Ronan Farrow and Andrew Marantz on Sam Altman’s Trust Problem

22:21The people who are scared of nukes are going to be mad at you for that.

22:24Peter Kafka:When that race kicked off in November 2022, when they introduced ChatGPT, I would talk to various people involved in tech. And on the record, they would all say the same thing, which is when you'd ask them about, you know, all this power, who's going to be in control of it? They'd say, well, we believe there should be smart regulation and we really are looking forward to that. And there was sort of the idea that, all right, we've been thinking about how we regulate or don't regulate social media. That's going to influence how we think about AI and we're going to head down that direction. Then Donald Trump gets elected in November 24.

22:57Peter Kafka:And some of his biggest boosters, Mark Andreessen, David Sachs, say essentially there should be no guardrails on this tech at all. And any guardrails we have are going to come back and bite us in the ass and they're going to allow China to leap ahead of us. Does that influence Altman's thinking or was he already there? Did Altman look around and say there's a new political landscape and I'm going to change the way that I've been talking about this? Or was he already at that point? Yeah, well, he definitely shifted. He was one of the doomeriest of doomers from not just 2015, when he's sort of courting, you know, Elon Musk and Peter Thiel, who was then interested in these arguments.

23:42But up until, you know, 2023, he was going around saying this technology could be lights out for all of us, going to Congress and saying, please tie my hands. And then, as you say, the wind starts to shift. These safety doomer arguments start to go out of favor. Trump comes back for a second term, and then the rhetoric shifts very abruptly. And so we don't know what was in his mind or in his heart at the time. Unlike Greg Brockman, he doesn't write down all of his thoughts in a digital diary that later comes out through Discovery. So we don't know. But the rhetoric publicly and privately definitely changes dramatically.

24:17Peter Kafka:Over the last year, really last six months, really last three months, there's a whole sort of bunch of storm clouds around OpenAI. Anthropic may be surpassing them. Anthropic has gone after the enterprise market. And they're doing really well. And OpenAI had sort of turned itself into a consumer-facing organization, and they've launched products like Sora 2 and then canceled them and said they're going to do adult erotica and said, no, actually not. And they went and bought a streaming talk show for some reason. Does anything you learned about Sam Altman over the last year and a half you were doing the story, does any of that reflect it in sort of the ups and downs they're going through right now?

25:01Peter Kafka:Or do you think this is just happens to be what happens to any tech company that starts out really strong and then has to face headwinds? I think this is sui generis in some ways. You know, we talked about before the baseline expectation that people don't really care that much about lying in Silicon Valley. And yet there is this huge preponderance of sources around Sam who just can't stop talking about it. And I think that is precisely because of the ramifications for the company. You know, I'm not going to play stock analyst, but what I can say from the reporting is that this management pattern and all of the implications of it, siloing information and sidelining dissenters and making different promises to different factions, apparently, this creates an environment of dysfunction.

25:50direction and the unbridled optimism and this this thing that the former board member I quoted earlier talked about of him believing the shifting sales pitches all the time it leads to the kind of mission creep that you're talking about right you know you've got all of you tell one group yes

26:06Peter Kafka:pursue that and then actually another group no you should pursue that instead and those things are going to conflict yes but also in terms of the company's direction and the product announcements you can see echoes of that same trait right you know we're going to do this Disney deal we're doing this Amazon deal. Sometimes, since I mentioned the Amazon deal, it's deal announcements on the same day where different companies are saying, hey, these things are in conflict. You know, there's a day where they announce, they reaffirm their exclusivity on stateless models with Microsoft. And then they turn around and they announce this new Amazon deal, which is to do with agents and stateful models in their enterprise solutions.

26:46Peter Kafka:And the importance that was Microsoft was their first big corporate backer and sort of solidified their lead. Microsoft is their big backer. And there are Microsoft executives saying, without getting into the technicalities too much, that basically this whole concept of building stateful agents relies on the underlying stateless models that Microsoft has exclusivity with respect to. So, you know, there are people saying that's where you get the Microsoft executives in this piece saying he's reneged on deals. He's made conflicting announcements. And I think that that goes to the question you just posed.

27:24What does it mean for the company? There are people within this company that feel like there is a practical inability to govern. You can see in the news cycle that like the CFO is publicly at odds with the CEO on the IPO timing. um you know the the person we report was being pretty pretty important conflict uh yeah if you're

27:43Peter Kafka:yeah because the cfo is the one who's going to take you public and if they're the one saying we shouldn't go public this year that's a problem and there was a huge concentration of somewhat chaotic seeming announcements around the launch of this piece for what it's worth you know um we have all this new reporting about them reneging on some of their safety commitments there's an announcement the day we launch our piece of a new safety fellowship that's very sort of airy and ethereal and doesn't really commit them to much. There's all these implications in the piece about their potentially chaotic geopolitical plans and employees threatening to quit over some of those, you know, a plan to auction AGI to the highest bidder amongst world powers.

