OpenAI審判續集：馬斯克暴走、輸血特斯拉與「鬼屋會議」內幕

Original｜Odaily Planet Daily（@OdailyChina）

Author｜Wenser（@wenser 2010 ）

On April 28, Musk and a group of OpenAI executives faced off in federal court in Oakland, California. (See “Musk vs. Altman: The Biggest Case in AI History Goes to Trial”)

The man, bearing the title of the world’s richest person, co-created OpenAI and then, like Jobs before him, stormed out in frustration. With the release of insider information such as opening statements and court evidence on May 6 and 7, more unknown details of this case involving two trillion-dollar tech giants, SpaceX and OpenAI, have gradually been exposed to the world. These include both Musk’s various maneuvers and the private interests and public principles of OpenAI executives.

Based on public information and court documents, Odaily Planet Daily has compiled eight stories for readers to gain insight into the hidden circumstances behind this “biggest AI trial in history,” which involves a potential settlement of up to $134 billion.

Insider Story 1: Musk Threw in $38 Million in Seed Funding, OpenAI Exec Greg Brockman Sat on $30 Billion in Equity

During the second week of Musk’s lawsuit against OpenAI, OpenAI co-founder and President Greg Brockman and his personal diary from 2015 to 2023 appeared as a witness and evidence in California court. Musk’s lawyer, Steven Molo, quickly found a “weak spot”—compared to Musk’s injection of $38 million in hard cash to support OpenAI’s founding and early development, Greg Brockman himself stated that he “didn’t invest a single cent in OpenAI’s founding and development.” Yet today, based on OpenAI’s valuation of $825 billion after its latest funding round, his personal OpenAI equity is worth approximately $30 billion.

Greg Brockman’s diary also laid bare his “wealth ambitions,” mentioning:

• “Financially, what will take me to $1B?” ; “It would be nice to be making the billions.” / “We’ve been thinking that maybe we should just flip to a for profit. Making the money for us sounds great and all.”
• “Can’t see us turning this into a for-profit without a very nasty fight... It’d be wrong to steal the non-profit from him. That’d be pretty morally bankrupt... and he’s really not an idiot. His story will correctly be that we weren’t honest with him in the end about still wanting to do the for-profit just without him.”
• Furthermore, he also wrote, “This is our only chance to get rid of Musk... Let me make $1 billion.”

Although Greg Brockman argued that some of these were hypothetical scenarios after the board ousted Musk, it hardly painted a picture of someone indifferent to fame and fortune.

The reason this is emphasized is that Greg Brockman had previously promised to donate $100,000 to the OpenAI non-profit foundation, a pledge never fulfilled. When asked, “Did you want to fund a non-profit, or did you want to become a billionaire through OpenAI?” he magnanimously replied, “Having $10 billion in stock is already very satisfying.” However, when Musk’s lawyer Molo asked why he didn’t donate the remaining $29 billion in equity to the OpenAI non-profit foundation, Greg was left speechless on the spot.

For comparison, Greg invested $471 million in his previous company, Stripe, and holds shares in cloud computing provider Corweave, one of OpenAI’s partners. The co-founder, who joined via technical contribution, suddenly became the center of public opinion for having it both ways.

Insider Story 2: Former OpenAI Board Member Met Musk at a Company Retreat, Then Accepted Sperm Donation and Had Four Children

On May 7, Beijing time, Shivon Zilis, a former OpenAI board member and mother of four of Musk’s children, also appeared as a witness in court.

According to her, she first met Musk at an OpenAI company retreat. She joined OpenAI in 2016 and served on the board for several years.

After deciding to become a single mother and have children, Musk—who has long championed “human-centric” ideals—volunteered to act as a sperm donor for her IVF process.

Regarding their current emotional status, Shivon Zilis stated, “We are in a romantic relationship. Musk visits regularly.” However, she denied claims of being Musk’s “secret agent” or “channel of information”; Musk, in turn, described her as a “close advisor.”

