What does the reversal of OpenAI’s power struggle reveal?

Unveiling the Truth Behind OpenAI's Power Struggle Reversal

Author: Wu 00000 Kong; Source: Qin Shuo’s Friends Circle

On November 17th, after being swiftly removed by the OpenAI board of directors, the “Father of ChatGPT,” Ultraman, experienced three reversals in just one week, leaving all the onlookers dizzy. Now that the dust has settled, let’s take a look at what this power struggle story teaches us.

First, let’s review the bewildering process.

1. Twist and Turn

On Friday, November 17th, the OpenAI board of directors announced the dismissal of CEO Ultraman. Microsoft, the largest investor, immediately stated that this would not affect their cooperation with OpenAI and that Microsoft would continue to support them.

• Robots, terminals, and LLM – The encrypted user experience is gradually becoming inward.
• LianGuai Morning News | CZ unable to return to UAE before next year’s verdict, Google Cloud confirms participation in EigenLayer testnet
• Shiba Inu: Is it Time to Go Fetch Some SHIB?

A few hours after Ultraman’s dismissal, Chairman and President Greg Brockman announced his resignation. Following suit, three top researchers, Jakob LianGuaichocki, Aleksander Madry, and Szymon Sidor, also resigned in protest, stating that they would stand with Ultraman and Brockman. More key personnel expressed their intention to take similar actions.

The investors became anxious, including Microsoft. On Saturday, November 18th, Microsoft, along with major investors such as Tiger Fund and Thrive Fund, stepped in to mediate. Ultraman hesitated about returning and proposed that the prerequisite for his return was the resignation of board members. A large number of employees also demanded the resignation of the board and set a deadline. If the board did not resign by 5 pm, they would resign. However, the board evaded a clear answer, and when 5 pm passed, they did not resign. The employees extended the deadline to Sunday afternoon.

On Saturday night, Ultraman sent a message on X platform saying, “I love the OpenAI team too much.” Many OpenAI employees replied with heart-shaped emoticons, indicating their intention to follow Ultraman. It was like a public opinion poll within OpenAI.

On Sunday, November 19th, Ultraman sent a picture message on X platform, indicating that he was once again at the company for final negotiations with the board. Microsoft CEO Satya Nadella, who presided over the negotiations, clearly hoped that both parties would reach an agreement as soon as possible, preferably before the stock market opens on Monday.

However, things didn’t go as planned. In fact, while investors were actively searching for new board members in preparation for Ultraman’s return, the board quietly searched for a CEO to replace interim CEO Mira Murati. Murati had previously stated that she would bring Ultraman and Brockman back to assume some role.

Later on Sunday, the board announced that Twitch co-founder Emmett Shear would become the new CEO of OpenAI. Within three days, the company changed its CEO three times, an unprecedented situation.

A few hours later, on Monday (November 20th) morning, Microsoft CEO Nadella shared “very exciting news” with everyone. Ultraman and Brockman will join Microsoft’s newly established advanced AI research team, with Ultraman being the CEO. Nadella said, “We will quickly take action and provide them with all the resources they need for success.”

Ultraman replied, “The mission continues.”

Nadella responded to Ultraman, citing successful examples of internal innovation within Microsoft, reassuring Ultraman, “Over the years, we have learned a lot about how to provide founders and innovators with space within Microsoft to establish their own identity and culture, including GitHub, Mojang Studios, and LinkedIn. I look forward to collaborating with you in the same way.”

On Monday (November 20th) morning, more than 700 OpenAI employees sent a joint letter to the board of directors, requesting to bring back Ultraman and Brockman. If the board does not do so, they will collectively resign and join Microsoft. Among the signatories is Sam Altman, who served as interim CEO for one day, which is understandable, but what’s strange is that Chief Scientist Ilya Sutskever is also among them.

The letter also mentioned, “Microsoft has assured us that if we are willing to join, their new subsidiary will have enough positions to welcome all OpenAI employees.”

In the afternoon of Monday (November 20th), the Vice President of Global Affairs at OpenAI sent a message to all employees, stating that Ultraman, the remaining three directors, and the new CEO Hill are still discussing, and they cannot reach a conclusion tonight. The plan is to continue the discussion on Tuesday morning.

Microsoft CEO Nadella is a bit confused, but regardless, he supports Ultraman whether he comes to Microsoft or returns to OpenAI. The Microsoft CTO also stated that the compensation for OpenAI employees joining Microsoft will be as high as or higher than before.

