笔记本电脑，全靠iPhone上的备忘录应用来报道整件事。）

在这场风波中，一个看似不起眼的人物几乎无处不在：OpenAI首席技术官米拉·穆拉蒂。起初，她被任命为临时CEO，但很快就把这个位置让给了外来者埃米特·谢尔。几天之内，奥特曼重新执掌公司，而当初联合起来罢免他的董事会也几乎被彻底换血。

穆拉蒂曾公开支持奥特曼复职，并在网上发文力挺他重返公司。但随着时间的推移，有报道称她在奥特曼被解职的过程中扮演了重要角色。据一些说法，她几乎可以说是开启了内部关于对奥特曼种种担忧的讨论，并将大量信息——包括截图、短信记录，以及奥特曼在Y Combinator期间管理不善的指控——传递给了联合创始人伊利亚·苏茨克弗，后者随后将这些担忧以一份52页的备忘录形式提交给了OpenAI董事会。在本周的证词中，前董事会成员海伦·托纳表示，穆拉蒂和苏茨克弗的担忧确实推动了董事会本身的想法，这些担忧涉及一系列欺骗行为、奥特曼“抵制”董事会监督，以及他“操纵”董事会流程和管理问题。

2023年11月16日，OpenAI董事会的四名成员——托纳、伊利亚·苏茨克弗、亚当·德安吉洛和塔莎·麦考利——一致签署了解除奥特曼在OpenAI职务的文件，并任命穆拉蒂为新任临时CEO。

尽管按照许多说法，穆拉蒂在奥特曼被解职的整个前期过程中发挥了关键作用，但她几乎立刻就将支持转向了奥特曼。

在从周日傍晚到周一清晨的14个小时内，两人来往了78条短信，详细讨论了奥特曼复职的可能性以及接下来会发生什么。奥特曼说，董事会成员德安吉洛“正试图让董事会同意一个重组方案”，但奥特曼和微软CEO萨提亚·纳德拉已经告诉德安吉洛，那“行不通，并且（他们）需要开始准备B计划了”。

大约在周一凌晨2点30分，奥特曼问道：“你能大概说一下是好是坏吗？萨提亚和其他人很着急。”

穆拉蒂回复：“非常坏。山姆，情况非常糟糕。”

奥特曼要求加入会议，但穆拉蒂说董事会不希望他参与。随后奥特曼发短信说：“你想怎么补救？如果对事情有帮助，我仍然愿意直接走人。但如果他们准备对我发起疯狂的诉讼，那我就不确定了。”

穆拉蒂表示，董事会坚信他们让奥特曼离开公司的决定是正确的，并补充道：“他们向我详细说明了所有理由，以及与你相关的问题，为什么你不能继续担任CEO。”

奥特曼问，那为什么董事会“整个周末都在说希望我回去”。

穆拉蒂回答：“他们希望今晚就有一位新CEO上任（不是我）。”

奥特曼问是谁，穆拉蒂回答：“新来的家伙是Twitch的那个怪人”，指的是谢尔。

穆拉蒂告诉奥特曼，她“希望萨提亚能帮忙扭转局面”。

在11月17日至20日期间，穆拉蒂和纳德拉——后者在这场冲突中坚定地站在奥特曼一边，并曾提出聘请所有OpenAI员工到微软，在奥特曼手下工作——也互通了大量短信（基本上是一边倒的，主要是穆拉蒂联系纳德拉）。在其中一条短信中，穆拉蒂提到她“不会把自己的名字放在这上面”，这似乎指的是董事会在那个周日发布的一份声明，声明称“董事会坚定地认为，其决定是推进和维护OpenAI使命的唯一途径”，并且“简而言之，山姆的行为及其与董事会互动中缺乏透明度，削弱了董事会以其被授权的方式有效监督公司的能力”。

几天之内，超过750名OpenAI员工签署了一封致OpenAI董事会的信，威胁要集体辞职，加入由奥特曼和OpenAI总裁格雷格·布罗克曼领导的新微软子公司。

他们写道：“你们解雇山姆·奥特曼并将格雷格·布罗克曼从董事会除名的过程，危及了我们的所有工作，损害了我们的使命和公司。”

