Adi Robertson

Musk tells the court that “at various times we discussed, we brainstormed about different ways to fund the charity,” including a for-profit structure. “We did talk about establishing a for-profit or Tesla providing some of the funding — there were a bunch of ideas that were brainstormed — I was not averse to a small for-profit that would provide funding to the nonprofit as long as the tail didn’t wag the dog.” These were “informal verbal conversations and email and text discussions” between Musk, Brockman, Altman, and Sutskever, he recalls.

Musk is trying to establish how instrumental he was to getting Huang and Nvidia to supply OpenAI with hardware. “I was asking Jensen for access to the first AI supercomputers they were making,” he says — referring to Nvidia’s DGX lineup — as an exhibit of a 2016 email conversation enters the record. “This was the first AI supercomputer ever developed and it would make a huge difference to OpenAI if we could get one … I wanted to be clear with Jensen that this was — I’m emailing from my Tesla email here, but I wanted to be clear this is not a Tesla request.” The conversation shows Huang promised “I will make sure OAI is one of the first ones” to get one.

“I came up with the idea, the name, recruited the key people, taught them everything I know, provided all the initial funding. Besides that, nothing,” he says (to a few chuckles in the courtroom). “It was specifically meant to be a charity that does not benefit any individual person. I could have chosen to start it as a for-profit and I chose not to.” He calls himself “instrumental in recruiting Ilya Sutskever and most of the initial team.”

Through email exhibits and testimony, Musk emphasizes he believed OpenAI would be a 501(c)(3) nonprofit that would “aim to bring in more money than it spends, and in doing so, that becomes cash reserves or savings” for the charity. “I was fundamentally involved” in the announcement, he says — “reviewed and drafted part of the announcement, reviewed the webpage, and did media interviews and whatnot.” Throughout this, he indicates, Musk thought he and Altman were on the same page.

“To be clear I would have created a nonprofit open-source foundation with or without Sam Altman or [OpenAI cofounder] Greg Brockman, but at the time I was glad they were doing so with me,” he says. The name came about “as the result of discussions with Greg, Sam, and Ilya [Sutskever]. We bounced around a few different names, the one I thought made sense was OpenAI because the ‘open’ in OpenAI represents open source.” (A lack of open-source releases, of course, is one of Musk’s bones to pick with OpenAI.)

Musk says Altman, then at Y Combinator, was “not well known at that time” — but he agreed with Musk on AI safety, which was very important. He reiterates that, by contrast, Larry Page called Musk a “species-ist” for being concerned about humans being wiped out by AI. “OpenAI exists because Larry Page called me a species-ist,” he says.

Musk continues to talk about the dangers of AI, saying he warned President Barack Obama in a meeting years before it was a serious concern. “There is no agreement to limit AI and so there is currently an AI acceleration that’s happening very quickly with a number of companies primarily in the US and with China,” he says. Back in 2015, he recalls, “I had many dinners with many people talking about AI safety” — one of those people was Sam Altman.

He tells the jury that OpenAI exists because of conversations he had with Larry Page of Google, who Musk says did not care sufficiently about the dangers of AI. Musk decided “we’ve got to have some counterpoint to Google,” which at the time was leading in AI research. But Musk says he thought about the technology earlier than that, in the ’90s, when he believed it could “solve all the diseases and make everyone prosperous, or it could kill us all.”

After noting that computers and chatbots have gotten “to the point where they feel almost human,” Musk brings up artificial general intelligence, or AGI. “AGI is when the AI becomes as smart as any human — arguably smarter than any human, and I think we are getting close to that point. My guess is that AI will be probably as smart as any human as soon as next year,” he says.

“The incorporation docs had been filed but there were no employees or IP at that time, so it depends on people’s perspective, but essentially there were five cofounders of the company and I was one of them,” he says. It’s actually, of course, a matter of legal agreement — here’s a little more background.