The high-stakes world of artificial intelligence has always been fraught with intense personalities and ambitious visions. A pivotal moment in the early days of OpenAI, an organization now valued at an astounding $852 billion, unfolded in August 2017 at Elon Musk’s sprawling 47-acre estate in Hillsborough, California. Co-founders Greg Brockman and Ilya Sutskever met with Musk, then-girlfriend Amber Heard, who served whiskey, before the conversation turned to the future of AI development.
According to Brockman’s federal court testimony in the recent Musk v. Altman trial, the atmosphere was already charged. Musk had pre-emptively gifted Brockman and Sutskever brand new Tesla Model 3 cars, a gesture Brockman interpreted as an attempt to “butter us up” and make them feel “indebted.” In return, amateur artist Sutskever presented Musk with a painting of a Tesla, setting the stage for negotiations that would dramatically shape OpenAI’s trajectory.
The Billion-Dollar Standoff: A Mansion, Teslas, and AI Dreams
The core discussion revolved around establishing a for-profit arm for OpenAI, a necessary step to attract the billions of dollars in investment needed for extensive compute resources. However, Musk demanded absolute control of the company, which Sutskever and Brockman vehemently opposed, viewing it as a “dictatorship” over the future of AI. They proposed a model of shared control, aiming for a more collaborative leadership structure.
Musk’s reaction was swift and explosive. After several minutes of deliberation, he rejected their offer, reportedly storming around the table and, in Brockman’s words, appearing as if he might “physically attack” him. Musk then grabbed Sutskever’s painting, declared he would cut off his funding to the nonprofit, and demanded Brockman and Sutskever quit before abruptly leaving the room. However, later that night, Musk’s chief of staff, Shivon Zilis, called them, signaling that “it’s not over.”
This dramatic account emerged during Brockman’s testimony, painting a picture that OpenAI believes undermines Musk’s current legal arguments against the company. Musk contends that his roughly $38 million in donations were improperly used, leading to the creation of a massive for-profit entity responsible for innovations like ChatGPT and Codex. Brockman, along with OpenAI CEO Sam Altman, and the company itself, deny any wrongdoing, with a jury expected to deliberate on an advisory ruling soon.
Vision Clash: “You Needed to Dream a Little Bit”
Further illustrating his concerns about Musk’s leadership, Brockman recounted another telling incident. He recalled an early OpenAI researcher, Alec Radford, demonstrating an nascent AI chatbot to Musk, who was unimpressed. Musk allegedly dismissed the system as “so stupid, that a kid on the internet could do better,” deeply demoralizing Radford to the point where he almost abandoned AI research entirely.
Brockman and Sutskever spent significant time rebuilding Radford’s confidence after this encounter. In Brockman’s view, Musk’s inability to recognize the potential in this early technology—a precursor to today’s ChatGPT—rendered him unfit to control OpenAI. He emphasized, “You needed to dream a little bit,” a quality he felt Musk lacked in that crucial moment of innovation.
As negotiations about a for-profit entity dragged on for months, Brockman, Sutskever, and Altman considered voting Musk off the OpenAI nonprofit board. Despite meeting again to discuss alternative funding, a consensus proved elusive, though they ultimately decided removing Musk felt “wrong.” Consequently, Musk eventually resigned from the board on his own in early 2018, having emailed that OpenAI was on a path of “certain failure.”
Boardroom Battles and Shifting Alliances
Shivon Zilis, then an adviser to both OpenAI and Musk, maintained her unique position. She acted as a “proxy Elon,” keeping him abreast of developments at OpenAI even after his departure. Brockman described her as a friend, with whom he first connected in 2012 or 2013, highlighting the complex web of relationships at play.
Zilis herself faced scrutiny after joining OpenAI’s board in 2020 and giving birth to Musk’s twins in 2021. Brockman learned of Musk’s paternity through news articles, not directly from Zilis. When confronted, she claimed the children were conceived via IVF and her relationship with Musk was “entirely platonic.” Despite calls from several board members for her removal, Brockman and Sutskever successfully argued for her to remain, believing she helped manage Musk’s frustrations with OpenAI.
Zilis ultimately left the OpenAI board in 2023, following Musk’s launch of his rival AI lab, xAI. This wasn’t the only instance of internal board conflict for OpenAI. Brockman also advocated for the removal, or at least partial recusal, of Quora CEO Adam D’Angelo from the board after Quora launched a chatbot in February 2023 that directly competed with OpenAI’s ChatGPT. D’Angelo remains on the board today.
Brockman additionally supported the removal of AI safety researcher Helen Toner from the board, though he did not specify the exact reason during his testimony. Toner eventually resigned from her role in 2023 after her involvement in the short-lived firing of Sam Altman, a move that quickly backfired and led to his reinstatement. These intricate dynamics underscore the intense pressures and evolving allegiances within the cutting-edge world of AI development.
Source: Wired – AI
