The other side of the father of ChatGPT: lies, manipulation, infighting

The board can fire me.

I think this is very important.

Sam Altman said at an event in June 2023.

Altman has always been proud of the particularity of OpenAI's management structure – a non-profit organization supervising a for-profit organization, with the former responsible for researching safe AGI that is beneficial to human development, and the latter making money to raise funds to "support" research.

Because of this, to a certain extent, the former's board of directors is not responsible to investors, but to "all mankind." If necessary, the CEO can even be fired.

Until this actually happens.

After last month's "Palace Fight", Altman has used actions to prove that the possibility he mentioned at the beginning was just a mirage.

As the "Palace Fight" came to an end, more details of the incident began to be revealed, and another side of Altman gradually emerged:

Four people who have worked with Altman over the years also say he can be a cunning man — sometimes deliberately misleading and deceptive.

"Times" admitted that the image of this "CEO of the Year" is becoming complicated. As more information emerges, another side of Altman is becoming difficult to ignore.

Infighting maker

When former OpenAI board member Helen Toner published a paper on AI safety, she might not have expected that it would reveal a worrying side of Altman.

In October of this year, Toner, who also serves as the director of strategy and research funding for the Center for Security and Emerging Technologies at Georgetown University, co-authored a paper on AI security. In the AI ​​industry, Toner is also regarded as someone who dares to speak out.

This paper mentioned that the release of ChatGPT by OpenAI has made technology giants nervous, prompting these companies to accelerate the launch of AI products to keep up with the trend. The paper also mentions that OpenAI’s competitor, Anthroic, is waiting to release its chatbot to avoid “the fireworks that fuel AI hype.”

▲ Helen Toner

Altman believes that Toner's article will damage OpenAI – if company members say such things, it will make OpenAI, which is already attracting regulatory attention, even more ugly, and may lead regulators to speculate that there are deeper problems within the company.

In response, Toner sent a group email to other members of the board of directors to remind them of the possible impact of the paper and respond to relevant questions.

Altman, meanwhile, began talking privately to other board members.

Altman and some board members said director Tasha McCauley made it clear "it was clear Helen had to leave." McCauley clarified that he never said such a thing .

Later, board members began to talk and learned that such a situation existed.

He has a way of preventing others from seeing the whole picture.

One person familiar with the matter said such behavior left people wondering what other people were thinking.

This may be considered a common business tactic, but this person familiar with the matter believes that Altman's behavior has "crossed the line" and affected the board's effective supervision of his behavior.

▲ In addition to Altman and Greg Brockman, former OpenAI board members also include Ilya Sutskever, Tasha McCauley, Adam D'Angelo and Helen Toner

"He's trying to pit people against each other by lying about what they think," said another person familiar with the situation. "This has been going on for years."

People familiar with Altman's perspective said Altman later admitted that he was "really clumsy in trying to remove board members" but said he was not trying to manipulate the board.

It is worth noting that multiple reports pointed out that this incident was not the direct reason for the board of directors to decide to fire Altman, but it was indeed representative in clarifying Altman's behavior pattern and sounded an alarm for the board of directors.

The voice that really lets the board of directors decide on action comes from OpenAI's management team.

Born from idealism, is OpenAI also a “toxic workplace”?

This fall, the board of directors encountered a small group of OpenAI managers "coming to their door."

The group, which included key members of the company as well as leaders responsible for running large teams, told the board that Altman had been "psychologically abusive" to members of the company.

They pointed out that Altman deliberately "stoked fights" among employees, leading to unhealthy competition within the company.

They believed that Altman's behavior was quite "toxic" when interacting with him, and they were also worried about Altman's retaliatory behavior:

One told the board that when an employee provided critical feedback to Altman, Altman would discredit that employee on that person's team.

Although this all feels quite "familiar" to the board of directors – thanks to Toner's paper – the board still needs to consider the extremely high level of loyalty Altman has cultivated in the company's content and the possible long-term impact of the incident.

At the same time, this small group of managers who filed complaints are also very important members of OpenAI.

Obviously, the board of directors ultimately believed that Altman's behavior was more harmful, at least at the level of the board's exercise of oversight over the development of secure AGI.

They decided to take a risk and fire their star CEO, and they had to do it quickly.

One person familiar with the board's decision said:

Once (Altman) gets a slight sense that there's a chance it's going to happen, no matter how small the chance is, he's going to go all out.

The board of directors who were forced to give in, and their "final stubbornness"

On November 7, the OpenAI board of directors suddenly announced the dismissal of Altman. The official announcement stated:

Mr. Altman's departure is the result of a thoughtful review by the Board of Directors.

The board of directors believed that he had not been candid in his communications with the board of directors and had hindered the board of directors from performing its duties.

