IT'S OVER! Sam Altman BACK as CEO of OpenAI

0:00Today on the AI Breakdown, it is finally over. Sam Altman is officially back as the CEO of OpenAI. the start of a new board has come into place. Here's everything you need to know. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown.net for more information about our Discord channel, our newsletter, and our YouTube.

0:24Welcome back to the AI Breakdown. Well, the folks at OpenAI have heard our pleas, and of course the pleas of the team there, which were of course for a quiet Thanksgiving. And so yes, Sam Altman is coming back to OpenAI to be their CEO. Now, that's not to say that there aren't some changes. The board is getting a shakeup and the board will not include Sam Altman and Greg Brockman. What's more, the board that fired him is not being replaced entirely. There's one key member who will continue to be a part of it. Now, we're going to get into how this all resolves. But first, let's take a look at where we were heading into Tuesday night when reporting from the New York Times and the Wall Street Journal has started to hone in on the conflict between Sam Altman and board member Helen Toner as perhaps the genesis of the entire controversy.

1:10For those playing along at home, there have been a never-ending series of villains on the OpenAI board side. Initially, it seemed like it was Ilya, the co-founder and chief scientist, who had had issues with Sam of late, and who was in fact the one to deliver the news to Sam that he was being fired. Over the weekend, the popular narrative was that Ilya had a fundamental disagreement with Sam about safety, that he had seen something that spooked him, and that that is what was the cause of all of this controversy. However, then it started to seem like maybe Adam D 'Angelo, the CEO of Cora, someone who has this seeming pretty significant conflict of interest given that he runs a competing product in Cora's Po, and the fact that he did not take too kindly to Mira Mirati, the former CTO, who had been elevated to the interim CEO position trying to plot to bring Sam and Greg back.

2:00Reporting started to suggest that perhaps Adam was the lead representative of the board and was the one who had reached out and hired Emmett Shear. But now a set of reports from The Information, the New York Times and the Wall Street Journal, have honed in on Helen Toner and her issues with Sam as the biggest driver of this insane situation we find ourselves in. The New York Times published last night, before Altman's ouster, OpenAI board was divided and feuding. Sam Altman attacked a member over a research paper that discussed the company, while directors disagreed about who should fill board vacancies open for months.

2:31Now, this piece frames the whole thing as instigated by Sam, who, quote, made a move to push out one of the board members because he thought a research paper she had co-written was critical of the company. So what was this paper? Well, let's just read the excerpt from the New York Times. A few weeks before Mr. Altman Ouster, he met with Ms. Toner to discuss a paper she had recently co-written for Georgetown University's Center for Security and emerging technology. Mr. Altman complained that the research paper seemed to criticize OpenAI's efforts to keep its AI technology safe while praising the approach taken by Anthropic, according to an email that Mr.

3:03Altman wrote to colleagues and that was viewed by the New York Times. In the email, Mr. Altman said that he had reprimanded Ms. Toner for the paper and that it was dangerous to the company, particularly at a time, he added, when the Federal Trade Commission was investigating OpenAI over the data used to build its technology. Ms. Toner defended it as an academic paper that analyzed the challenges that the public faces when trying to understand the intentions of the countries and companies developing AI. But Mr. Altman disagreed. He wrote in the email, I did not feel we're on the same page on the damage of all this.

3:30Any amount of criticism from a board member carries a lot of weight. Now, this is an excerpt from that paper, and it is incredibly laudatory towards Anthropic. One section reads, a different approach to signaling in the private sector comes from Anthropic, one of OpenAI's primary competitors. Anthropic's desire to be perceived as a company that values safety shines through across its communications, beginning from its tagline, an AI safety and research company. A careful look at the company's decision-making reveals that this commitment goes beyond words. A March 2023 strategy document published on Anthropik's website revealed that the release of Anthropik's chatbot Claude, a competitor to ChatGPT, had been deliberately delayed in order to avoid, quote, advancing the rate of AI capabilities progress.

4:09Basically, this whole page talks about how ChatGPT and OpenAI were responsible for starting an arms race, and Anthropik and Claude were just overall much better. Now, a Wall Street Journal piece that came out about this last night also pointed out Toner's ties to the effective altruism movement and why that was a concern internally. She had previously worked with Open Philanthropy, which is an organization that has in many cases been associated with EA. And of course, Anthropic is also an EA-aligned organization, which is why Sam Bankman-Fried was, before he went to jail, its lead donor. From the WSJ, OpenAI leadership and employees were growing increasingly concerned about being painted in the press as, quote, a bunch of effective altruists, as one of them put it.