On Wednesday (November 22nd) morning, OpenAI announced that a preliminary agreement has been reached. Ultraman will return to OpenAI as the CEO, and the new board will consist of Bret Taylor, Larry Summers, and Adam D’Angelo. D’Angelo is one of the original three independent directors. The task of this board is to identify, review, and appoint a new board consisting of up to nine people. Microsoft has applied for one seat, and Ultraman has also applied for one seat. Everyone expressed their intention to invite an independent third party to thoroughly investigate this incident.

Afterwards, Microsoft CEO Nadella posted his congratulations and hoped that there would be no more surprising things happening. Ultraman expressed his excitement to return to OpenAI and continue developing the “deep partnership” with Microsoft, stating that “Microsoft is the biggest financial supporter of ChatGPT.”

Another major investor in OpenAI, the Thrive Fund, stated that Ultraman’s return is the best outcome for the company, its employees, other companies developing products based on their technology, and the world as a whole.

The story ends here, and so far, there have been no new twists. It seems like the dust has settled. Now let’s analyze the different roles.

2. Microsoft, to the rescue or taking advantage of the situation?

After the incident, Microsoft stated that they were not aware beforehand and only found out after the board announced the decision. But right after the incident, they immediately withdrew their support for OpenAI without questioning the board’s decision. It was only when the president, three top researchers, and a large number of key employees expressed their intention to leave OpenAI alongside Ultraman, that Microsoft panicked and rushed to mediate between the board and Ultraman, along with other investors.

However, the result of the mediation was that the board quickly hired CEO Hill, while Ultraman announced joining Microsoft. Microsoft CEO Nadella wasted no time in announcing this “great news” in the early morning, trying to turn the situation into a done deal, fearing any unexpected changes. He also stated that Ultraman would be given full freedom and support, reassuring him and avoiding any change of heart.

This raises the question: is Microsoft here to put out the fire or take advantage of the situation? The next day, almost all OpenAI employees expressed their willingness to join Microsoft if the board didn’t bring Ultraman back. Microsoft’s CTO even stated that there was plenty of room and the same or even higher treatment for all OpenAI employees.

If all OpenAI employees indeed join Microsoft alongside Ultraman, it means that Microsoft obtained a company valued at 90 billion dollars for a very low price.

If the story were to end like this, it would be quite ridiculous. Over the years, Ultraman and Microsoft have been in constant struggle. On the one hand, Ultraman sought funding from Microsoft for development, while on the other hand, they strived to remain independent and avoid being “owned,” ensuring that the most advanced AI technology would not be monopolized by a single company. But now, they have been cheaply swallowed up. Isn’t that absurd? Fortunately, Ultraman doesn’t seem that foolish and has returned to the negotiating table, reaching an agreement with the board and returning to OpenAI.

Following this, Microsoft CEO Nadella stated that Microsoft supports Ultraman regardless of his choice.

3. Chief Scientist Sutskever, changing his stance

After the news of Ultraman’s dismissal came out, the general consensus was that Chief Scientist Sutskever was behind this “coup.” However, the next day, Sutskever’s name appeared on the joint letter from employees demanding the board to bring back Ultraman. He also posted, saying, “I regret participating in the actions of the board. I never intended to harm OpenAI. I love everything we have created together and will do my best to reunite the company.”

Obviously, Sutskever underestimated the influence and appeal of Ultraman. He wanted to drive Ultraman out, but what he didn’t expect was that this would lead to the disintegration of OpenAI, which he didn’t want to see, so he changed his stance.

In addition, there are reports that Brockman’s wife also came forward to request that Sutskever think twice. Sutskever is the best man at the wedding of the Brockman couple, and they have a close relationship.

In 2015, Sutskever resigned from Google, and Ultraman, Musk, and others founded OpenAI around him, establishing it as a non-profit organization to prevent Google or any other company from monopolizing artificial intelligence technology and ensuring it benefits all of humanity. If Ultraman now joins Microsoft and makes them the dominant power, then his years of effort would lose meaning.

Last year, after ChatGPT became popular, Ultraman, as the CEO of OpenAI, was hailed as the “father” of ChatGPT, but from a technical standpoint, Sutskever is the true soul behind it.

Sutskever (right)

Sutskever was born in Russia in 1985 and moved to Israel with his family when he was 5 years old, and later moved to Canada with his family. He studied under Professor Geoffrey Hinton at the University of Toronto for his Ph.D.

