Microsoft wins the schism at OpenAI
Making sense of the recent developments at OpenAI
For some time, I have been wondering if or when OpenAI would join Microsoft, similar to how Google owns DeepMind. Now, it seems I have my answer, though the circumstances were unexpected.
Let's unpack the events that have unfolded over the last few days and how the schism at OpenAI is reshaping the AI landscape.
The timeline of events
According to Greg Brockman’s tweet, Sam Altman (who at that time was in the Middle East) received a text message from Ilya Sutskever, chief scientist at OpenAI, inviting him to join a call on Friday at noon, November 17th. During that call, Sutskever and OpenAI’s board informed Altman that he was being removed from his role as CEO of OpenAI.
About 15 minutes later, when the call with Altman was done, Sutskever asked Brockman to have a call. A couple of minutes later, Brockman learned what happened and that he had been removed from the board. However, he would retain his role as the president of OpenAI. Around the same time, OpenAI published a statement announcing the change in leadership. Mira Murati, who held the role of CTO at OpenAI, was appointed as an interim CEO until a permanent replacement for Altman had been found.
Up to this point, it seems only Sutskever, the board and Mira Murati knew about what was going to happen (Murati was apparently informed the night before, according to Brockman’s timeline of events).
Three hours later, Brockman announced that he was leaving OpenAI. Shortly after that, three OpenAI employees - Jakub Pachocki (GPT-4 lead and the director of research), Aleksander Madry (head of AI risk and the leader of OpenAI Preparedness team) and Szymon Sidor (open-source researcher at OpenAI) - joined Altman and Brockman in leaving OpenAI.
In the meantime, Microsoft issued a statement reassuring that Microsoft has a long-term agreement with OpenAI and that they are committed to this partnership. It’s fair to assume that Microsoft was not aware of the upcoming changes until they were made public.
That was Friday. What followed was shock and a barrage of speculations about what actually happened (I will examine that later in this post).
OpenAI investors pressured the company to bring back Altman but the negotiations have failed. On Sunday, OpenAI announced that Emmett Shear, the co-founder of Twitch, has been appointed as an interim CEO. On Monday, Satya Nadella revealed that Altman, Brockman and their team will be joining a new advanced AI research team at Microsoft. “We are going to build something new & it will be incredible”, tweeted Brockman while Altman just said that “the mission continues”.
Exposed fractures at OpenAI
Let’s start by examining why was Sam Altman fired from OpenAI. The reason the board gave for its decision was that the board “no longer has confidence in his ability to continue leading OpenAI” and that Altman was “not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities”. A leaked memo published by Axios adds more context. “We can say definitively that the board's decision was not made in response to malfeasance or anything related to our financial, business, safety, or security/privacy practices. This was a breakdown in communication between Sam and the board", the memo states.
So the next question then is what caused this “breakdown in communication between Sam and the board”? Everyone at OpenAI is aligned with the company’s mission - “to ensure that artificial general intelligence benefits all humanity”. The source of the schism lies in the answer to another question: what is the best and safest way to bring AGI?
One approach, championed by Altman, is through incremental releases. This part of OpenAI believes in gradual, iterative deployments of advanced AI to the public to let people start exploring what is possible with these new, powerful tools. That’s the part of OpenAI that released ChatGPT and turned OpenAI into a leader in the AI industry worth $29 billion.
The second approach, the one that the board has chosen to align with and was led by Ilya Sutskever, calls for more research and development to make sure AGI’s goals and values align with ours first before releasing. That’s the part of OpenAI that represents the original spirit and mission of the company.
As The Atlantic reports, this tension between “research vs product” started to brew a year ago. The overnight success of ChatGPT caught everyone at OpenAI by surprise. This suddenly changed the dynamics inside the company. OpenAI had a product and resources have been shifted away from the research teams to serve the traffic coming to ChatGPT. The company itself started to change from a research lab into a product-based company. Sutskever did not approve this change and the rift between him and Altman started to grow.
But why now? Why was Altman kicked out of OpenAI now and not when GPT-3 or GPT-4 were released?
