Source: Tencent Technology
According to a recent report by , in 2016, OpenAI began negotiations with Microsoft on a cooperation agreement, as this newly established startup hoped to obtain critical computing resource support to build the chatbot ChatGPT. However, as the CEO of OpenAI, Sam Altman first had to secure the support of the biggest "patron", Elon Musk.
According to the latest disclosed emails, in September 2016, Altman sent Musk a message saying: "We only need $10 million to get $60 million worth of computing resources, and we can also provide advice on Microsoft's cloud deployment technology." Microsoft hoped that OpenAI could provide feedback and promotion for Microsoft's AI tools, such as Azure Batch. However, Musk expressed disgust at this idea, saying it made him "sick".
Altman then proposed another deal: "Microsoft is now willing to reach an agreement with us for $50 million, and promises that 'OpenAI will independently assess whether it has made its best efforts' and both parties have the right to terminate the agreement at any time. The agreement does not contain any promotional content or other additional conditions. This way, we won't be seen as Microsoft's clumsy marketing pawns."
Musk responded: "If they don't use this feature in the messages they send out, I think it's fine. But if we can avoid being seen as Microsoft's 'marketing whores', then the value of this agreement will far exceed $50 million."
Last week, Musk publicly released these emails and other related emails as part of his lawsuit against OpenAI and Microsoft. On the surface, Musk's move was to prove the anti-competitive partnership between Microsoft and OpenAI. But more crucially, these emails revealed the early collaboration details and subsequent power struggle between Altman and Musk. Musk invested $50-100 million in OpenAI in its early days.
These emails also trace the evolution of OpenAI from an open-source non-profit organization to Microsoft's "de facto closed-source subsidiary", during which the company abandoned its original intention to develop AI for the benefit of all humanity. They also witnessed the complete breakdown of the collaboration between Musk and Altman.
An OpenAI spokesperson said in a statement: "Musk has now tried to change his story three times in less than a year, and this latest statement is even more unsupported and exaggerated than the previous ones."
Musk and Altman co-founded OpenAI because they were concerned that AI technology equivalent to human-level intelligence would fall into the hands of tech giants like Google. However, they ultimately found that OpenAI had gradually evolved into the kind of tech giant they were concerned about. After successfully securing the CEO position of OpenAI, which Musk had long desired, Altman decided to keep OpenAI's cutting-edge AI technology strictly confidential, claiming it was too dangerous to be publicly released.
This decision angered Musk, who then left the OpenAI board and founded his own AI startup xAI last year. Nearly a decade after co-founding OpenAI, Musk is now taking this competition to court, fighting a fierce battle over "the unstoppable future of computing".
1. Caught between two forces, lacking both talent and funding
From the beginning, OpenAI was caught between two conflicting forces: its lofty mission to benefit humanity on the one hand, and the fierce competition with tech giants on the other. Musk and Altman unanimously believed that attracting top talent and raising large amounts of cash were crucial, regardless of their respective motivations. This early compromise laid the groundwork for Musk's later criticism of OpenAI's pursuit of profit over its founding principles.
In 2015, OpenAI was still known as YC AI, an experimental lab under the non-profit research division YCR of the incubator Y Combinator. As the then-president of Y Combinator, Altman actively leveraged his extensive network and resources to attract researchers and funding. Musk urged Altman and CTO Greg Brockman (now president) to seek over $100 million in funding, warning them that any amount less would be negligible compared to the deep-pocketed tech giants like Google and Facebook.
In an email that year, Musk wrote: "I think we should say we will start with a $1 billion funding commitment. This is true, and I will pay any amount that others do not provide."
However, despite Musk's support and tens of millions in funding, this nascent organization still faced the challenges most startups encounter: fierce competition for top talent. Although OpenAI may now be the hottest employer in Silicon Valley, a decade ago (or before the launch of ChatGPT), many leading AI researchers were not enthusiastic about it.
To attract the best AI researchers, Altman and his team designed a unique compensation plan: a $175,000 base salary, the title of "part-time partner" at YC, and 0.25% equity in each batch of YC startups. Today, OpenAI's AI researchers earn close to $1 million per year. In an email to Musk, Altman referred to this plan as the "AI Manhattan Project", confident that he could attract the top 50 researchers to join this non-profit effort that would benefit the world while providing startup-level compensation.
Altman's initial goal was to build a founding team of 7-10 elite members, and he was determined to attract the industry's top talent at all costs. However, they faced fierce competition from Google's AI lab, DeepMind.
In a letter to Musk in December 2015, Altman expressed his concerns: "DeepMind plans to make lucrative offers to every member of OpenAI tomorrow, trying to kill us. Do you object to giving each person an additional $100,000 to $200,000 per year in salary? Although I believe they are all passionate about the mission, the raise will be a positive signal that we will take good care of them." He added, "It sounds like DeepMind is gearing up to wage war with us."
Musk approved the raise proposal. By February 2016, the founding team of OpenAI received $275,000 in annual salary and YC equity, while other employees enjoyed $175,000 in salary and $125,000 in performance bonuses, or equivalent YC or SpaceX stock. OpenAI co-founder Ilya Sutskever, the former Chief Scientist, was an exception, reportedly earning over $1.9 million in 2016 after "rejecting offers several times the compensation OpenAI provided".