28:25They dispute this characterization, but it really does appear to have been considered at a high level for some time. These are all things that seem to tie into them announcing this sort of new deal for AI policy. And like we talk about how a lot of this has caused spasms of doubt in Altman's leadership in the present day and in recent history and that Fiji Simo was recently sort of floated as a successor. She's now going on leave for medical reasons. So it's a lot of the COO also just got moved to special projects. That is a lot of churn for a company preparing for a trillion dollar IPO.

29:02Peter Kafka:Right. And again, you could say, Lexan, Mark Zuckerberg fired a million people when he started Facebook and all those guys left. And he melted down in front of Walt Mossberg and Kara Swisher when they asked him a marginally difficult question. And certainly he couldn't be the CEO of a publicly traded company. And that wasn't going to work. and he just blustered his way through it and we know what happened and it has worked out for him. There could be an argument that says, look, this is just the, this is a super fast growing company. And even though, even though Sam Altman is not a child, he's an adult man, that he's never had a position like this.

29:39Peter Kafka:He can also learn and work his way through this. Yeah. And the piece for what it's worth is this is not like a monolithically negative hit piece. I've actually been really encouraged by the range of reactions. There are people who read this and say, Sam Altman is a monster. He needs to be kept away from authority rules. And there are people who read this and say, like, I find Sam Altman likable and I understand where this trait comes from. That flows from a very deliberate approach we took. You know, it's a product of deep conversations with him about this trait. It allows for optimism about all of our ability to grow in the face of criticism.

30:15I think you can read the piece and judge for yourself whether this is someone who seems to be reflecting and growing. And for what it's worth, there are people in the non-safetyist, nuts and bolts, pragmatic investor community that I referred to earlier, who talk about having given Sam the benefit of the doubt around the time of the blip, this episode where he was fired, and who now say that they look back and think that he should have been dealt with more harshly because it seems like this alleged dishonesty trait has now become stable and that there hasn't been the kind of growth that you are talking about.

30:53So that's one opinion.

30:55Peter Kafka:And those are capitalists saying he's breaking the rules of capitalism and we can't abide by that. Exactly. We'll be right back. But first, a word from a sponsor.

31:09Springfest is heating up at Lowe's and for a limited time, we have the extra big deals you need to impress guests. Get your outdoor space ready and save$50 on the select Cobalt 24-volt blower kit. Now$79. Plus, save$80 on the Char-Broil Performance 4-Burner Grill. Now$199. It's springtime, and our best lineup is here at Lowe's. Valid to 413. While supplies last, selection varies by location. When you think of a phone, not a smartphone, but like a telephone, you almost certainly think of one particular device. This boxy thing that sat on your desk, it has a bunch of buttons on it, it has a handset that you pick up, it has a braided cable.

31:51You're thinking of a phone called the Western Electric 500. And for decades, it was absolutely everywhere in American homes. It was essentially illegal not to have it in your house if you had a phone. This week on Version History, our chat show about the best and worst and most interesting products in tech history, we're telling the story of the Western Electric 500, which is actually the story about the AT &T monopoly and how it fell apart. All that on version history on YouTube and wherever you get podcasts.

32:27Peter Kafka:And we're back. You mentioned the interviews you did. You said you did more than a dozen interviews with him. You guys are from the New Yorker. You're not doing puff pieces. Ronan is celebrated for bringing down Harvey Weinstein. How do you convince Sam Altman to sit for one interview with you? And why does he do it a dozen times? What does he think is coming out of this? Well, I think a couple of things. One is the trait that we're talking about, that Sam genuinely believes he can persuade anyone of anything. That includes reporters. You know, he's out there in the press a lot. and you know he was generally cooperative often charming clearly thought he could help shape the story and I for what it's worth credit to him for engaging because a lot of people in his position would have lawyered up and gone silent and I do think I can see that in the first the first setting and maybe he believes you're like one of the many interviews he does with people who fawn over him or don't ask him a difficult question but I think you have to get pretty gnarly yeah no no he he knew very clearly because also I made a point of him knowing, you know, this kind of a long lead investigative piece has no gotcha component, right?

33:40You're checking basically every sentence with the people affected. You know, the volume of responses from OpenAI, the hours and hours on the phone with Sam as we were finishing this out, it was a lot. And those conversations were at times contentious, as you might imagine. But they were also really geared towards fairness. And I was really listening deeply. And if there was anything where I felt like Sam was making a decent case, hey, this might be true, but also it could be taken as sensationalist, that stuff went out. This is truly a generous piece of reporting to him in many ways, despite the fact that it carries this really bracing criticism.