这封信的第一个签名者是谁？正是穆拉蒂本人。

这封信最有趣的部分之一是在接近结尾处，签名者特别向董事会指出：“在你们做出最初决定的两天内，你们再次更换了临时CEO米拉·穆拉蒂，这违背了公司的最佳利益。”

但请记住：根据托纳的证词，穆拉蒂显然曾告诉董事会，除非董事会能够使这个决定“合法化”，否则她不想担任临时CEO。托纳说，穆拉蒂“似乎并不明白，无论是有意还是无意，她自己在使这个决定合法化方面可以发挥关键作用”。

“她一直在观望风向，却没意识到她自己就是那股风，”托纳说。

托纳还表示，在奥特曼被解职后，穆拉蒂表现得“出奇地不支持”且“异常被动”，并补充道：“她似乎完全没兴趣告诉她的团队，她与我们之间的对话是我们决定解雇山姆的一个重要因素。”

在穆拉蒂和奥特曼的78条短信往来中，奥特曼问是否到了把员工联名信发给董事会的时候了；穆拉蒂告诉他那“没用的”，董事会成员“根本不在乎所有人是否辞职”，他们只是不希望奥特曼“染指通用人工智能”。奥特曼问德安吉洛是否知道穆拉蒂重新雇了他，穆拉蒂说知道。

11月20日周一清晨，穆拉蒂给微软CTO凯文·斯科特发短信说，他们“快要让董事会辞职了”。斯科特回复：“这次是真的吗？”穆拉蒂说：“看起来是的。伊利亚（苏茨克弗）在我们的请愿书上签了字。”

当天上午晚些时候，穆拉蒂请求纳德拉“请尽快发表一份公开声明，表明对（OpenAI）联合团队的支持，基本上就是把团队凝聚起来？我们绝不能让研究员流失给戴米斯（哈萨比斯）或埃隆（马斯克），这一点非常重要。”

大约一年前，在一份日期为2022年9月30日的文件中，穆拉蒂列举了一系列对奥特曼及其管理风格的抱怨，这份文件似乎是直接与他分享的。她写道：“围绕我们的项目、人员、目标等不断的恐慌，制造了混乱和人员流失”，并且“我们口口声声说要专注，但实际执行起来却是什么都做、还要做得快，因为我们不断承受着改变优先事项、调整人员和项目的压力。”她还写到了奥特曼与高管团队在应用AI团队重要性上的分歧，并要求奥特曼直接与她沟通他的担忧：“我不想从别人那里得知……这对我们解决公司的关键问题来说是个错失的机会，而且你这样做会削弱公司的领导力。”

在2022年的那份文件中，穆拉蒂还提到了一个观点，即“满足用户的需求并非OpenAI的基因”——公司首要目标是创造1亿美元的收入，而奥特曼的立场是“我们如何达到这个数字并不重要，我们必须达到。”穆拉蒂还说，为奥特曼解决这些问题提出的首要解决方案之一，就是“去了解情况”，并通过官方渠道提出变革建议。

“我经常从你那里听到两件似乎相互矛盾的事情：（1）我们行动不够快，或者某个特定领域或人失败了；（2）你不知道发生了什么，所以你可能是错的，”她在2022年的文件中写道。“当不确定事情进展如何，或者感觉事情进展不顺利时，请你直接来找我米拉获取信息，并安排深入审查，直到你确信自己了解了情况。”

在本周马斯克诉奥特曼案的庭审中，作为穆拉蒂证词的一部分，她表示自己坚持对奥特曼的批评，并称她的担忧“完全与管理有关……我在一个非常复杂的组织中承担着一项极其困难的工作。我要求山姆去领导，清晰地去领导，而不是破坏我履行自己职责的能力。”

