【动向】OpenAI董事会解雇CEO，非营利灵魂与野心家的矛盾

投资100亿美元的微软没有OpenAI董事会席位，微软管理层在消息公布前几分钟才得知。

11月17日，美国大模型领域最炙手可热的创业公司OpenAI发布声明称，山姆·奥特曼（Sam Atlman）将辞去首席执行官职务并离开董事会。该公司首席技术官米拉·穆拉蒂（Mira Murati）将担任临时首席执行官，立即生效。

声明称，Sam Altman在与董事会的沟通中始终不坦诚，阻碍了董事会履行职责的能力。董事会不再对他继续领导OpenAI充满信心。同时，另一位创始人兼总裁格雷格·布罗克曼（Greg Brockman）将辞去董事会主席职务，并继续担任公司职务，向首席执行官汇报。CXO UNION CXO联盟 cxounion.cn

11月18日，山姆在其社交账号上表示，很喜欢在OpenAI工作的时间，稍后会公布他接下来的安排。随后，格雷格转发这条内容并表示，得知这一消息后，他给OpenAI发了一条信息，称他决定退出OpenAI。

也就是说，成立8年后，OpenAI成为最受全球瞩目的科技公司还不到1年，两位核心创始人同时离开。山姆都对这次事件表示震惊，他们都还在努力弄清楚原因到底是什么。CXO UNION CXO联盟 cxounion.cn

他提到，11月16日，山姆收到首席科学家Ilya Sutskever发来的短信，要求周五中午谈话。山姆参加了一次Google线上会议，除了格雷格外，其他董事都在场。Ilya告诉山姆他将被解雇。

11月17日中午12:23，格雷格被告知，他将被从董事会中除名（但他对公司至关重要，将保留他的职位），而山姆已被解雇。随后，OpenAI对外公布了这一消息。截至目前，OpenAI并未对这一人事变动有更多新的解释，包括外界最关心的山姆具体有哪些“不坦诚的行为”。山姆一直是OpenAI的代言人，上周的开发者大会上，他主持并发布了OpenAI的一系列新产品和服务。CXO UNION CXO联盟 cxounion.cn

11月16日，也就是被解雇的前一天，他还代表OpenAI在亚太经济合作会议上发表了讲话。

据多家美国媒体报道，11月17日，OpenAI临时CEO米拉在公司全体会议上表示，目前公司与微软的关系保持稳定，微软CEO和CTO都对OpenAI的未来表达了“极大的信心”。

一、谁做的决定？

据美国媒体报道称，微软的领导层是在此消息公布前的几分钟才得知。此前，OpenAI的董事会由六名成员组成，格雷格和山姆退出后，现由OpenAI首席科学家Ilya Sutskever、Quora首席执行官Adam D’Angelo、技术企业家Tasha McCauley以及乔治城安全与新兴技术中心的Helen Toner组成。CXO UNION CXO联盟 cxounion.cn

向OpenAI公司投资了100亿美元的微软并没有该公司董事会席位。声明称，董事会大部分成员都是独立董事，不持有OpenAI股权，包括前任CEO山姆。山姆的收益来源于Y Combinator早期对OpenAI的小额投资。董事会的基本职责是推进OpenAI的使命并维护其章程。

OpenAI成立于2015年，是一家非营利性组织，核心使命是确保通用人工智能（AGI）能造福全人类。OpenAI的受益者是全人类，而非投资人。考虑到AGI的发展需要大量资金支持，2019年OpenAI宣布调整为“利润上限”模式，董事会是所有OpenAI活动的总体管理机构。具体的模式为，成立一个新的营利性子公司，用于募集资金，但有义务履行非营利性组织的使命，也就是安全和造福全人类。CXO UNION CXO联盟 cxounion.cn

营利性子公司由OpenAI完全控制，允许赚钱但会设置利润上限，包括微软的投资收益也会受到限制。做出“把核心创始人山姆·奥特曼赶出公司”决定的OpenAI董事会构成有一些特殊。目前的四名董事中，只有Ilya是OpenAI员工，他是一名AI科学家，加入OpenAI之前在谷歌担任科学家，过去他出现在公众视野中时，基本都在讨论OpenAI的技术细节。目前他主要负责人工智能的监督工作。