34:14And how I rendered it to him was this is going to be tough, right? The design of the enterprise is to zero in on the biggest points of contention and criticism and inspect them. He got that. And I think he still felt it was in his interest, A, to shape it. And B, there's another dimension, which is there is a specific incentive for him to engage on some of the personal aspects. You know, you may have seen they're invoking this piece in court now. They sent a letter to relevant attorneys general.

34:44Peter Kafka:Yeah, we should pause for a second. The OpenAI official response to your story that they gave me on Monday does not say this is not true. In fact, it doesn't specifically take aim at any of your reporting. And obviously, you guys have a rare and super diligent fact-checking process in journalism. But at no point does OpenAI say, this is not true. And you're saying not only that, but they're using it in a separate court case. Yeah, that's exactly right. And obviously, we can't control what people do with the information we surface after the fact. This is an interesting one where OpenAI is relying on this as reliable information in their letter to these AGs saying, you know, look at this and the way it documents the way Elon is going up against Sam.

35:27They're accusing Elon of anti-competitive behavior. This is all germane to their long and winding lawsuit with Elon Musk. On the other hand, Elon Musk is simultaneously, you know, pushing this through his various, you know, Twitter channels and posting a lot of interviews with us and so on. So that's all out of our control.

35:44Peter Kafka:Katie Miller's wife, Katie Miller, retweeted my tweet that had the open AI response. You can sort of see the you can see how who who might have a problem with Sam Altman in the current moment. Yeah, you know, the information is going to go into the bloodstream and people are going to use it how they may. I think for me, the important thing here is the way in which this describes problems that are in some way uniquely extreme with Sam, but that also do open up a window into a wider set of questions about accountability for AI labs and AI leadership. One more process journalism question for you.

36:21Peter Kafka:You guys mentioned this Dario dossier, Dario Amidi, one of the many people who worked with Sam Altman, who no longer works with him and in fact is one of his main rivals now, co-CEO of Anthropic. You say there's this 200 pages of documents, internal emails, memos, et cetera, have been circulated by colleagues in Silicon Valley, but never been disclosed publicly. I was astonished to read that there is that a file that big that is circulating wildly that hasn't leaked out WikiLeaks style. How could this both circulate widely and be kept under wraps? How do you think about that? I think that's true of a number of documents in this story.

36:58right? I mean, the Ilya memos are another great example where like they're at the heart of a contentious legal case. They were at the heart of this firing. People knew they existed. There has been incredible reporting on this subject by Keech Hagee, Karen Howe, and a slew of other people, you know, and we talk about the books about Altman. A lot of the general themes have been excavated. But, you know, it is just a familiar reality for me of many stories that that it takes, I mean, to use the cliche, it does take a village and like a community of reporters chipping away over time to more actively surface some of those particular.

37:35Peter Kafka:I guess I would think that one of Altman's many, many enemies would wanted to get this out sooner. If these documents are circulating that wildly, that someone would have dropped this somewhere and said, hey, look at all this shit. Yeah, there's a lot of things in this story that you would think. I mean, the amount of speculation around the firing was so intense that the number of things that are only coming out now is kind of surprising. And look, I think it speaks to like a lot of the people in this industry are very calculated and very pragmatic. Yes, there are these wild, ruthless rivalries, of course.

38:13And we go into great depth about that in the piece. But a lot of people think this through and I think fairly come to the conclusion like, OK, if if Sam Altman isn't the AGI dictator, you know, do we want Elon to be the AGI dictator? Do we want it to be Ilya? Do we want it to be Greg Brockman? Do we want it to be Mira? And I think it's really crucial that, yes, this is a piece that is using one person as a lens. Yes, it has his name in the headline, so I'm not disputing that. But I do think that when you come away from it, hopefully you don't come away thinking, wow, I have problems with Sam. I really hope someone else becomes AGI dictator.

38:53Right. The problem is structural here.

38:55Peter Kafka:That's Andrew. You brought me right back to where I where I wanted to go for to wrap this up. So like you say, there's there's Sam Altman, but there are several other men who it's not just that they're billionaires, not just they're wealthy. You see a lot of people saying people are complaining about billionaires, but they don't understand that they they earn their wealth, honestly, and they're it's not just the billionaires. They have enormous power. They have enormous power over their companies, whether these are publicly traded companies or not, they control them. And their wealth and power, we used to say this all the time about big tech companies, they're essentially nation states.

39:28Peter Kafka:And now these individual men are that way. And they can try to use that power to colonize space. They can use that power to try to influence elections in the U.S. and abroad. So I'm wondering how you think about that structural issue, which you just brought up, Andrea. What should a normal person think about that? And what, if anything, should we be doing about that? Well, I mean, I think Ronan also has a lot of thoughts about this. I mean, I'll just briefly say, like, I think we're well past the time when people can kind of think, oh, maybe this is sort of a parlor trick. Maybe these are sort of stochastic parrots and they don't affect my life in any way.