穆拉蒂本人或许没有出现在法庭上，但她的证词——以及文件中披露的内容——却是最令人印象深刻的环节之一。

英文来源：

The week leading up to Thanksgiving 2023 was the AI industry’s biggest soap opera moment. OpenAI CEO Sam Altman was abruptly ousted from his role at the ChatGPT maker. The explanation? That Altman was “not consistently candid in his communications with the board.” Now, via witness testimony and trial exhibits in Musk v. Altman, the public is getting a concrete look behind the scenes of that dramatic weekend for the first time, much of it centered on former CTO Mira Murati.
Mira Murati’s deposition pulled back the curtain on Sam Altman’s ouster
The former OpenAI CTO had receipts. But they mostly confuse her own story.
Mira Murati’s deposition pulled back the curtain on Sam Altman’s ouster
The former OpenAI CTO had receipts. But they mostly confuse her own story.
It was a unique situation in that the roller coaster of a power play — which seemed to change every hour — took place, in many ways, publicly. The board’s strikingly vague blog post announcing Altman’s ouster was posted on OpenAI’s website, immediately sparking a laundry list of conspiracy theories bandied about on X. (It turned out that the impetus had allegedly been a pattern of lying or omission by Altman, whether about OpenAI’s safety processes, about his own ownership stake in OpenAI’s startup fund, or about the release of certain tools or features like ChatGPT.) Other OpenAI executives and AI industry leaders made public statements in support of Altman. An online campaign began among hundreds of OpenAI employees in which they posted a heart if they supported Altman’s reinstatement, and many posted the phrase, “OpenAI is nothing without its people.” Rumors swirled as countless onlookers waited with bated breath for any new kernel of information. (I covered the whole thing from a backpacking trip in Patagonia, armed with only an iPhone notes app and no laptop.)
Throughout it all, one unassuming character seemed to be everywhere at once: OpenAI CTO Mira Murati. At first, she was made interim CEO, before immediately ceding the position to outsider Emmett Shear. Within days, Altman was back at the helm of the company, and the board that had come together to remove him was largely gone.
Murati had publicly supported Altman’s reinstatement and posted online in favor of him returning to his role at the company. But over time, reports surfaced that she had had a significant hand in his ouster. She had, by some accounts, more or less started the internal conversation about concerns surrounding Altman and funneled a significant amount of information — including screenshots, documentation of text messages, and allegations of mismanagement during Altman’s time at Y Combinator — to cofounder Ilya Sutskever, who then took his concerns to the OpenAI board in the form of a 52-page memo. In testimony this week, former board member Helen Toner said that Murati and Sutskever’s concerns had materially advanced the board’s own, relating to a pattern of deceit, Altman’s “resistance” of board oversight, and his “manipulation” of board processes and management problems.
On November 16th, 2023, four members of OpenAI’s board of directors — Toner, Ilya Sutskever, Adam D’Angelo, and Tasha McCauley — unanimously signed a document terminating Altman’s employment with OpenAI and naming Murati the new interim CEO.
Though Murati had, by many accounts, played an integral part in the entire lead-up to Altman’s ouster, Murati almost immediately seemed to switch her support to Altman.
In 78 text messages exchanged over a 14-hour period, between early Sunday evening and Monday morning, Murati and Altman talked at length about whether his reinstatement would be possible and what would happen next. Altman said that D’Angelo, a board member, was “trying to get the board to agree to a configuration” but that Altman and Microsoft CEO Satya Nadella had told D’Angelo that that “doesn’t work and that [they] need to start preparing for plan b.”
Around 2:30AM on Monday morning, Altman asked, “can you indicate directionally good or bad? satya and others anxious.”
Murati responded, “Directionally very bad. Sam this is very bad.”
Altman asked to join the meeting and Murati said the board didn’t want him to. Altman then texted, “what do you want to make it better? i’m still willing to just walk away if that helps. if they are ramped up for crazy lawsuits against me then i’m not sure what.”
Murati said the board was convinced of their decision for Altman to leave the company, adding, “They’ve walked me through all the reasons and the issues with you and why you can’t be ceo.”
Altman asked why the board, then, had been “saying all weekend they wanted me back.”
Murati responded, “They want to have a new ceo in place tonight (not me.”
Altman asked who, and Murati responded, “New guy is rando twitch guy,” referencing Shear.
Murati told Altman she was “hoping Satya can help undo this.”