Musk responded, "We need to do our utmost to attract top talent. We need to continue to increase our investment. If we need to review the compensation of existing employees in the future, that's fine. Either we get the best talent in the world, or we will be defeated by DeepMind. I am willing to support it as long as we can cultivate excellent talents." He also warned that DeepMind's victory has brought him "tremendous mental pressure", "they are obviously making significant progress, and given the talent level there, they should be making such progress".
2.Musk and Altman both want to be CEO,their motives are questioned by other co-founders
Over time, significant differences have emerged between Musk and Altman.
In August 2017, just as OpenAI was finalizing the details of a $200 million to $1 billion Series A round, some internal conflicts began to surface. Shivon Zilis, a former OpenAI board member and also the operations director of Musk's brain-computer interface company Neuralink, wrote to Musk, mentioning that Brockman and Sutskever were concerned about the equity and control distribution of OpenAI's newly established for-profit entity, and also expressed doubts about whether Musk could devote enough time while serving as CEO. Musk responded, "This is so annoying. Please encourage them to start a company. I'm done with this."
The following month, Sutskever and Brockman jointly sent an email to Musk and Altman further elaborating on their concerns. They pointed out that if Musk served as CEO, he might "unilaterally and absolutely control" the development of artificial general intelligence (AGI). They also questioned Altman's motives, asking why he was so keen on the "CEO title" and inquiring whether "developing AGI is really his primary motivation? What is the relationship between this and your political goals? How have your thoughts changed during this period?"
Although the email did not elaborate on the specific meaning of the term "political", Altman had indeed begun to actively voice his political views in electoral campaigns in California earlier that year. Sutskever and Brockman stated that they were blinded during the early fundraising process and failed to raise these concerns in a timely manner.
DeepMind CEO Demis Hassabis
They continued, "OpenAI's goal is to create a better future and prevent AGI (artificial general intelligence) from taking over everything. We are also concerned that Demis Hassabis of DeepMind may create an AGI hegemony. So if you plan to become a ruler, or create a system that can make you a ruler, that's definitely not a good idea, especially considering the efforts we are making."
However, Musk was not convinced. He responded, "I won't fund OpenAI anymore unless you make a commitment. This makes me feel like a fool, basically I'm providing free funding for you to create a startup. Discussion over."Altman, on the other hand, responded that he was still "passionate about the non-profit structure", a stance that ultimately led Sutskever and Brockman to concede.
After this confrontation, the situation became even more complicated. Zilis relayed a conversation she had with Altman, in which Altman expressed that his trust in Brockman and Sutskever had significantly decreased, feeling that their information transmission was inconsistent and sometimes even childish. As a result, Altman decided to take a 10-day leave to reassess his level of trust in the two co-founders and whether he was still willing to continue working with them.
However, just five months after Sutskever and Brockman expressed concerns about the power struggle, the situation reached another critical turning point. In a heated argument, Musk was convinced that OpenAI had fallen far behind Google, and he proposed to take control of the company. This was precisely the scenario that Brockman and Sutskever had warned about, where Musk might attempt to unilaterally control the company's direction.
In 2018, Musk expressed a pessimistic outlook on OpenAI's competitive prospects against DeepMind and Google. Due to the management team's rejection of his proposal, Musk left the board in February 2018 and cut off financial support, but he still provided advisory services to the company.
Musk's departure posed a serious threat to OpenAI's non-profit model. Since he had previously invested $100 million, his departure left OpenAI in urgent need of finding new sources of funding. In 2019, to reduce its reliance on Musk and provide funding for its training data centers, OpenAI designed a unique structure: a for-profit company controlled by a non-profit organization. This structure attracted the attention of investors such as LinkedIn co-founder Reid Hoffman and venture capitalist Vinod Khosla, who participated in the first round of funding and promised nearly $1 billion, although the initial capital was only $130 million.
During this period, Musk sent an email to Altman, hinting that he was involved in the company's structural adjustments and requesting that Altman explicitly state that he had no economic interest in OpenAI's for-profit division. Altman simply replied, "Okay." However, this email later became part of the litigation.
3.The power struggle at OpenAI is not just a personal feud
OpenAI wields tremendous influence in the artificial intelligence industry, and the battle for control within the organization has been closely watched. Although the struggle between Musk and Altman was intense, Altman ultimately emerged victorious, gaining almost complete control over OpenAI.
Musk's allegations against OpenAI and Microsoft may have legal value, and his lawsuits have largely focused on accusing Altman of hypocrisy, which is often difficult to substantiate in court. However, this lawsuit has provided Musk with a rare opportunity to publicly reveal some details about OpenAI's origins and his own role. The evidence shows that Altman had already amassed a considerable degree of power within the company from the very beginning, which may not have fully aligned with the initial intentions of the other co-founders.
Furthermore, this lawsuit has also exposed the complex relationships and power struggles within OpenAI. Altman's willingness to confront for-profit competitors from the outset highlights his dominant position within the company. At the same time, this incident also reveals that Musk's desire for power has not diminished, and he may have hoped to use this lawsuit to destroy Altman's reputation and consolidate his own position as a legitimate manager of AGI.
However, we must also recognize that the narrative presented in this lawsuit may not be complete. Many important conversations and decisions may have occurred privately and were not included in the litigation materials. Therefore, we cannot fully understand all the details and background of this power struggle.
In summary, the battle for control of OpenAI is not just a personal feud, but also a struggle over the future direction and strategic positioning of the artificial intelligence industry. The outcome of this struggle will have a profound impact on OpenAI and provide valuable insights for the entire industry.