40:07Like whether your concern is this sort of existential sci-fi stuff or not, which I think is worth thinking about personally. But even if that's not your bag, right? If you're worried about the economic bubble that could cause potentially a market panic or a recession. If you're worried about, you know, slop, you know, choking off all of your information points. If you're worried about copyright stuff. I mean, you know, you know the whole list. I just think we're well past the point when we can say, like, I wonder if some future this will start to affect my life. It's clearly already affecting all of us.

40:41And I think one point that Altman does make that I think is really legitimate in his piece, The Gentle Singularity, is that because this stuff is happening so quickly, but just slow enough that we can sort of accommodate and adapt to it. We sort of don't notice even when we're past a number of singularities, which I think we kind of already are. And I think to go back to a point that I was making earlier, a lot of what's new here in this piece is actually about what was omitted and about lack of past transparency and accountability. So like with respect to this pivotal law firm investigation that was part of rubber stamping Sam coming back, people knew there was an investigation.

41:20They didn't know that there was no written report. They didn't know that the decision to not write it down was made on the advice of the lawyers of two new board members that Sam had been personally involved in picking to oversee this and steer it in this direction. You know, they didn't know that Altman has been claiming that all new board members after he came back got oral briefings on this, you know, unreported report that wasn't in writing. And that then we really learned that's not the case. you know, and we have one of the people overseeing it saying there was no need for a written report.

41:56And we get into all the details of that legal analysis. There's a number of cases like that, right? There's a case where we didn't know before that a board member wanted to vote, it appears, against the nonprofit conversion into a for-profit when OpenAI was doing that. And we have a lawyer for the board saying, oh, well, if we record it that way, it could raise questions about the legitimacy of this in the future. And they record it as an abstention, You know, OpenAI disputes this, but these are all examples of failures of structures and accountability. And so when I look at the central question, Andrew is exactly right in laying the foundation of people need to care about this because they're not just building any widgets.

42:35Their own pitch is that this could endanger all of us in all these different ways. You know, many people's.

42:43Peter Kafka:But let's say we had another blip and you swap out Sam Altman with a Dario Amodi or just whatever your fantasy person is, a responsible human being who believes in the greater good and is super responsible. Is that adequate that we just put all our hope in any person? No, there is no fantasy person. And that's part of what I'm getting at with the bigger point. This is about structural failures. We are in this sort of post-Citizens United landscape where money from the private sector and specifically Silicon Valley has flooded Washington, where the hopes for meaningful guardrails through regulation and legislation are dim in the view of many of the critics who raise these concerns, and where there are specific steps that we could take that could apply those kinds of guardrails to any deficient leader in this field.

43:32You know, that is the point. Sam Altman is a uniquely extreme example of some of these allegations around lack of integrity. But the problem is that any of these guys could fail you. You talk about Anthropic. Just look at the recent history of Anthropics leaks, you know, their source code going out all over the place. the question raised about Sam is by one colleague, you know, do we want this guy to be the one with his finger on the button? Any of these people with the finger on the button, they are private companies that are really unaccountable. Sam Altman is one of the people who actually very actively made the argument, you know what, the government may need to take control of this.

44:11That's not the policy proposal I'm making. The point I'm making is we need guardrails on whomever it is leading this technology. We need oversight. This is not a pipe dream, right? We do this with pharmaceuticals. We do this with food. There are policy steps being taken in regulatory and legislative environments elsewhere. Europe has much more mandatory pre-deployment safety testing for AI models already. You could have written public record requirements for internal investigations like the one that I mentioned. You could have more robust required national security reviews of the kinds of infrastructure projects and autocracies that we talk about in this piece.

44:54You could have whistleblower protections specific to the AI industry, right? People were very fearful to talk to us in many cases. There right now is no federal statute protecting AI company employees who disclose these kinds of safety concerns about frontier models. that's easy to solve, right? You have Sarbanes-Oxley for financial fraud. You can put in place protections, and we need those protections regardless of who it is.

45:21Peter Kafka:These things are all possible. They seem impossible in America in 2026, but things could change. Andrew Ronan, thank you for your work. Thank you for your time. I really appreciate it. Yeah, thank you, Peter. Thank you. This was great. thanks again to ronard farrow thanks again to andrew morance really good to talk to them thanks to charlotte silver who produces and edits this show thanks to our advertisers who bring it to you for free thanks to you guys for listening we have another podcast for you next week i'm not sure exactly what day it's going to come out maybe monday maybe wednesday we'll see see you soon

46:22We'll see you next time. zero pressure. Right now get up to 40 % off site-wide plus get a free professional measure at blinds.com. Rules and restrictions apply.