Could anything but profit steer AI? The OpenAI trial offered clues but no verdict

The trial pitting Elon Musk against OpenAI CEO Sam Altman made clear the two billionaires agreed on one thing: building artificial intelligence would require significant resources — and enormous amounts of money.

It may seem obvious now, as an AI-obsessed stock market helps finance a global construction boom of chipmaking factories and energy-hogging data centers to keep chatbots running, but testimony and evidence showed how people with outsized control of the AI industry were privately debating its costs nearly a decade ago.

“Even raising several hundred million won’t be enough,” Musk said in a 2018 email to Altman and other OpenAI co-founders about what he increasingly saw as a futile attempt to compete with Google. “This needs billions per year immediately or forget it.”

The soaring costs factored into the trajectory of OpenAI, which began in 2015 as a nonprofit dedicated to developing AI for the common good and is now a capitalistic enterprise valued at $852 billion. As San Francisco-based OpenAI and other AI companies move toward historically large Wall Street debuts, the trial also raised questions about whether anything but commercial interests can steer AI’s future.

It is possible to build big things only with nonprofit money, but in the case of OpenAI’s early years, the uncertainty around AI also made it a risky investment, said Karan Girotra, a professor of operations, technology, and innovation at Cornell Tech. Now, he said, investment in AI is no longer speculative.

“Now it’s traditional investment in something we know works,” Girotra said. “People want your car, you need to build the factory ahead of demand.”

In his lawsuit, Musk accused OpenAI of betraying its charitable mission for building AI, saying Altman and fellow co-founder Greg Brockman went behind his back and unjustly enriched themselves. OpenAI, in turn, has said Musk supported plans to form a for-profit company and filed his 2024 lawsuit to undercut the ChatGPT maker’s success as he built his own AI company, xAI.

The jury did not touch questions about AI’s future

The federal jury in Oakland, California, never got to deliver a verdict on the merits of the case, determining Musk’s lawsuit missed a statutory deadline and dismissing it Monday after a three-week trial.

But the trial put on record details of internal battles that presaged today’s societal and political debates over AI’s impacts and costs.

“It’s sort of hard to imagine at this point, given where AI has gotten,” testified Kevin Scott, Microsoft’s chief technology officer, as he explained to jurors why his company opted to invest billions of dollars to help build OpenAI’s technology after founding donor Musk quit OpenAI’s board in 2018.

“It was before ChatGPT,” Scott said. “It was before these remarkable things that are happening right now and so most of the people at Microsoft were very skeptical about whether or not all of these claims were going to materialize into reality.”

Microsoft, a defendant in the lawsuit, at the time was also looking for a way to compete with Google in AI research. OpenAI told Microsoft what they needed was more data and more computing resources — and if they had that, their AI systems would grow far more powerful.

“The things that they wanted and ultimately that we helped them do were very capital-intensive projects like building giant data centers, full of very expensive computers and networks,” Scott said.

Testimony detailed how costs limited options

It remains in dispute how much profit was the prime motivator for the shift to OpenAI’s capitalistic enterprise, which is not yet profitable but likely headed for an initial public offering as soon as later this year.

What is clear, however, is how the costs involved constrained the company’s options.

More than five years before OpenAI introduced ChatGPT, the company had a breakthrough when it taught an AI system to beat professional players of Dota 2, a multiplayer video game featuring ogres, centaurs and other fantastical creatures.

“Honestly, the world reacted to it somewhat less than I thought they should have, but to us internally, it really felt like a moment where we had shown that our technology, using something called reinforcement learning, could take on an enormously complex task,” Altman testified.

OpenAI’s livestreamed victory against a top Dota 2 player at a Seattle competition in 2017 made the tiny nonprofit a major contender against Google, which was then seen as the leader in AI research. It also led to some soul-searching about how OpenAI could compete when it was a nonprofit, largely dependent on Musk and other donors.

“He was impressed,” Altman said of Musk. “And then immediately after the Dota win, Mr. Musk said he thought we really need to get more serious and figure out how to get way more capital.”

For another co-founder and OpenAI’s former chief scientist, Ilya Sutskever, the Dota victory was the beginning of discussion about whether OpenAI should create a for-profit company to more easily raise money.

“The realization is that to make progress in AI, you need a big computer,” Sutskever told jurors. “And you need the big computer because the brain is a big computer. You have a hundred billion neurons and a hundred trillion synapses in the brain.”

What followed was a battle of wills — with Altman and Musk vying for leadership of OpenAI and Musk later trying to fold the AI laboratory into his car company Tesla. The other OpenAI leaders resisted, and Musk eventually quit.

___

AP Technology Writer Barbara Ortutay contributed to this story.