Sam Altman’s Trust Issues at OpenAI
00:02This is The New Yorker Radio Hour, a co-production of WNYC Studios and The New Yorker.Welcome to The New Yorker Radio Hour. I'm David Remnick.Just a few months ago, Andrew Ross Sorkin, the financial journalist, was on this program,and he quoted a figure that was really remarkable.Virtually all of the recent economic growth in the United States, Sorkin told me,is investment in artificial intelligence.A lot of people are concerned that a huge bubble of...around AI is about to pop and take the economy with it.
00:38And a few people continue to feel that AI is just overhyped.But I don't think there's really much doubt at this point that in our lifetimes, at least,AI is going to bring changes as significant as the Industrial Revolution 200 years ago.At the center of this world-changing technology is a man named Sam Altman, the CEO of OpenAI.It was OpenAI that really brought artificial intelligence into most of our lives with chat GPT,and that exploded into our consciousness in 2022.But the chat bots are just the tip of the iceberg.OpenAI is planning to go public this year, and it recently fundraised more money than any company ever.
01:25Ronan Farrow and Andrew Morantz have spoken with over 100 people closely connected to Sam Altmanand with Altman himself many times.They began by looking in particular at the week when Altmanwas very suddenly fired from OpenAI,and days later reinstated a CEO.That whole episode has been mired in secrecy and confusion.Ronan and Andrew see the firing, the blip, as they call it,as a key to understanding Altman and the problems with his leadership.
01:55Their extraordinary investigation in the New Yorkeris called Sam Altman,may control the future. Can he be trusted? Now, Andrew Ronan, you compare Sam Altman to Robert Oppenheimer,who of course was pivotal in developing the A-bomb. Oppenheimer not only developed a technology,but in a sense he defined an age in American life, the atomic age. But there's, of course,something extremely ominous about that comparison too. So let's begin this way, Andrew.Who is Sam Altman? And why would you compare him to Robert Oppenheimer?Well, for one thing, we compare him to Oppenheimer because he compares himself to Oppenheimer.
02:38I mean, constantly.Constant throughout the rhetoric before Open AI existed for why it needed to exist, there's this constant thread of analogies to the Manhattan Project.So when he emails Elon Musk out of the blue in May of 2015, he says, hi, Elon, this is Sam.He says, I think we need a Manhattan project for AI.And it does have this dual-edged nature to it, which is both we're going to be the good guys and defeat the bad guys.Right?We're the Americans and we're going to defeat the Nazis.Yeah.
03:10But so in this case, instead of the Nazis, it's either China in a national security context or Google in a competitive corporate context.But it's also the ominous part, you know, and is the notion that the atomic bomb defined an age, the atomic age.It still looms over our politics and global security.what is the potential ominous aspect of AI?We hear about it as something that could be fantasticfor the development of drugs,for all kinds of things,but it could also wipe out,
03:40God knows how many jobs,but it goes darker than that.I will say at the outset of this reporting,I was not myself convinced of the much-Balihoodtransformative impact of this technology.I really emerge from this more convinced.There's the scenarios that I,I think you're alluding to, right? The atom bomb-esque ones, the idea that this really could lead to a kind of Terminator skynet scenario where a rogue artificial intelligence falls out of alignment with what we want and takes control of nukes.
04:13But you don't have to buy into those extreme scenarios to also see the immediate impact that the entire U.S. economy is now propped up by a few companies that are all in on AI with Open AI at the center of it, that most credible projections very immediately foresee millions of jobs being exposed to disruption by this, that it's already taking form as a significant change agent in the way weapons are used in battlefields, that they can now,just fire without human operators.That is something in the very immediate future.It's taking form in bioweapon development.These are all things that are happeningor on the verge of happening,not far off fantasies.Now, just to be clear, Andrew,
04:59what most of us have in mindin terms of AI is chatbots,which run on large language models.How is that different from the aspirationof an AGI, artificial general intelligence?And that's something that doesn't exist yet.ChatGPT is a useful tool that can help you write emails.And AGI is supposed to be able to do any cognitive task that any person can do arguably better.
05:24So it's general because it can not only write emails or play chess, but it can also discover new drugs or solve new medical problems or break new ground in physics.So we're not quite there yet.What's interesting, Andrew, is that you quickly describe Sam Altman as somebody who really doesn't have that much.computer science knowledge that when it comes to the actual science, the actual technology,Sam Altman is no great genius. So what does he bring to the table that makes him such a transformative figure?Yeah. He, in a sense, you know, what every entrepreneur does is, you know, bring together technical talent and investors and merge them into a solvent company.So in that sense, it's not that remarkable. I think what makes him unusual is,he's not bringing this kind of bull in a China shop.
