The founding story of OpenAI and the inside story of the split between Musk and Altman

Author: Lan Xi Source: Weibo

Because Musk is suing OpenAI, many private emails that were not originally made public have been disclosed. Based on the email conversation records, The Verge sorted out the timeline of OpenAI's early top-level design. Although there are a lot of nonsense, it is very interesting to read them together. The real business war is so plain and unpretentious:
- In 2015, OpenAI had not yet appeared. Its predecessor was Y Combinator's AI laboratory. Sam Altman, who was still the president of YC at the time, invested 10 million US dollars out of his own pocket and began to recruit people everywhere;
- At this time, Musk suggested that he should raise at least 100 million US dollars, because any capital below this number would be easily crushed by Google and Facebook;
- Sam Altman sent a heartfelt email to Musk, saying that he had been thinking about whether he should choose to stop the development of AI for the sake of mankind. If he really couldn't stop it, then it might be correct if the leadership of AI could fall into the hands of a company other than Google;
- Then OpenAI went out independently. Musk proposed a small goal of raising $1 billion first, and asked Sam Altman to find as much money as possible, and promised that he would make up the final gap;
- Because OpenAI was inconspicuous in Silicon Valley at the time and could not recruit top talents at all, Sam Altman designed a radical salary plan: a basic annual salary of $175,000, plus a 0.25% YC equity grant for each batch;
- Sam Altman told Musk that this would be a "Manhattan Project" in the artificial intelligence industry, hoping to attract 50 top researchers to join. Although the company remains non-profit, employees who work hard can still get similar returns to entrepreneurship;
- In 2016, DeepMind, an AI laboratory under Google, noticed OpenAI's actions and began to call everyone contacted by OpenAI, intercepting them with multiples of salary;
- Sam Altman was panicked and quickly wrote an email to Musk, hoping to approve his annual salary to increase by another $100,000-200,000, otherwise DeepMind would have won the war in advance;
- Musk agreed, and the early team received a salary of nearly $300,000 + floating equity. Musk even contributed SpaceX's equivalent stock to fill the gap;
- There were three exceptions, Greg (later president), Ilya (later chief scientist), and Trevor (later researcher), who received higher salaries, but Sam Altman did not seem to include himself;
- Ilya's income in 2016 exceeded $1.9 million, which to a certain extent helped him resist the recruitment from Google;
- Musk agreed with OpenAI's generous salary when it had nothing, "We continue to do our best to attract top talents. If adding money can solve the problem, it would be great. As long as we are not defeated by DeepMind, I can accept any payment";
- Musk was extremely disgusted with Google at that time. He told Sam Altman that as long as he thought that DeepMind might become the winner, he would explode with mental stress, and he must not let Google rule the world;
- In 2016, OpenAI was still short of money, especially chips. Sam Altman went to Microsoft and discussed a plan to exchange 10 million US dollars for computing resources equivalent to 60 million US dollars, on the condition that OpenAI would use Microsoft's Azure cloud service and help promote it in the technology circle;
- Musk told Sam Atlman that this was too disgusting;
- There was no way, Sam Altman went to talk to Microsoft again, and the new offer was that Microsoft would invest 50 million US dollars without any additional conditions, and the cooperation could be terminated at any time. Sam Altman said to Musk with resentment, "You don't look like a Microsoft bootlicker anymore, are you satisfied?";
- Musk replied mercilessly, saying that Microsoft had to be asked not to issue a press release to mention this matter. Just not being a Microsoft bootlicker was worth several 50 million US dollars;
- In 2017, when OpenAI was preparing for a new round of financing, someone Musk had planted—a woman who later gave birth to three children for Musk was on the board of directors—told Musk that the founding team was not happy with Musk being the CEO because they felt that Musk was managing too many companies and could not devote 100% of his time to OpenAI;
- At that time, Tesla was on the verge of bankruptcy. Musk was visibly overwhelmed and had no interest in OpenAI's little tricks. He emailed the insider and said, "I'm so annoyed. Let them start a company. I've had enough."
- Not long after, Ilya and Greg pushed forward the ice-breaking plan again, writing long emails and copying Sam Altman and Musk at the same time, saying that they were worried about Sam Altman's obsession with the title of CEO and doubting whether he was loyal to the ideal of AGI. At the same time, they explained to Musk that the reason for rejecting him as CEO was that they knew that his personality would definitely pursue absolute control over AGI;
- In other words, the core of their doubts is that although the risks of AI are huge, picking one person to control AI may bring greater risks. They want to conceive a more decentralized governance structure.
- Musk and Sam Altman were both devastated at this time. Nothing was done and they put aside this analysis, all kinds of grand narratives on how to determine the fate of mankind. Musk directly replied that the discussion was over. If you just treat me as a rich man who only pays money, I will run away.
- Sam Altman superficially comforted his colleagues, saying that he was confident in the non-profit nature of the company, but privately said that Greg and Ilya were sometimes too naive, and he wanted to take a vacation to think clearly whether he could still trust them not to stab him in the back.
- In 2018, Musk judged that OpenAI's progress had completely fallen behind DeepMind, and once again proposed that he should be the CEO, otherwise it would be impossible to beat Google, but the founding team of OpenAI still rejected him, causing Musk to withdraw from the board of directors and stop capital donations.
- The sudden loss of Musk's money put OpenAI in a passive position for a short time. In order to obtain financing, OpenAI changed the company for the first time and designed a structure of a limited profit company controlled by a non-profit organization;
- Musk had begun to realize that he needed to keep the evidence for future lawsuits. He sent an email to Sam Altman, asking him to clearly prove that he (Musk) had no benefits in the new company structure, and Sam Altman replied neatly: "Received";
- The Verge concluded that the power struggle around OpenAI essentially came from the collision of two visions in different directions. At least the information shown in the email was that OpenAI's ability to turn the tide and reach today's height was inseparable from the driving force provided by Sam Altman. He could always get what he wanted at all costs, and Musk was not short of power. He wanted to retaliate against all behaviors that despised him and firmly believed that he was the steward of AGI.