4:48Two days before Altman's ouster, they were discussing these concerns on a Slack channel, which included Ilya. One senior executive wrote that the company needed to up-level its independence, meaning create more distance between itself and the EA movement. Now, if you were hoping that this reportage might cause a little bit of a shake-up in who believed who and who was inclined to trust who, think again. Instead, it has acted simply as a Rorschach test for people's priors, as witnessed by basically infinite comments on Twitter slash x. Gary Marcus, whose position this whole time has been roughly that he trusts Sam less than he trusts the other members of the board, writes, scoop from the information and New York Times that is not a good look for Sam if true.

5:25A board member, Toner, had done safety research, which OpenAI was set up to do. Her research pointed to a rival approach. Altman argued with her and then, story goes, tried to push her off the board. Now, many people responded to Gary with some version of, wait, but she was a board member talking up the most direct competitor. AJC writes, Is it common for a board member to publish negative or any reporting about their employer and company? Feels like a conflict or poor judgment at best. Jennifer Camille Lee writes, this does not make Sam look bad. It makes him look like he was protecting the company.

5:56Now on the other end of the spectrum is Chris at NoClarified who writes, this is Helen crapping on OpenAI's rollout of ChatGPT. While not perfect, apparently they spent several months working on safety to get it out. They were just much faster and more well capitalized. Why is Ms. Toner crapping on the product that she's on the board for? This would be unthinkable and any other company. Somewhere in the middle is Rob Benzinger, who observes, Seems like the first news article with leaks from the board and possibly the first to represent something like their perspective. I have to say, if Sam was trying to keep board members from saying anything negative about OpenAI's safety practices in public, I think this is really bad behavior on Sam's part.

6:30Sam acknowledges that the stuff OpenAI is trying to build toward has a strong chance of killing literally everyone on earth. This is not a game. These are not normal tech companies. If no one with privileged information about these companies is allowed to say anything public about the relative merits of DeepMind's approach, OpenAI's approach, Anthropik's approach, etc., then that severely limits humanity's ability to use any tools at all to improve our approaches and converge on sane practices. If you disagree with Toner's analysis, you should argue against it, not try to suppress it in order to protect your company's reputation and bottom line.

6:58This seems totally obvious to me. Though again, maybe the New York Times article here isn't fairly summarizing what happened. The article was co-written by Cade Metz, who often gets lots of basic facts wrong in his articles. Indeed, I can spot some obvious errors in this one already. I do have to say I remain confused about why the board has said so little about their reasoning. I don't know what the real story is here, but if the real story is, EAs trust Dario to safely steward AI but don't trust Sam, then even if they're right to think Sam is too reckless for this role, it's important that they be open and public about their reasoning.

7:25Now, when someone asked Rob what the obvious errors were, he responded, The first one that jumped out at me was that the reporters completely failed to understand what's maybe the most basic fact in this story. When these folks say they're worried about AI risk, what sort of risk do they have in mind? The New York Times wrote, The Rift also showed how building new AI systems is testing whether business people who want to make money from artificial intelligence can work in sync with researchers who worry that what they're building could eventually eliminate jobs or become a threat to humanity if things like autonomous weapons grow out of control.

7:54Rob continues, When Sam says that AI may be lights out for humanity, he's not talking about technological unemployment, and he's not talking about conventional weapons like military drones getting too strong. This is either a hilarious misunderstanding or a willful decision to misinformed. Now, like I said, I'm noting Rob here for trying to have some sort of middle take that criticizes both the board and Sam, but I still think it misses the fundamental fact that your job as a board member is not to be writing very public research papers about how much better your competitors are. Even if that's your analysis, that's the type of thing that you bring internal to the company or to the nonprofit or whatever the structure is and try to instigate change.

8:32If you want to be someone who's an objective observer from outside, who can lob suggestions in and make unbiased comparisons, that's fine. But there are plenty of roles to do that from that aren't on the frickin' board of directors. Now, the other part of that Wall Street Journal piece, which I found even more insane than this, was summed up by Angelist founder Nivy, who wrote, The OpenAI board's hallucination about Sam's lack of candor spurred law enforcement inquiries, according to Wall Street Journal. This would be gross negligence and defamation. On Friday, after the board defamed Sam in their blog post, the company got calls from law enforcement, including the U.S.