Between November 17th and 20th, Murati and Nadella, who was squarely on Altman’s side during the conflict and had offered to hire every OpenAI employee over to Microsoft to work under Altman, also exchanged a number of text messages (largely one-sided, with Murati reaching out to Nadella). In one, Murati mentions that she’s “not putting [her] name on this,” seeming to reference a statement by the board issued that Sunday that “the board firmly stands by its decision as the only path to advance and defend the mission of OpenAI,” and that “put simply, Sam’s behavior and lack of transparency in his interactions with the board undermined the board’s ability to effectively supervise the company in the manner it was mandated to do.”
Within days, more than 750 OpenAI employees signed a letter to OpenAI’s board, threatening to quit and join the new Microsoft subsidiary that would be led by Altman and OpenAI president Greg Brockman.
They wrote that “the process through which you terminated Sam Altman and removed Greg Brockman from the board has jeopardized all of this work and undermined our mission and company.”
The very first signatory on that letter? Murati herself.
One of the most interesting parts of the letter is near the end, when the signatories specifically note to the board that “within two days of your initial decision, you again replaced interim CEO Mira Murati against the best interests of the company.”
But remember: Murati, apparently, had told the board that she didn’t want to serve as interim CEO unless the board was able to “legitimize” the decision, according to Toner’s testimony. Toner said Murati “did not seem to understand, either willfully or not, that she had a pivotal role to play in legitimizing this decision herself.”
“She was waiting to see which way the wind would blow, and she didn’t realize that she was the wind,” Toner said.
Toner also said that Murati had been “strikingly unsupportive” and “remarkably passive” after Altman’s removal, adding, “She seemed totally uninterested in telling her team that her conversations with us had been a significant factor in our decision to fire Sam.”
During the 78 text messages between Murati and Altman, Altman asked if it was time to send the board the letter from the employees; Murati told him it “wouldn’t matter” and that the board members “don’t care if everyone quits,” just that they didn’t want Altman’s “hand on agi.” Altman asked if D’Angelo knew that Murati had rehired Altman, and she said yes.
Early on the morning of Monday, November 20th, Murati texted Microsoft CTO Kevin Scott that they were “close to having the board resign.” Scott responded, “For real this time?” Murati said, “It seems so. Ilya [Sutskever] signed our petition.”
Later that morning, Murati asked Nadella to “please make a public statement soon that shows support for the joint [OpenAI] team, basically bringing the team together? It’s very important that we don’t lose researchers to Demis [Hassabis] or Elon [Musk].”
A little over a year before, in a document dated September 30th, 2022, Murati had written a list of complaints about Altman and his management style that was apparently shared directly with him. She wrote that “constant panic around our projects, people, goals, etc generates chaos and churn,” and that “we talk about focus but in practice our approach is do-everything and do it fast because we constantly get pressure to change priorities and shuffle around people and projects.” She also wrote about Altman and the executive team’s misalignment about the importance of the applied AI team, and requested that Altman talk about his concerns with her directly: “I don’t want to find out from others … It’s a missed opportunity for us to resolve important issues for the company and it undermines the leadership of the company when you do this.”
Murati also mentioned, in that 2022 document, the idea that “doing what the users want is not in the DNA of OpenAI” — that the company’s top-cited goal was to generate $100 million in revenue, and that Altman’s position was that “it didn’t matter how we got to this number, we needed to get there.” Murati also said that one of the top proposed solutions for Altman to remedy these issues would be to “get informed” and use official channels to bring up proposed changes.
“Often I hear from you two things simultaneously, that to me seem in conflict: (1) We’re not moving fast enough or a particular area or person is failing & (2) You don’t know what’s going on, so you might be wrong,” she wrote in the 2022 document. “When unsure of how things are going or if there’s a feeling that things are not going well, go directly to Mira to get information and set up in-depth reviews until you are satisfied that you understand the situation.”
As part of Murati’s deposition played at the trial in Musk v. Altman this week, she said she stood by her criticisms of Altman and that her concerns were “completely management related … I had an incredibly hard job to do in an organization that was very complex. I was asking Sam to lead, and lead with clarity, and not undermine my ability to do my job.”
Murati may not have been present in the courtroom, but her testimony — and what was revealed in documentation — was among the most memorable.