Open(Drama)AI: A story of a lawsuit in three acts.

There is something rather poetic about OpenAI’s name. When its founders first chose it, they were thinking of a world where AI models would be freely accessible for all to use. Instead, almost a decade after conception, the openness could well be a suggestion of its internal drama more than its models.?

For months, OpenAI has become a hegemon for the news world and a perpetual presence in popular discussions. First, as the world discovered the potential of its tools with great awe and skepticism, articles galore began to sing praise of the company. Then, out of the blue, the company fired its chief executive, Sam Altman, and became the center of attention once more. Now, after tech billionaire Elon Musk sued the company for betraying its original mission, OpenAI is back on the spotlight for even more confusing reasons than before. While the openness of its software might be a large topic for discussion—one which we will soon encounter—, its internal drama is not.

This week, we wanted to recount at least a fraction of OpenAI’s story: the part needed to understand Elon’s current lawsuit and the many interests involved. There is quite a bit of drama—which we know you’ll enjoy—, but there is also, in it, a core debate that will shape humanity for years to come. As we develop the world’s most powerful technology, who should be in charge of it and how should it come about? Is OpenAI in the right to privatize this knowledge or should we aspire for full openness in AI??

We do not know the answers to the above, but at least we have some initial facts that can help us find out, in a world of public drama, a small and guiding light.

Note: Although we do cover a great part of OpenAI’s story, we are, by no means, covering it all. Most notably, we will not cover Sam Altman’s ouster in any depth—to do so would make this essay too long to read. If you wish to learn more of that saga, we recommend the following essay on it: The Inside Story of Microsoft’s Partnership with OpenAI.

Act I. A need for change

Napa Valley; Palo Alto

(Elon Musk, Larry Page, Sam Altman)

Sometimes a conversation is enough to change the course of history. At least, to inspire certain ideas that will, in due course, shift humanity towards a given path. We owe the existence of Open AI to one such conversation.?

The setting was a birthday party in Napa Valley; the year, 2013; the subjects, a bunch of tech entrepreneurs. The party itself was in honor of now-billionaire Elon Musk who, back then, was better known for being a co-founder of PayPal and oddly interested in building commercial rockets. Amongst those invited were Luke Nosek, Reid Hoffman, and, crucial to the story, Google’s co-founder Larry Page.?

At some point, as many conversations go, the subject turned to speculation; more specifically, around the merits of AI—then only a theoretical possibility. Many were familiar with the work of a startup called DeepMind which promised to create tools that could, one day, rival the human race through sophisticated artificial intelligence models. Musk, for his part, found such a prospect to be a large threat to humankind, even positing the possibility humans could one day be replaced by robots. Page, however, saw the exact opposite, arguing in favor of a world where human consciousness wasn’t much more valuable than that created by machines.

The conversation—soon a heated debate—left Musk with a genuine fear of Page and a deeply held belief that Google would push for the creation of AI without the proper safeguards to protect humanity. All of this would be further intensified when Google acquired DeepMind the year after, positioning itself as the most important prospect to develop Generative AI, and giving Musk further reasons to fear the future of technology was now in the wrong hands. (Much of these accounts stem from Isacson’s seminal biography of Musk).

This is when OpenAI was born and we should keep this inception in mind as we go forward. The company was the product of Musk’s existential angst and desire to defeat Google in the world of AI.?

To achieve a better path to AI, everything had to be different to Google. The initial idea was to create a nonprofit company that would stand in direct contrast to private AI research. Its goal was to “advance digital intelligence in the way that is most likely to benefit humanity as a whole” for which it raised $135 million ($45 million from Elon and $90 from other donors).

Instead of keeping models behind closed doors, they would make them readily accessible for programmers to use and improve—hence, the name, OpenAI. Musk soon brought Sam Altman to become his co-chair to the foundation, a young Stanford dropout, former founder, and head of Y-Combinator. Not long after, he stole one of Google’s main engineers, Ilya Sutskever, by offering him a $1.9 million bonus.?

Slow and steady, the team to defeat Google was falling into place.

Act II: The original sin

OpenAI and Tesla HQ

(Elon Musk, Sam Altman)

For the next two years, OpenAI made the best efforts to enter the AI war. Yet even with $135 million in donations, it was fighting some of the largest tech companies around. It was clear that, just from a resource perspective, there was no way in which OpenAI could ever compete with Google. If it ever hoped to bring AI to life—and do so in a safe way—, it had to receive large amounts of capital that could likely only come from the private sector.