9:04Attorney's Office in Manhattan. When the board was asked for details by the team, they said they couldn't provide a specific example because Sam had been so deft, as if he had cast a magic spell on them. Now, this was truly an incredible section of this piece. Quote, People familiar with the board's thinking said there wasn't one incident that led to their decision to eject Altman, but a consistent slow erosion of trust over time that made them increasingly uneasy. Also complicating matters were Altman's mounting list of outside AI-related Adventures, which raised questions for the board around how OpenAI's technology or intellectual property could be used.

9:34The board agreed to discuss the matter with their counsel. After a few hours, they returned, still unwilling to provide specifics. They said that Altman wasn't candid and often got his way. The board said that Altman had been so deft they couldn't even give a specific example, according to people familiar with the executives. So they're literally saying, Sam is such a bamboozler that he bamboozled us so good that we can't even tell you how he bamboozled us. In what universe did they think that this would fly? It's just absolutely shocking to me. So this was where things were late Tuesday night as I headed to bed on the East Coast.

10:07And of course, I woke up to the news that Sam Altman had finally been reinstated, that five days of this crazy back and forth had come to a close. OpenAI tweeted, We have reached an agreement in principle for Sam Altman to return to OpenAI as CEO with a new initial board of Brett Taylor, who's chair, Larry Summers, and Adam D 'Angelo. We're collaborating to figure out the details. Thank you so much for your patience through this. Now, obviously, this board composition is quite significant, and it turns out was a key part of the deliberation, so we'll come back to that in just a moment. Allman retweeted this with a salute emoji flanked by two hearts, and then added, I love OpenAI, and everything I've done over the past few days has been in service of keeping this team and its mission together.

10:47When I decided to join Microsoft on Sunday evening, it was clear that was the best path for me and the team. With the new board and with Satya's support, I'm looking forward to returning to OpenAI and building on our strong partnership with Microsoft. Now, Satya Nadella, the CEO of Microsoft, who reporting had suggested was hoping for this outcome, even though he had arranged for Sam and Greg and their team to come to Microsoft, it still seemed like he thought that this was the best outcome, if they could get there, retweeted Sam's post and said, we are encouraged by the changes to the OpenAI board.

11:15We believe this is the first essential step on a path to more scalable, well-informed and effective governance. Sam, Greg and I have talked and agreed they have a key role to play along with the OAI leadership team in ensuring OAI continues to thrive and build on its mission. We look forward to building on our strong partnership and delivering the value to this next generation of AI to our customers and partners. Now, if you're interested in the Microsoft side of this, on Friday's episode, I'm getting deep into how I think this all played out for them and why even though they appeared to pull off their own little coup, I still think it was a net negative for them.

11:46Helen Toner, who we have just been talking about, retweeted OpenAI's announcement and added, and now we all get some sleep, generating hundreds of responses like this one from Rick Burton, who said, you destroyed the trust of hundreds of employees, thousands of customers, and millions of users. A change in the board could have been accomplished without all of this drama. Sleep well. Emmett Scheer, the former CEO of Twitch who had been pulled in to be the new interim CEO over the weekend, wrote, I am deeply pleased by this result. After 72 very intense hours of work, coming into OpenAI, I wasn't sure what the right path would be.

12:15This was the pathway that maximized safety alongside doing right by all stakeholders involved. I'm glad to have been a part of the solution. And it appears from all the reporting that we're getting that that's more than Emmett patting himself on the back, that in fact he was an essential part of the resolution of this story. Bloomberg's Emily Chang writes, Emmett Shear joining the talks was a turning point in the negotiations. He asked the board to provide concrete evidence of Sam's wrongdoing. When they did not do so, he said he could not remain in the CEO role and helped both sides find common ground to get Sam reinstated.

12:47Gary Marcus writes, Whatever they paid Emmett for his service, I believe it was worth it. He both got Sam back and got Sam to agree to have someone look at his behavior, which for whatever reason was legitimately viewed as potentially improper. That's a hell of a delicate balance and exactly what the company needed. So let's talk about how the negotiations actually resolved here and where the key sticking points were. Going back to the thread from Emily Chang, she writes, The biggest sticking point was Sam being on the board. Ultimately, he conceded to not being on the board at least initially to close the deal.

13:16The hope slash expectation is that he will end up on the board eventually. Another sticking point? Adam being on the board. He will remain on the board for now. In general, agreeing on new board members seem to be tough, given there are only three to start. Microsoft will likely have board representation, at least as an observer, and possibly a board seat or more. All of this still has to be worked out, given potential regulatory issues, etc. So, basically it sounds like, the big issue quickly became, whether Sam was going to be on the board if he was reinstated, and whether he was going to be able to stack that board with loyalists.