The board no longer has confidence in him continuing to lead OpenAI.

Due to legal and confidentiality reasons, the board of directors has not disclosed the specific circumstances behind the "failure to remain candid."

This vacancy also gives Altman room to tell his own story.

As soon as the dismissal was announced, Altman's computer was locked.

He started looking for his investors and mentors and told everyone that he was going to start a new AI company.

One of Altman's first calls was Ron Conway, founder of venture capital fund SVAngel. In Conway's view, the board fired Altman for "nitpicking, which is far from warranting dismissal":

It would be reckless and irresponsible for a board to fire a founder for emotional reasons.

Within OpenAI, the team was also dissatisfied with the lack of explanation due to legal reasons.

Jason Kwon, OpenAI’s chief strategy officer, accused the board of directors of violating its fiduciary duties:

It is impossible for you to watch your company collapse.

Toner could only reply: "The destruction of the company may also be in line with the mission of the board of directors."

Not long after, after hearing the threat from employees that "if Altman doesn't come back, we will resign," the board of directors invited Altman to the company to discuss the next steps.

There are two versions of the story here.

In Altman’s explanation, the board of directors invited him back to persuade him to return to OpenAI:

I went through a lot of emotions.

At first I was very resistant, but soon I shifted to a sense of responsibility and wanting to protect something that I hold dear.

But people close to the board of directors said that the offer made by the board at that time was actually for Altman to come back and discuss how to stabilize the company.

During the discussion, the two parties discussed when and what information should be disclosed to the media.

Altman also once proposed a public apology for the "Toner incident", but the board of directors had doubts and worried that once this was said, the whole thing would be more like the board of directors' motivation to retaliate against Altman.

Behind the seemingly loving resignation letter signed by 700 people, although there is loyalty, there is also the factor of money.

According to the original plan, OpenAI employees would be able to sell their shares in the company valued at US$86 billion and quickly realize cash shortly after Altman's resignation.

But if Altman's return is not negotiated, OpenAI looks like it will collapse, and the cash it looks like will evaporate.

In the workplace application Blind, which requires work email authentication, an OpenAI employee revealed that when the joint letter appeared, he had to sign it under tremendous pressure from colleagues.

Some employees disagreed with this statement, saying, "Half the company signed at 2-3 o'clock in the morning. This is not something peer pressure can do."

There was another important signal in the joint letter at that time, that is, the name of OpenAI chief scientist Ilya Sutskever also appeared on the request.

Until then, Sutskever had been supportive of the board's decision to fire Altman.

The night before the joint letter appeared, Sutskever met Anna Brockma, the wife of OpenAI president Greg Brockman, at the OpenAI offices. When Altman was fired, Greg immediately followed suit.

In 2019, Sutskever hosted the wedding of Anna and Greg in the OpenAI office. At that time, the robot was responsible for helping the groom pick up the wedding ring.

And on that night in November this year, Anna walked up to Sutskever crying, took his arm, and persuaded him to reconsider his decision to fire Altman. According to people familiar with the matter, Sutskever's face was expressionless at the time.

We have no way of knowing what kind of struggles Sutskever went through next. We only know that the next day, his name appeared in the joint letter.

When Altman returned, Sutskever posted on X:

I've learned a lot of lessons over the past month, one of which is that "if morale doesn't improve, the whipping will continue" happens more often than it should.

"The beatings will continue until morale improves." The irony is a vicious circle: low morale causes punishment, and punishment further worsens the low morale.

The post was deleted the day after it was posted, and Sutskever’s future at OpenAI is full of uncertainty.

In the end, although Altman's return was decided, the board of directors also obtained some concessions from the other party.

The newly formed independent board will oversee Altman's conduct and investigate the previous board's decision to fire him. Neither Altman nor Greg will get a seat on the board again.

While Toner and McCauley have both resigned from the board, original director Adam D'Angelo will retain his position.

▲Adam D'Angelo

One insider said D'Angelo's retention was fought hard by the departing directors.

They want to leave someone in the company who understands what happened here.

Reference sources:

TIME: 2023 CEO OF THE YEAR: Sam Altman

New York Times: Inside OpenAI's Crisis Over the Future of Artificial Intelligence

Washington Post: OpenAI leaders warned of abusive behavior before Sam Altman's ouster

Bloomberg: OpenAI's Altman Ouster Was Result of Drawn-Out Tensions

WSJ: The OpenAI Board Member Who Clashed With Sam Altman Shares Her Side

New Yorker: The Inside Story of Microsoft's Partnership with OpenAI

# Welcome to follow the official WeChat public account of aifaner: aifaner (WeChat ID: ifanr). More exciting content will be provided to you as soon as possible.

Ai Faner | Original link · View comments · Sina Weibo