It seems the trigger for the coup was the recent OpenAI Dev Day. Altman revealed the new GPTs - small, specialised versions of ChatGPT that anyone can create - and a marketplace for people to publish their GPTs. Altman said this is a step towards the introduction of agents - highly capable AI bots that can plan and execute complex actions on their own. That’s when Sutskever thought he had to take action.
OpenAI has an unusual corporate structure, stemming back to the founding of the company and its mission. OpenAI was founded in 2015 as a non-profit organisation with a clear goal - “to ensure that artificial general intelligence benefits all of humanity”. After being a non-profit company for a couple of years, in 2019, in order to attract more investments and talent, OpenAI went through a reorganisation. A new, capped-profit company named OpenAI LP has been created under the non-profit OpenAI, Inc. company. OpenAI LP is the company that was receiving the investments and is charging for OpenAI services, such as ChatGPT Plus and OpenAI API. Meanwhile, the non-profit OpenAI, Inc., with the board of directors, is ensuring that the company is following its original mission.
Ilya Sutskever convinced the board of directors that the direction in which OpenAI was heading was against the company’s values and the board ousted Sam Altman.
If you're enjoying the insights and perspectives shared in the Humanity Redefined newsletter, why not spread the word?
The aftermath
Altman and his team are now at Microsoft, free of any shackles they may have had while they were at OpenAI. California has no no-compete laws, meaning that Altman, Brockman and whoever joins them can take their knowledge and expertise from OpenAI and start building their own models to challenge what OpenAI, Google and others have to offer.
The recent events have eroded Microsoft’s trust in OpenAI. A little over two weeks ago, Nadella joined Altman on the stage at OpenAI Dev Day reaffirming Microsoft’s commitment to “build AGI together”. Just a couple of days ago, at the Microsoft Ignite conference, Nadella was presenting the AI-focused future of Microsoft with OpenAI playing a central role in those plans. Despite being OpenAI's largest investor and its primary partner in both technology and business, Microsoft was not informed about the upcoming changes in OpenAI's leadership. Apparently, Nadella was “livid” when he heard the news. Moreover, Microsoft, which invested over $10 billion into OpenAI, does not even have a seat on OpenAI's board. When the news broke, Microsoft's stock dropped by 2%, showing the extent to which Microsoft's future is intertwined with OpenAI.
But Satya Nadella has proven once again how good of a CEO he is. Microsoft has got OpenAI’s core team for free. Sam Altman, Silicon Valley’s golden boy and the personification of the AI revolution, is now working at Microsoft. Greg Brockman brings in heaps of technical experience, as well as others who have left and will leave OpenAI.
Altman’s new team will need some time to set up and get up to speed. In the meantime, Microsoft will be supporting OpenAI and using their models. But I expect that over time, resources will shift from OpenAI to the new team. We don’t know the details of the deal between OpenAI and Microsoft but as soon as the new team matches what OpenAI has to offer, I expect Microsoft to reduce its support for OpenAI, if not completely cut the ties.
And then there is the entire ecosystem of not only startups but also large companies that are building products and services on top of OpenAI models. These companies might be asking themselves the question of what to do next. They want to have stability which OpenAI, in the light of recent events, cannot deliver.
What we have witnessed is a massive reshuffling of the AI landscape and the dust hasn’t settled yet. One of the most important AI labs is imploding and we have a new AGI lab fully owned by Microsoft with access to enormous amounts of computing power. The probability of AGI arriving from Microsoft has now increased.
OpenAI will continue, but in what form remains to be seen. More people are likely to leave and many places, especially Microsoft, will welcome them with open arms. OpenAI will now have to reinvent itself and choose what it wants to be.
As I am writing this article, reports emerge that OpenAI employees call for the board to resign. Wojciech Zaremba, one of the co-founders of OpenAI, joins this call. Meanwhile, Sutskever regrets the chain of events he set in motion.
Thanks for reading. If you enjoyed this post, please click the ❤️ button or share it.
Humanity Redefined sheds light on the bleeding edge of technology and how advancements in AI, robotics, and biotech can usher in abundance, expand humanity's horizons, and redefine what it means to be human.
A big thank you to my paid subscribers, to my Patrons: whmr, Florian, dux, Eric, Preppikoma and Andrew, and to everyone who supports my work on Ko-Fi. Thank you for the support!