Adam D’Angelo于2018年4月加入 OpenAI董事会。当时他有一个观点：安全的AGI工作既重要又没有得到充分重视。他的观点是OpenAI不应该成为一家科技大厂，因为这不会让世界变得更好。Tasha McCauley今年年初加入OpenAI董事会，她曾创办过多家初创公司。目前她与另一位董事Helen Toner一起担任英国成立的国际人工智能治理中心的顾问委员会成员。Helen Toner于2021年9月加入OpenAI董事会，她的职责是AI的安全问题。格雷格曾表示“非常重视Helen对人工智能长期风险和影响的深入思考。”通常公司董事会成员会分别负责不同领域的监督职责，但OpenAI完全不同，四名董事的主要的职责都是AI安全。作为OpenAI最高的管理机构，董事会代表了这家公司的灵魂和初衷：非营利和安全，但在山姆领导下的OpenAI，逐渐偏离了这个方向。CXO UNION CXO联盟 cxounion.cn

二、非营利、安全与野心

与山姆一同辞职的格雷格是OpenAI的创始人之一。据格雷格的文章，2015年山姆询问他未来的计划，他表示想做AI，但还没清楚。山姆说他考虑通过YC孵化器建立一个AI实验室，目标是“建立安全的人类级AI”，或许可以合作他们很快达成一致，认为必须要通过非营利性组织，才能实现这一目标，因为不会有任何利益冲突来影响这一使命。

但山姆在硅谷的“人设”是野心家，YC创始人保罗·格雷厄姆（Paul Graham）曾评价他“野心超过了硅谷能容纳的边界”、“如果把他扔到某个食人族之岛，5年后他将成为这个岛的国王”。2014年，不到30岁的山姆担任YC创业孵化器CEO，5年时间带领YC市值接近1500亿元。CXO UNION CXO联盟 cxounion.cn

2015年，他与马斯克、格雷格等人共同成立OpenAI，2019年之前，他的身份是OpenAI的投资人。

2019年，他离任YC总裁，成为OpenAI的CEO。今年，OpenAI开始在商业化之路上一路狂奔，包括推出付费版产品、定制化企业版等。

今年8月，有消息称OpenAI2023年营收将达到10亿美元，并在2024年继续增长几十亿美元。今年10月，山姆告诉OpenAI员工，今年公司营收将达到13亿美元。而2022年，该公司营收仅2800万美元。在他的带领下，OpenAI的最新估值已经达到900亿美元。CXO UNION CXO联盟 cxounion.cn

2022年9月，山姆和Reid Hoffman（Linkedin联合创始人）的一次对谈上，他提到目前AI领域真正的难题是财富分配、AGI使用权和AGI治理问题。随着AI技术的发展，AGI会变成一种商品，需要考虑让所有人平等获得使用AGI的机会。

除了商业化，山姆与董事会另一层可能的矛盾是“安全”。山姆领导下的OpenAI营收快速增长，具体措施包括C端收费，定制化B端服务，开放GPT接口和应用生态等，但这些都让OpenAI面临安全指责。这是AI安全和商业化之间天然存在的矛盾。CXO UNION CXO联盟 cxounion.cn

11月8日，ChatGPT和API接口出现“严重停机”事件，整个故障时间超过12小时。主要原因是11月7日开发者日公布ChatGPT更新后，有大量用户涌入，以及由于DDoS（分布式拒绝服务，通常由网络超载引起）和身份验证失败，导致ChatGPT全线宕机。     CXO UNION CXO联盟 cxounion.cn

随后山姆在社交平台上道歉。他说，新功能的使用情况远远超出预期，公司原计划周一为所有订阅者启用GPT，但仍未能实现。由于负载的原因，短期内可能会出现服务不稳定的情况。围绕OpenAI安全的担忧主要来自数据问题，不少人担心和ChatGPT的对话过程中会泄露数据。

今年年初，意大利数据保护局（GPDP）指控OpenAI违反了欧盟的数据保护法，目前该案件还在审理中。最近的一次关于OpenAI安全性的争议是11月9日，微软在内部网站的更新中提到了“安全和数据问题”，因此切断了员工使用ChatGPT等AI工具。但一天后11月10日，微软就恢复了使用。微软发言人称，微软正在测试一个端点控制系统，中间出现了错误，所以很快恢复了。CXO UNION CXO联盟 cxounion.cn