Hinton, born in the UK in 1947, obtained his Ph.D. in artificial intelligence from the University of Edinburgh in 1978, and then went to the United States for teaching. Dissatisfied with the political ideas of the Reagan era, he moved to Canada. He is the recipient of the 2018 Turing Award and is a Fellow of the Royal Society of Britain, a Fellow of the Royal Society of Canada, and a Foreign Fellow of the National Academy of Sciences in the United States.

Hinton has been studying in the field of artificial intelligence for many years and chose the neural network’s technical route. However, this route was not well regarded at the time because computer computing power was lagging behind, and progress was slow. Hinton almost gave up. He recruited some Ph.D. students, hoping that they would persist in the direction of neural networks, generation after generation, and perhaps there would be a breakthrough 100 years later.

Unexpectedly, 10 years later, computer computing power greatly increased, and the neural network technology route became a reality. At this time, he recruited Sutskever and another Ph.D. student, Alex Krizhevsky. In 2012, the mentor and two students completed the development of a new type of neural network, and they won the championship at the ImageNet Artificial Intelligence Contest on September 30th that year. Many companies approached them. The mentor and two students formed a company and accepted bids.

At the time, there were four companies participating in the bidding: Baidu, Google, Microsoft, and the newly established UK company, DeepMind.

The three mentors felt that Google was more ethical, as their company charter clearly stated “do no evil,” so they accepted Google’s offer. Google then created a new division called Google Brain for artificial intelligence, which was led by the three mentors. Little did they know, Google would later collaborate with the US Department of Defense on an artificial intelligence research project called Project Maven, which was exposed in 2018. Employees protested and demanded an explanation from management, citing the “do no evil” clause in the company charter. Surprisingly, management responded by removing the “do no evil” clause entirely.

In January 2014, Google also acquired DeepMind, with the three mentors leading the way.

In 2014, DeepMind began developing a Go software called AlphaGo, and in 2016, AlphaGo defeated the world champion Go player, Lee Sedol. In 2019, AlphaStar, specifically designed to play Starcraft, achieved a breakthrough in the field of collaborative artificial intelligence by defeating professional human players with a 10:1 win ratio. In 2018, AlphaFold disrupted the field of protein folding structure prediction.

One of Susswein’s senior colleagues, Kritchenski, retired from Google in 2017, completely leaving the field of artificial intelligence.

Professor Sinton resigned from Google in 2023 and warned the world that autonomous artificial intelligence systems could pose a serious threat to humanity in the long run. Reports indicated that Sinton regretted some of his research work.

Susswein shared similar views with his mentor, believing that General Artificial Intelligence (AGI) would inevitably surpass human intelligence soon. Thus, preparations should be made from now onwards to face AI that exceeds human wisdom. The existing methods of controlling AGI are not suitable for AI smarter than humans.

In July 2023, Susswein established a small department within OpenAI called “Superalignment” with the aim of developing a set of fault-tolerant procedures to control AGI technology. The plan was to allocate one-fifth of OpenAI’s computing resources to this department and solve this problem within four years. AGI technology would only be made available to humans once this control technology was developed.

Ultraman and investors clearly couldn’t wait that long. They wanted to commercialize more aggressively, as did numerous startups within the OpenAI ecosystem who were eager to quickly grow into unicorns. Their vision contradicted Susswein’s.

Susswein likely initiated a coup and fired Ultraman because of this fundamental difference in approach, with the specific trigger possibly being Ultraman’s speech at the APEC summit. The board of directors announced Ultraman’s dismissal, stating, “The board believes that Mr. Ultraman has not been fully transparent in his communications with the board, which has hindered the board’s ability to fulfill its duties.” No further explanations were provided regarding his lack of transparency. The exact reason for Ultraman’s dismissal is currently unknown, and both parties have stated that they will investigate the matter through a third party.

4. The Invincible Independent Directors, in the Name of All Humanity

Sutsweik was supposed to be the mastermind behind the coup, persuading the three independent directors to act together. But he quickly turned, signing the joint letter and demanding the board bring back Ultraman. Strangely, after the mastermind’s change of heart, the three independent directors did not follow suit, but insisted on “resisting,” even quickly bringing in a new CEO, further worsening the situation.

The independent directors are external individuals who do not participate in the management of the company, and they do not hold any shares of OpenAI. In fact, the three executive directors involved in the management, Ultraman, Brockman, and Sutsweik, also do not hold any shares of OpenAI.