13:46For much of the negotiation, Sam apparently wanted to be on the board again. And on top of that, he wanted everyone from the board that had fired him gone. Now, the deal they ultimately landed on was that Sam conceded that he would not be on the board at least initially. However, of course, the board is now three members with an intention to bring it up to the size of nine. And what's more, Sam ultimately also conceded to allow Adam D 'Angelo, the CEO of Cora and current board member, to remain as a consistent through line. Brett Taylor, who was an early Facebook team member and then co-CEO of Salesforce and former director of Twitter, was one of the only consensus choices.

14:21And so we'll take the role of chair. And then somehow Larry Summers came out of nowhere. Of course, the former Treasury Secretary, the former president of Harvard. And while we don't have a ton of reporting yet on how they landed on Larry Summers, my guess is that they wanted some amount of continuity in a Washington position board member. Helen Toner being at Georgetown had sort of played that role, and so perhaps that's what they were thinking with Larry Summers. Although, again, as I said, we don't have that much information about their thinking yet, and I'm sure we'll get more over time. Now, the other concession that Sam agreed to was an investigation about his behavior.

14:52The only detail that we have comes from the information who write, Sam Altman has agreed to an internal investigation into alleged conduct that prompted the board to oust him, according to a person with knowledge of the situation. And so while yes, Altman is coming back. These concessions are real. Having to face down that internal investigation, not having control of the board at least to start, not even being on the board to start, Altman's power is diminished relative to where it was last Thursday, for example. Now, most of the community is still left gasping at just the insanity of the whole situation.

15:23Siki Chen writes, this is actually remarkable. So the new CEO asked the board for proof that Sam did anything wrong or he's going to walk. The board was so incapable of doing so that they preferred to just bring Sam back instead of providing any proof at all. Imad from Stability AI echoed a lot of comments that I've seen, asking for broader representation as that board gets filled out. He writes, Brett is fantastic and given he made Google Maps, the AGI should know where it is going. Larry should be able to give insights on job transformation. Adam has been answering questions. But please, more representation and balance here.

15:54Powerful white guys are less than 0.01 % of the world. Continuing beyond the joke, he wrote, Diversity of thought is essential to follow charter for open AI and ensure the benefits of AGI that impact us all are distributed as widely as possible. Just as no taxation without representation, AGI built on a collective output needs diverse input. Others are reflecting on the potential long-term damage done by this whole episode. Discussing Helen Toner's comments that it would be compatible with their mission as the board to destroy OpenAI, Rune wrote, If someone holds the view that burning the entire company down is compatible with the mission of building safe AGI, It seems logical, ethical, and moral to try your best to remove them from your board.

16:33AI alignment content creator Rob Miles, however, wrote, For a normal company, sure, the board's job is the well-being of the company. OpenAI is not a normal company. Its board's job is the well-being of humanity, and the governance structure was deliberately set up to allow exactly this to happen. This is a point you've seen lots of people make. However, I find myself agreeing with this next point from Lewis, the founder at Spellcraft AI, who writes, True, it was set up to allow for exactly the situation. But here's the other part. You guys pulled the fire alarm so quickly that you've destroyed any trust there.

17:02It's over. Now, of course, Rob wasn't involved in that decision, and Lewis is using the you guys pejoratively to describe the entirety of the AI safety space. But the point remains. Again, this is something that I'll get into a lot more in my Friday breakdown episode of just the winners and losers out of this whole episode. So what happens next? Well, Swiggs argues that the new chip on the shoulder for this team is going to lead them to produce even faster and more furiously. He writes, Finally, the OpenAI saga ends and everybody can go back to building. Three things that turn things around in my opinion.

17:32One, 95 % of employees signing that letter. Two, Ilya and Mira turning Team Sam. Three, Microsoft pulling credits. Things aren't back to where they were. OpenAI has been through hell and back. This team is going to ship like we've never seen before. They're certainly going to have to to start building back trust. I agree that things aren't back to where they were. I think this is an OpenAI that is significantly weakened. I think developers have a new fiduciary responsibility to be more careful about their affiliation, about building on OpenAI exclusively, and I think there's going to be real challenges.

18:04Siki once again sums up just the utter absurdity of the whole situation, writing, In the history of corporations, has a company ever fired a CEO, hired a new CEO, fired that new CEO, hired another new CEO, and then rehired the CEO they originally fired in two and a half business days? The answer, of course, is no. But as we always said, AI is something different. Anyways, guys, that hopefully wraps our coverage for now of this crazy situation. I hope you are getting ready for a wonderful Thanksgiving holiday if you are in America and a wonderful weekend everywhere else if you are not. Until next time, peace.