今年5月，苹果限制员工使用ChatGPT及其他外部AI工具，因为苹果在研发相关技术，担心机密数据外泄。同月，三星电子发布内部备忘录，禁止员工使用包括ChatGPT、BingChat、谷歌Bart在内的生成式AI工具，因为三星工程师曾在4月将内部原始代码上传至ChatGPT。

今年7月，美国联邦贸易委员会（FTC）对OpenAI发起调查，调查是否通过发布虚假信息损害了人们的利益，以及OpenAI是否存在“不公平或欺骗性”的隐私和数据安全行为。该机构担心ChatGPT和其他AI服务被输入大量数据，而AI公司没有对数据类型进行检查。随后，山姆表示会和FTC合作，并相信OpenAI的技术是安全的。大模型依靠源源不断的新数据来优化，而商业化又要求大模型在专业领域不断进步。只要OpenAI还在追求营收增长，就必然会面临数据安全的挑战。曾是OpenAI早期投资人及董事会成员的马斯克就多次表示他对OpenAI的发展不认可，包括微软对于OpenAI的强控制权（这一说法被微软否认），OpenAI对于AI安全方面的把控不够等。

今年2月，他在X账号上写道，“OpenAI最初是作为一家开源的非营利性公司而创建的，为了抗衡谷歌，这也是我把它命名为OpenAI的原因。但现在它已经成为一家闭源的营利性公司，由微软有效控制，这完全违背了我的本意。”    

马斯克在2018年就退出了OpenAI董事会。另一位前董事会成员Reid Hoffman于今年3月退出，因为他计划投资一家使用OpenAI产品的软件公司，担心会有利益冲突。CXO UNION CXO联盟 cxounion.cn

翻译：

OpenAI board fires CEO, non-profit soul and careerist conflict

Microsoft, which has invested $10 billion, does not have a seat on OpenAI’s board, and Microsoft management only learned of the announcement minutes before it was announced.

On November 17, OpenAI, the hottest startup in the United States in the field of large models, released a statement that Sam Atlman will step down as CEO and leave the board of directors. Mira Murati, the company’s chief technology officer, will serve as interim CEO, effective immediately.

Sam Altman has been consistently untruthful in his communications with the board, hindering the board’s ability to carry out its duties. The board no longer has full confidence in his continued leadership of OpenAI. At the same time, Greg Brockman, the other founder and president, will step down as chairman of the board and continue in his role with the company, reporting to the CEO.

On November 18, Sam said on his social media account that he enjoyed his time at OpenAI and would announce his next arrangement later. Greg then retweeted the post and said that after learning of the news, he sent a message to OpenAI saying that he had decided to quit OpenAI.

In other words, eight years after its founding, OpenAI has become the most watched technology company in the world for less than a year, and the two core founders have left at the same time. Both Sam expressed shock at the incident, and they were still trying to figure out what the cause was.

He mentioned that on Nov. 16, Sam received a text message from chief scientist Ilya Sutskever asking to talk at noon on Friday. Sam attended a Google online meeting where all the board members were present except Greg. Ilya tells Sam he’s going to be fired.CXO UNION CXO联盟 cxounion.cn

On November 17, at 12:23 PM, Greg was informed that he was being removed from the board (but he was vital to the company and would retain his position) and Sam had been fired. Later, OpenAI announced the news. As of now, OpenAI has no more new explanations for this personnel change, including what specific “dishonest behavior” Sam has that the outside world is most concerned about. Sam has been the face of OpenAI, and last week at the developer conference, he hosted and announced a series of new OpenAI products and services. CXO UNION CXO Union cxounion.cn

On Nov. 16, the day before he was fired, he also spoke on behalf of OpenAI at the Asia-Pacific Economic Cooperation Conference.

According to several US media reports, on November 17, OpenAI interim CEO Mira said at the company’s plenary meeting that the current relationship between the company and Microsoft remains stable, and Microsoft’s CEO and CTO have expressed “great confidence” in the future of OpenAI.

Who made the decision?