This is a very peculiar arrangement. OpenAI Lab, registered in Delaware in 2015, is a non-profit organization that completely controls OpenAI Global, registered in 2019. Investors such as Microsoft hold shares in this company, but do not have seats on the board. They only provide funding, are not allowed to speak, and have no decision-making power. The six directors who have the power to speak and make decisions also have no shares.

This does not quite adhere to the principles of modern corporate governance. The basic principle of modern corporate governance, in plain terms, is that “money talks.” In other words, whoever provides the money makes the decisions. This is the most efficient arrangement. The underlying logic is the relationship between property rights and efficiency.

In simple terms, when you use your own money to do your own things, you do your best. When you use other people’s money to do other people’s things, it’s hard to say. When you use other people’s money to do your own things, it’s either illegal or unethical. When you use your own money to do other people’s things, that’s charity.

The executive directors of OpenAI are quite unique. Chief Scientist Sutsweik is an idealistic scientist who is not short of money and not interested in money. CEO Ultraman is a serial entrepreneur and investor who is not short of money and not interested in money. President Brockman is also a serial entrepreneur who is not short of money and not interested in money. They all uphold the purpose of OpenAI, to use general artificial intelligence (GAI) for the benefit of all humanity, rather than making the investors wealthy.

They believe that investors are tied up by capital, have short-sighted vision, and are focused on immediate gains, so they cannot be allowed to make decisions. Therefore, they invited three independent directors who have no investment relationship with OpenAI.

Adam D’Angelo, the CEO of the American version of Quora; Tasha McCauley, a female entrepreneur, former CEO of GeoSim, and the wife of Hollywood star Joseph; Helen Toner, Director of Strategy and Foundations Research at Georgetown University’s Center for Security and Emerging Technology.

These three individuals have no investment in OpenAI, but does that mean they have no conflicts of interest with OpenAI? It seems not necessarily.

Adam D’Angelo is hailed as one of the smartest CEOs in Silicon Valley, but his own company is also developing an artificial intelligence model called Poe. Recently, Ultraman launched GPTs at the OpenAI developer conference, which almost certainly spelled the end of Poe’s development.

Insiders say this has frustrated Adam greatly. He founded and operated Quora for 14 years, and former employees have revealed that Adam would suddenly dismiss executives without considering the impact on employees, and would not provide an explanation for the dismissals. One time, the employees at Quora were so unhappy with a dismissal that dozens of them went on strike the next day.

As for another board member, Helen Toner, she was previously criticized by Ultraman for her paper criticizing OpenAI’s efforts in artificial intelligence safety and praising their competitor, Anthropic. She and Ultraman had heated debates. Ultraman even discussed with executives the possibility of removing her from the board, but before he could take action, he himself was fired.

It seems that Chief Scientist, Suotsuike, was used as a scapegoat.

A few hours after Ultraman was dismissed, Jason Kwon, the Chief Strategy Officer of OpenAI, had a conflict with the remaining board members in a video call. Kwon condemned the board’s removal of Ultraman as endangering the company’s future and violating the responsibilities of board members. In response, Toner said, “The mission of the board is to ensure the creation of artificial intelligence that benefits all of humanity, even if it means the dissolution of the company, it still aligns with the mission of the board.”

It’s estimated that investors like Microsoft, upon hearing this statement, would be astonished. Just a few seemingly insignificant individuals with their own agenda could make their billions of dollars vanish in the name of “benefiting all of humanity.” With such a governance system, who would dare to play with you?

No wonder the outcome of the negotiations was the eventual formation of a nine-member board of directors, with Microsoft insisting on obtaining a board seat and Ultraman also insisting on obtaining a seat.

In May of this year, Ultraman testified at Capitol Hill, and a senator asked him, “How much money can you get from OpenAI?” Ultraman replied, “I don’t have any shares in OpenAI, my salary is only enough to cover social security.” The senator reminded him, “Then you might have some trouble, you’d better consult a lawyer.” And trouble promptly followed.

It can be seen that no matter how high-tech a company is, no matter how pure its purpose is, and no matter how noble its founder is, without considering human nature and the relationship between property rights and efficiency, the design of governance systems will ultimately result in significant losses.

Let’s hope that Ultraman learns from this experience.

We will continue to update Blocking; if you have any questions or suggestions, please contact us!