OpenAI: From Non-Profit to the Company That Never Fades from Headlines
This article dives into some history and legal constitution of OpenAI and how the company has evolved in the past few years and why Elon has been so mad at them for quite some time now.
AI enthusiast or hater, you can’t deny that last year belonged to OpenAI, the American company behind one of the most popular tools right now - ChatGPT. But ChatGPT is not the only thing OpenAI has been on the news for. The CEO firing and rehiring debacle, and now a lawsuit from the on and off richest person of the planet - Elon Musk. This article dives into some history and legal constitution of OpenAI and how the company has evolved in the past few years and why Elon has been so mad at them for quite some time now.
In their own words,
“OpenAI is an AI research and deployment company. Our mission is to ensure that artificial general intelligence benefits all of humanity.”
This looks all sweet and shiny, but what many don’t know is that OpenAI consists of 2 legal entities: OpenAI Inc. (a single member Delaware LLC controlled by OpenAI non-profit) and OpenAI LP (a capped for-profit organization)
So a non-profit is seeking profit now?
OpenAI was officially announced as a non-profit that would “freely collaborate” with researchers, institutions and companies to research and deploy new technologies. Their research and patents were proposed to be shared with the world. This initiative was the brainchild of some of the biggest names in industry - Sam Altman (the ex-president of startup incubator Y Combinator), Greg Brockman, Reid Hoffman, Jessica Livingston, Peter Thiel, Elon Musk (who later had to depart from the board due to conflict of interest to recruit similarly skilled talent for Tesla. What really led to this might be somewhat different though). Other than these big names, some organizations like Amazon Web Services (AWS), Infosys, and YC Research, in combination, committed donations worth USD 1B+ to the company in December 2015. The actual donations, however, amounted to just above 1/10th of that number up untill 2019.
Fast forward to 2019, the company realized they don’t have enough resources to compete with tech giants like Google and Facebook (now Meta). They admittedly needed more funding for the compute to be used for model training and to afford other big names in the AI industry like Iya Sutskever, chief scientist at OpenAI, who in 2016 was reportedly paid $1.9 million.
So OpenAI was restructured in 2019 to create a new subsidiary:
OpenAI LP (capped for profit, profits being capped at 100 times the investment. This means investors can only get 100 times their investment, any further profits would go back to OpenAI.), and
OpenAI Inc. which continues to ensure that OpenAI LP stays aligned with the foundation’s mission to utilize Artificial General Intelligence for the benefit of all. Majority of OpenAI Inc.'s board is legally barred from having financial stakes in OpenAI LP.
The for-profit subsidiary, unlike the non-profit, is under no obligation to publicly report salaries now and offers stock options to their employees, and now owns ChatGPT and other LLM models. This sparked criticism from industry leaders, including Elon Musk and Oren Etzioni from the nonprofit Allen Institute for AI, who challenged the idea that a "nonprofit can't compete." OpenAI LP now profits through the commercial licenses to their wide stream of AI models, ChatGPT being one of them.
Sam Altman, interestingly, doesn't own any equity in OpenAI. But this might not be as altruistic as it sounds. He has personally invested in a startup called Rain AI, that hopes to replicate the human brain for building new computer chips - Neuromorphic processing units, or NPUs, which could be roughly 100 times more powerful than a GPU. All of this makes even more sense when put together with the recent investment talks from Altman, for a whopping USD 7 trillion.
So where does Microsoft come into the picture?
With the announcement of a for profit subsidiary in 2019, came another announcement of an investment of $1B by Microsoft, led by Satya Nadela. This created an entangled partnership between the two, with OpenAI exclusively using Microsoft’s Azure Compute, collaborating on building a supercomputer and Microsoft getting a license to commercialize OpenAI’s advanced AI technology. Altman was quoted as saying “we’re working with Microsoft to build the supercomputing foundation on which we’ll build AGI.” This has and would be a valuable development for Microsoft for the future of the AI industry and model training in the coming decade. Microsoft again participated in investment rounds in 2021 and 2023.
This partnership grew further last year with Microsoft’s $10B investment in OpenAI, which has raised concerns of OpenAI becoming more controlled by Microsoft, which is the biggest shareholder in OpenAI LP (the capped for-profit organization). Elon Musk has alleged that Microsoft would be taking proxy control of OpenAI, an allegation that Satya Nadella was quick to refute previously. Microsoft announced integrations of the latest tech in their enterprise and consumer products like Github Copilot, Bing, Office over the last year, which took the world by a storm and brought Microsoft to a state that is now keeping Google on its toes, and that too on a slippery slope.
How would the lawsuit affect OpenAI?
“OpenAI, Inc. has been transformed into a closed-source de facto subsidiary of the largest technology company in the world: Microsoft,” Musk’s lawsuit claims.
Elon Musk's lawsuit against OpenAI and its CEO, Sam Altman, primarily focuses on a breach of contract and the alleged abandonment of the company's original mission to benefit humanity. Musk claims that OpenAI has shifted its focus to profit-making, particularly through its partnership with Microsoft, which has led to the release of AI models that are not open-source and are not available to the public as originally intended. Musk's lawsuit seeks to compel OpenAI to adhere to its founding agreement and possibly to prevent Microsoft from profiting from the company's AI technology. The lawsuit also alleges breach of fiduciary duty and unfair business practices. Elon had shedded millions in investment in what he was told would be used to build AI to benefit humanity (not Microsoft).
OpenAI moved their technology to closed-source (or should we say ClosedAI?) with the release of GPT-3, one of the predecessors of the models behind ChatGPT. Earlier models like GPT-1 and GPT-2 were completely open source. Some arguments against open source AI have been potential misuse by criminals.
OpenAI again made headlines by open-sourcing past emails with Elon, where the extent of openness of their research and technology was clearly communicated and Musk seemed to be in (terse) agreement.
Interestingly, Elon Musk also founded xAI with a goal to solve some existential questions like understanding the true nature of the universe, which, he believes, would ultimately have much bigger potential to benefit humanity (let’s see how that one turns out ;) )
Informative
From my viewpoint, this evolution underscores a fundamental challenge in the AI field: balancing the need for substantial funding to advance research with the commitment to open access and the ethical use of AI. The partnership with Microsoft, while beneficial for resource acquisition and technological advancement, raises valid concerns about the concentration of influence in AI development and the potential sidelining of the initial goal to democratize AI benefits.
From my viewpoint, this evolution underscores a fundamental challenge in the AI field: balancing the need for substantial funding to advance research with the commitment to open access and the ethical use of AI. The partnership with Microsoft, while beneficial for resource acquisition and technological advancement, raises valid concerns about the concentration of influence in AI development and the potential sidelining of the initial goal to democratize AI benefits.
- GPT4's comment after reading this blog.