According to US media reports, Microsoft’s leadership was only informed of the news a few minutes before it was announced. Previously, OpenAI’s board was made up of six members, and after Greg and Sam stepped down, It is now made up of Ilya Sutskever, chief scientist at OpenAI, Adam D’Angelo, CEO of Quora, technology entrepreneur Tasha McCauley, and Helen Toner of the Center for Security and Emerging Technologies in Georgetown. CXO UNION CXO联盟 cxounion.cn

Microsoft, which has invested $10 billion in OpenAI, does not have a seat on the company’s board. The majority of the board members are independent directors who do not own equity in OpenAI, including former CEO Sam. Sam’s earnings came from Y Combinator’s early, small investment in OpenAI. The Board’s primary responsibility is to advance OpenAI’s mission and uphold its charter.

Founded in 2015, OpenAI is a non-profit organization with the core mission of ensuring that Artificial General Intelligence (AGI) can benefit all of humanity. OpenAI benefits humanity, not investors. Considering that the development of AGI requires a large amount of financial support, in 2019 OpenAI announced the adjustment to a “profit cap” model, and the board of directors is the overall governing body for all OpenAI activities. The specific model is to create a new for-profit subsidiary, which will be used to raise funds, but will be obligated to fulfill the mission of the non-profit organization, which is safety and benefit all mankind. CXO UNION CXO联盟 cxounion.cn

The for-profit subsidiary, which is fully controlled by OpenAI, will be allowed to make money but will be subject to profit caps, including limits on Microsoft’s investment returns. The composition of OpenAI’s board, which made the decision to oust core founder Sam Altman from the company, is somewhat unusual. Of the four current directors, only Ilya is an OpenAI employee, and he is an AI scientist who joined OpenAI as a scientist at Google. In the past, when he appeared in the public eye, he mostly discussed the technical details of OpenAI. He currently oversees artificial intelligence.

Adam D’Angelo joined OpenAI’s Board of Directors in April 2018. At the time he had a point: Safe AGI work was both important and underappreciated. His view is that OpenAI shouldn’t become a tech giant because it won’t make the world a better place. Tasha McCauley, who joined OpenAI’s board earlier this year, has founded several startups. She currently serves on the advisory board of the UK-based International Centre for AI Governance, along with another director, Helen Toner. Helen Toner joined the OpenAI Board of Directors in September 2021 and is responsible for AI security issues. Gregg has said he “greatly values Helen’s deep thinking about the long-term risks and implications of AI.” Normally, a company’s board members have separate oversight responsibilities in different areas, but OpenAI is completely different, and the main responsibility of all four directors is AI security. As OpenAI’s highest governing body, the board of directors represents the soul and purpose of the company: non-profit and security, but under Sam’s leadership OpenAI has gradually strayed from this direction. CXO UNION CXO联盟 cxounion.cn

Non-profit, security and ambition

Greg, who resigned along with Sam, was one of the founders of OpenAI. According to Greg’s article, Sam asked him about his future plans in 2015, and he said he wanted to do AI, but he didn’t know yet. Sam said he considered setting up an AI lab through the YC Incubator with the goal of “building safe human-level AI,” perhaps in partnership, and they quickly agreed that this would have to be done through a non-profit organization because there wouldn’t be any conflicts of interest that would interfere with that mission.CXO UNION CXO联盟 cxounion.cn

But Sam’s Silicon Valley “people” are ambitious, YC founder Paul Graham (Paul Graham) once said that he “ambition beyond the boundaries of Silicon Valley can accommodate”, “if you throw him on some cannibal island, five years later he will be the king of the island.” In 2014, Sam, less than 30 years old, served as the CEO of YC business incubator, and led YC’s market value to close to 150 billion yuan in five years.

In 2015, he co-founded OpenAI with Musk, Greg and others, and until 2019, he was an investor in OpenAI.

In 2019, he stepped down as president of YC to become CEO of OpenAI. This year, OpenAI began to run on the road to commercialization, including the launch of a paid version of the product, customized enterprise edition and so on.CXO UNION CXO联盟 cxounion.cn

In August, it was announced that OpenAI2023 will reach $1 billion in annual revenue and continue to grow by several billion dollars in 2024. In October, Sam told OpenAI employees that the company would reach $1.3 billion in revenue this year. In 2022, the company’s revenue is only $28 million. Under his leadership, OpenAI was recently valued at $90 billion.