After Musk’s falling out with other OpenAI co-founders in 2018, she still acted as a bridge for communication. She only formally left the OpenAI board after Musk founded xAI, a competitor to OpenAI, in 2023.

Insider Story 3: OpenAI Was Originally Called the ‘AI Manhattan Project,’ Musk Then Named It

In May 2015, Sam Altman, then president of Y Combinator and OpenAI’s founder, sent Musk an email proposing that YC lead the creation of a “Manhattan Project-style AI lab.” (Note by Odaily Planet Daily: Referencing the atomic bomb project led by Oppenheimer.)

Ultimately, Musk himself named the new AI lab “Open AI Institute,” abbreviated as “OpenAI” (taking a cue from the Open Source philosophy)—a name that carried the core ideal behind OpenAI’s birth: openness, transparency, and service to all humanity.

Looking back, with the development of the AGI path and the awkward situation of the two now facing off in court, Sam Altman perhaps long saw that OpenAI and AI large models would, like the atomic bomb, transform from a hoped-for “tool to end war” into a “weapon of mass destruction threatening the world.” OpenAI’s complete transformation into a for-profit company planning an IPO also signals its departure from the once-open source and open spirit.

Insider Story 4: Musk Poached Former Google Researcher Ilya Sutskever to OpenAI with a Single Phone Call

In 2015, Ilya Sutskever, who had been acquired by Google, suddenly received a “cold email” from Sam Altman (Note by Odaily Planet Daily: A personal email sent directly to someone completely unknown with no prior interaction), inviting him to dinner with Greg Brockman and Musk to discuss forming a new AI research organization (later OpenAI). At that time, Altman emphasized the organization’s goal was to develop AGI benefiting all humanity, avoiding monopoly by a few giant companies.

By then, Sutskever had worked at Google for nearly three years and was highly regarded and strongly persuaded to stay by the Google Brain team and DeepMind founder Demis Hassabis.

Facing the “Chief Scientist” offer from the fledgling OpenAI, which hadn’t achieved much yet, Ilya Sutskever hesitated despite sharing their philosophy, harboring various concerns. Eventually, Musk called him on the day of OpenAI’s official public launch, convincing him to switch jobs, thereby bringing in a foundational figure for OpenAI’s early development.

Unfortunately, in November 2023, Ilya Sutskever, then a board member, participated in the ousting of Sam Altman as CEO due to disagreements over the direction of AI safety versus commercial development. The incident ended with Altman taking control of the board, Ilya publicly apologizing and resigning from his board position. In May 2024, he officially left OpenAI. Altman later publicly thanked him, saying, “Without Ilya, there would be no OpenAI today.”

Insider Story 5: OpenAI Nearly Became a Tesla Subsidiary, OpenAI Employees Once Worked for Tesla for Free

These two pieces of information come from the aforementioned former OpenAI board member Shivon Zilis and OpenAI co-founder and President Greg Brockman.

According to Shivon Zilis, around 2017, about two years after OpenAI’s founding, Musk, Sam Altman, and others were struggling with computing resources and funding, racking their brains daily to find money and GPU resources.

One day, Musk suddenly proposed merging OpenAI into Tesla as a subsidiary, turning it into an internal AI lab to seek more funding and resource support. Additionally, Musk prepared a Tesla board seat for Sam Altman. Of course, due to strong insistence from Sam Altman and Greg Brockman, the plan was ultimately not realized. It was after this that Musk gradually parted ways with OpenAI’s current founding team, leading to a complete breakdown in 2018 and his frustrated exit.

Furthermore, according to Greg Brockman, Musk once pulled several OpenAI employees to work for free on Tesla’s autonomous driving team, including former OpenAI researcher Andrej Karpathy, who later officially joined Tesla and is now one of the “big names in the AI circle.”

Insider Story 6: Musk ‘Bribed’ Greg Brockman and Ilya Sutskever with Founder Series Tesla Model 3s

In July 2017, on the eve of the “Haunted House Meeting” over control of the majority of OpenAI’s equity, Musk emailed some OpenAI executives, stating clearly: “As a thank you for your contributions to OpenAI, I’d like to give each of you a Founder Series Model 3. These are the first ones produced and not yet available to the public.”

