On Friday local time, Altman was ousted by board members led by OpenAI co-founder and board director Ilya Sutskever.
But in a surprising twist Saturday, Sutskever told an employee meeting that company executives were working hard to bring Altman back.
Then, in an eventual reversal, Altman decided not to return to OpenAI after a weekend of negotiations with the board, executives and investors.
According to the latest media reports, people familiar with the matter said that Sutskever is increasingly worried that OpenAI’s technology may bring dangers, while Altman does not pay enough attention to this risk. In addition, “commercialization” and profit also seem to be one of the reasons for internal divisions.
Concerns about AI security risks intensify
Altman's ouster has drawn attention to long-standing risks in the artificial intelligence community. Some people believe that artificial intelligence is the biggest business opportunity in a generation, while others are worried about the dangers of developing too fast. The vote to oust Altman shows how an ideological movement dedicated to fearing the risks of artificial intelligence has become an inevitable part of tech culture.
Nearly a year since ChatGPT was released, artificial intelligence has captured public attention amid hopes it could be used for important work such as drug research or to help educate children. But some AI scientists and political leaders worry about the risks, such as jobs being eliminated by automation or autonomous warfare developing beyond human control.
OpenAI's board of directors did not give a specific reason for ousting Altman, but said in a blog post that the board believed that Mr. Altman had not communicated candidly with them. According to a message from the media, OpenAI employees were told on Saturday morning that Altman was fired not because of "malfeasance or anything related to our financial, business, security or security/privacy practices" but because of "a breakdown in communication between Altman and the board of directors."
In recent weeks, Jakub Pachocki, who helped oversee ChatGPT’s core technology, GPT-4, was promoted to the company’s director of research. Pachocki, who previously held a lower position than Sutskever, was promoted to the same position as Sutskever, according to two people familiar with the matter. Pachocki resigned from the company late Friday, two people familiar with the matter said.
It is worth mentioning that Mr. Sutskever and three other board members are left: Adam D’Angelo, CEO of the question and answer website Quora, Tasha McCauley, part-time senior management scientist at RAND, and Helen Toner, director of the Strategic and Fundamental Research Fund at Georgetown University’s Center for Security and Emerging Technologies.
McCauley and Toner are associated with the Rationalist and Effective Altruist movement, a group deeply concerned that artificial intelligence could one day destroy humanity. Today's artificial intelligence technology cannot destroy humanity, but this group believes that as the technology becomes more powerful, these dangers will arise.
Sutskever also increasingly agrees with these concepts. Speaking on a podcast on November 2, Mr Sutskever said:
It doesn’t seem surprising that we will have data centers with computers that are much smarter than people, but what will this artificial intelligence do? I have no idea.
Does commercialization deviate from the “original intention” of non-profit?
In addition to security, another possible conflict between Altman and the board of directors is "commercialization."
Sutskever, Altman and Brockman, who founded OpenAI together as a nonprofit in 2015 with help from Tesla CEO Elon Musk, said the lab would not be driven by commercial incentives, unlike Google and other companies.
In 2018, Altman transformed OpenAI into a for-profit company and negotiated a $1 billion investment from Microsoft. This massive funding is critical to building technologies like GPT-4, which was released earlier this year. Since its initial investment, Microsoft has poured an additional $12 billion into the company.
But the company is still governed by a non-profit board of directors, and investors like Microsoft do make profits from OpenAI, but their profits are capped. Any funds over the cap are funneled back to the nonprofit.
In this company structure with dual attributes of "profit" and "non-profit", Altman and Suzkovo represent two positions.
Seeing the power of GPT-4, Sutskever focused more on security, creating a new "Hyper-Alignment" team within the company that was designed to explore ways to ensure future versions of the technology can't cause harm.
Altman hopes that OpenAI can stay ahead of competitors, so he prefers to quickly commercialize generative AI products and deploy consumer businesses across the world.
In late September, Altman flew to the Middle East to meet with investors, according to two people familiar with the matter. He is seeking up to $1 billion in funding from Japanese investment conglomerate SoftBank for a new venture under OpenAI that plans to produce a hardware device for running artificial intelligence technology like ChatGPT.
In addition, OpenAI is also negotiating a "tender offer" financing that would allow employees to cash out their company shares. This transaction would bring OpenAI's market value to more than $80 billion, almost three times what it was six months ago.