In September 2022, during a conversation with Reid Hoffman (co-founder of Linkedin), he mentioned that the real challenges in AI today are wealth distribution, AGI access, and AGI governance. As AI technology evolves, AGI will become a commodity, and consideration needs to be given to giving everyone equal access to AGI.

In addition to commercialization, another level of possible conflict between Sam and the board is “security.” Under Sam’s leadership, OpenAI’s revenue has grown rapidly, with specific measures including C-side fees, customized B-side services, open GPT interfaces and application ecosystems, but these have made OpenAI face security criticism. This is a natural contradiction between AI security and commercialization. CXO UNION CXO联盟 cxounion.cn

On November 8, ChatGPT and the API interface experienced a “serious outage” event, with the entire outage lasting more than 12 hours. The main reasons were a large influx of users after the ChatGPT update was announced on Developer Day on November 7, as well as a total ChatGPT outage due to DDoS (distributed denial of service, often caused by network overload) and authentication failures.

Sam later apologised on social media. Usage of the new feature has far exceeded expectations, he said, and the company had planned to enable GPT for all subscribers on Monday, but it still hasn’t happened. Service instability may occur in the short term due to load. The main concern surrounding OpenAI security comes from data issues, and many people are concerned about data leakage during conversations with ChatGPT.CXO UNION CXO联盟 cxounion.cn

Earlier this year, the Italian Data Protection Authority (GPDP) accused OpenAI of violating the European Union’s data protection law, and the case is still pending. The most recent controversy over OpenAI’s security came on November 9, when Microsoft cited “security and data issues” in an update to its internal website, thus cutting off employees from using AI tools like ChatGPT. But a day later, on November 10, Microsoft restored it. A Microsoft spokesperson said that Microsoft was testing an endpoint control system, and there was an error, so it was quickly restored.

In May, Apple limited employees’ use of ChatGPT and other external AI tools as it worked on the technology and was concerned about confidential data leakage. In the same month, Samsung Electronics issued an internal memo banning employees from using generative AI tools, including ChatGPT, BingChat, and Google Bart, because Samsung engineers had uploaded internal source code to ChatGPT in April.

In July, the U.S. Federal Trade Commission (FTC) launched an investigation into whether OpenAI harmed people by publishing false information, and whether OpenAI engaged in “unfair or deceptive” privacy and data security practices. The agency is concerned that ChatGPT and other AI services are being fed large amounts of data without AI companies performing checks on data types. Sam later said that he would cooperate with the FTC and that he believed OpenAI’s technology was safe. Large models rely on a constant stream of new data to optimize, and commercialization requires large models to continue to advance in the field of expertise. As long as OpenAI is pursuing revenue growth, it will inevitably face data security challenges. Musk, who was an early investor and board member of OpenAI, has repeatedly said that he does not approve of the development of OpenAI, including Microsoft’s strong control over OpenAI (this claim has been denied by Microsoft), and OpenAI’s insufficient control over AI security.

In February of this year, he wrote on the X account, “OpenAI was originally created as an open source non-profit company to compete with Google, which is why I named it OpenAI. But now it’s a closed-source, for-profit company effectively controlled by Microsoft, which is the opposite of what I intended.”

Musk stepped down from OpenAI’s board in 2018. Another former board member, Reid Hoffman, quit in March because he planned to invest in a software company using OpenAI products, fearing a conflict of interest.

本文由CXO UNION-CXO联盟（cxounion.cn）转载而成，来源于财经十一人；编辑/翻译：CXO UNIONCXO联盟小C。

如需加入CXO UNION（CXO联盟）高管社群，请联系社群小伙伴哦~

免责声明: 本网站(http://www.cxounion.cn/)内容主要来自原创、合作媒体供稿和第三方投稿，凡在本网站出现的信息，均仅供参考。本网站将尽力确保所提供信息的准确性及可靠性，但不保证有关资料的准确性及可靠性，读者在使用前请进一步核实，并对任何自主决定的行为负责。本网站对有关资料所引致的错误、不确或遗漏，概不负任何法律责任。

本网站刊载的所有内容(包括但不仅限文字、图片、LOGO、音频、视频、软件、程序等) 版权归原作者所有。任何单位或个人认为本网站中的内容可能涉嫌侵犯其知识产权或存在不实内容时，请及时通知本站，予以删除。