The debate was soon settled: OpenAI had to become a private company if it wished to lead in the quest for AI. The question that remained was how this should be done while preserving the company’s integrity.

Recent emails disclosed by OpenAI suggest that Musk himself first suggested the idea as early as 2017. In return, he would gain full leadership of the company and place it under Tesla for it to be adequately funded while it processed the millions of images Tesla gathered from its cars every day to create self-driving vehicles. It was around the time that Altman, already the chief executive for the non-profit, coined for himself what seems to be an adequate portrayal of Musk: “Elon desperately wants the world to be saved. But only if he can be the one to save it”

The board rejected the idea and, instead, created a complicated financial system that would allow it to maintain its nonprofit vision while entering the private world—crucially, keeping Musk away from controlling the company. Instead of abandoning the non-profit structure, OpenAI created a separate for-profit entity, OpenAI GP LLP. The non-profit—OpenAI, Inc—then became a shareholder of the for-profit company, owning some 51% of all stocks. The remaining 49% was given to investors, most notably Microsoft which put $13 bn to the new company’s goals. Quite crucially, despite these large investments, the company’s board remained fully controlled by the nonprofit, meaning that even Microsoft was exempt from board meetings in one of its largest investments.

This would become the founding sin of OpenAI and the cause of many of its troubles. In a quest for growth, the company abandoned its nonprofit nature and created a for-profit entity in 2019. To fully become profitable, it made its models proprietary—albeit letting many users reap their benefits for free—, thus breaking its original notion of AI being an open commodity for developers to explore. The company kept its nonprofit board as a guiding principle to ensure its tools would be well aligned with the interests of humanity, but it also entered the more complex world of capital incentives.?

Musk, for his part, left OpenAI’s board in defeat a year prior to the transition. The company cited a potential conflict with Tesla’s own AI research but, in reality, there was a sense of frustration that he wouldn’t lead the efforts towards the noble AI he had envisioned in Napa while debating Larry Page. With this, the war would open a new front for which, only now, are we witnessing the first skirmishes. At first, OpenAI was designed as the resistance to Google’s hegemony over AI; now, one of the company’s founders left in frustration—the same force that brought him to create it in the first place.

Act III: The lawsuit

San Francisco

(Elon Musk, Sam Altman)

We can speculate as to why it took so long for the next chapter of this story to unfold. Perhaps it was a matter of size. While Musk left the company in 2018, it wasn’t until 2020 that the company released its first model GPT-3, and, even then, it wasn’t until 2022 that it reached broad fame as it unveiled Chat GPT—its conversational AI tool. Perhaps it was a matter of benefit; recently Musk has devoted his time and energy to creating his own AI company: xAI. Regardless, it is true that the company is once again in the spotlight.

Musk has returned once more and this time he did so with weapons in hand. He is no longer sending emails or proposing a takeover. Now, he has filed a lawsuit claiming the company betrayed its original vision—his vision, perhaps, of a world with free access models to rival Google’s relentless march towards progress.

At its core, there are two main arguments for Musk’s lawsuit. On the one hand, it claims that OpenAI has betrayed its original vision by abandoning its practice of making its models publicly available—the company’s main revenue comes precisely from charging users for the use of its models. This has led Musk himself to claim he will drop the lawsuit if OpenAI changes its name to ClosedAI.?

On the other hand, quite interestingly, is a financial argument about OpenAI’s early funding—of which Musk contributed $45 million. By shifting from a nonprofit to a for-profit entity, it was able to live its first years without paying taxes on its funding. Why, then, would any company begin as an LLC if it can follow the OpenAI path instead?

As of writing this piece, we have little to no evidence on how the fight will actually evolve. OpenAI soon responded by publishing the emails we previously mentioned where Musk personally suggested the company should be made private to achieve its goals. Almost as if its own shift in principles could be justified by Musk’s past backing. Musk, as a response, announced his own company, xAI, will make its signature model, Grok, open source. A constant back and forth of moral signaling.

What follows? It is unclear. A court will soon have to rule on OpenAI’s future and settle the battle between its founders. For now, all we can do is sit back and enjoy the show—one of many the company has had to offer.