In August 2017, Sutskever texted Brockman about this, saying: “At least we’re getting our Teslas.” “Will a Model 3 make you be willing to accept massively unfavourable terms?” By then, OpenAI’s top leadership was already prepared for Musk’s “sugar-coated bullets.”

Greg Brockman also stated in court that Musk’s “Tesla car reward” wasn’t given for good work but to gain greater influence within OpenAI, a way to “curry favor” with him and Ilya Sutskever (Note by Odaily Planet Daily: arguably a form of bribery). Because these cars were delivered in late August 2017, just before the meeting on equity distribution for OpenAI’s for-profit entity.

Interestingly, as a goodwill gesture, Ilya Sutskever, then OpenAI’s Chief Scientist, specially commissioned a painting of a Tesla Model 3 to give to Musk at the subsequent “Haunted House Meeting.”

Insider Story 7: Musk Suddenly Flew into a Rage at the Haunted House Meeting, Almost Physically Threatening Someone

In August 2017, after OpenAI defeated top human players in a Dota 2 game competition, Musk suggested a “celebration,” inviting the OpenAI team to a party at his newly purchased mansion in Hillsborough, California, a 47-acre property worth $23 million.

Due to the property’s outdated decor, poor maintenance, and eerie atmosphere (reminiscent of a Gatsby-esque strange mansion), Musk jokingly called it the “haunted mansion,” even sending an email warning everyone they might see “party carnage.”

According to Greg Brockman’s recollection in court, aside from the OpenAI team, Musk’s then-girlfriend, Amber Heard, was initially there. She poured whiskey for everyone and then left with friends. The atmosphere started cordial, but things took a turn when they discussed the “next step plan” for “OpenAI transitioning to a for-profit entity.” Specifically, when Musk didn’t get the response he wanted regarding equity distribution and control, he seemed to transform, “suddenly stood up, walked around the table in a rage, looking very angry.” Brockman stated directly: “At that moment, I really thought he was going to hit me.”

Finally, Musk grabbed the painting given by Ilya Sutskever, announced he would cut off funding for OpenAI unless Brockman and Sutskever resigned, and stormed out of the room, ending the party on a sour note.

Insider Story 8: Musk’s ‘Biding His Time’ Fight for OpenAI Control Was Only to Realize His ‘Mars Dream’?

In court, when asked why Musk was so determined to gain control of OpenAI, Greg Brockman said that Musk had told him part of the reason was: He needed $80 billion to realize his grand vision of building a city on Mars.

On the other hand, SpaceX’s IPO was also proceeding rapidly, and its fundraising target happened to be around $75 billion, quite close to the $80 billion mentioned by Brockman.

Eight years later, the root cause of the love-hate relationship between Musk and OpenAI may finally be clear: Musk needed OpenAI as a key vessel to transfer resources to SpaceX. If he had successfully gained control back then, he might, like Sam Altman and Greg Brockman, have still transformed OpenAI into a for-profit entity. The difference is that Musk might not have needed to go to the trouble of creating xAI from scratch and eventually merging it into SpaceX.

Of course, according to the latest news, Musk has turned to cooperate with Anthropic, OpenAI’s biggest competitor, betting his existing computing resources on them to achieve his ultimate dream—landing on Mars by indirect means. See “Musk and Anthropic Go Looking for Electricity in Space”.

To conclude, we leave you with a small anecdote from the trial of this “first case in the AI field,” which involves potential damages of up to $134 billion.

According to documents filed by OpenAI’s lawyers, two days before the trial officially began, Musk texted Greg Brockman to sound out OpenAI’s willingness to settle. When Brockman mentioned that “both sides should drop their claims,” Musk retorted sharply: “This weekend, you and Sam will be the most hated people in America. If you insist, then so be it.”

Although Judge Yvonne Gonzalez Rogers, presiding over the case, ultimately did not admit the text message as evidence, given the current situation, Musk and OpenAI’s “legal battle” is far from over.