06:17You know, he's not Elon Musk or Jeff Bezos and just saying, I'm going to bulldoze my way to power.He brings this kind of subtle sales pitch around 2014, 2015, 2016, where the public is feeling kind of burned out on the bluster of tech tycoons.You know, everyone's suddenly mad at social media, Republicans for stifling their speech and Democrats for giving us Trump and whatever.And he comes in and instead of saying, who needs.Luddite regulators, he says, actually, please regulate mebecause the thing I'm building is so scarythat if you don't regulate me, everyone you love will die.And it turns out to be a counterintuitively
06:54very good sales pitch.And even more importantly, it's a good pitch for recruiting.So it's a pitch to what?He had been at something called Y Combinator,which was a kind of factory for technologyand technology investments, a development place, if you will.And then he begins right.And he begins something called OpenAI,
07:12which is what, Ronan?Well, it started out as a nonprofit.And this is really at the heart of the story.This is a distinction that Elon Musk is now suing Sam Altman and Open AI over.Well, who was involved in Open AI in the beginning?So at the very beginning, Sam Altman, who had a lot of irons on the fire, right?He was investing in a lot of different cutting-edge areas.And AI became a fascination.
07:37He saw the industry moving towards that.And an interesting thing happened where he had been very optimistic.about technology in general.But he, I think, saw an inflection point,according to people we talked towho were in conversations with him at the timeand knew his thinking,where a lot of the leaders in the industry
07:53were becoming more apocalyptic about it.And one of those people was Elon Musk.And in fact, the very people,the scientific researcherswho were most capable of building advanced AIwere some of the most freaked out about it.So in order to pitch to them and recruit to them,he had to come to them,
08:07according to a lot of people,and say, I'm as scared as you are.And then we learned very quickly,very, very quickly, that a lot of people around Sam Altman discovered that, well, they just don't believe him, and they rebel against him, and they think of him as a liar. Why is that, Ronan? These two things are deeply interrelated. Yes, Sam Altman could have presented this with a for-profit motive more openly and perhaps earlier, his critics say, but part of the power of his pitch was that he went to Elon Musk and he said, I hear you. I hear you.This could destroy humanity.We need to put something together, and we need it to be safety first, and it can't just be about racing to get the technology first.This has to be a scenario where we're willing to slow down on development to keep it safe.It was all rooted in a fear-driven argument.
09:00And that was why Google was the bad guy, because Google is for-profit.They're the mega-corporation, so we're going to be the good guys.And that was, as Andrew alluded to, a really powerful recruiting tool, because part of Open AI's strength was not onlydid Sam Altman through these quite extraordinary powers of persuasion present this sort of fear-drivenrationale why he needed to get the money for this. He also was able to go to the brightest minds in thefield and say, this is a non-profit. We may not be able to pay as much as Google, but we can give yousomething else, which is we're the good guys. So what happened a few years ago? The blip? Yeah, what was theblip? So one of the top people who they recruited, who took, who was offered $6 million a year at
09:41Google and turned it down in order to go work for the good guys was this guy, Ilya Sutzgevur,who was on the board in 2020, in 2023, and he started to get the feeling, as we quote him in thepiece, saying, I don't think Sam is the guy who should have his finger on the button to returnto the atom bomb analogy. And so he starts to rally the board against Sam. Now, this has been alingering question for years in Silicon Valley. What did Ilya see? What is in his secret memos thathe compiled, is there a smoking gun? Is there some one thing that explains it all? And what we found,and the reason that you graciously gave us 16,000 words to explain it, is that there is not onesmoking gun. There is this small accumulation of detailed patterns of behavior that add up to,
10:28in aggregate, what people like Sutske ever felt was someone who can't be entrusted with thisworld-altering technology. For example, there are certainly specific episodes over the history ofOpen AI that have not yet been extensively reported.We have a detailed account of a moment where critics within the company allege, and peoplethreatened to quit over this, a plot was entertained at the highest levels of the companyto sell the next generation of this technology, AGI, when it comes, to the highest internationalbidder, to pit China and Russia and the United States against each other.And that really did prompt some of the safety-minded people in the company to say, you know,
11:10as one tells us, this is insane.What would have been the repercussions?I mean, potentially a situation where...Military, right?Vladimir Putin has disproportionate controlof the most dangerous technology in the world.Now, we should say this is somethingthat ultimately did not come to fruition
11:24and Open AI says, oh, it was just one of many ideasthat was considered for a time.But there were people...That was just an ideathat we were considering selling thisto Vladimir Putin.Well, there you go.And, you know...
11:34It suggests kind of amoralism, no?That's what the critics say.One of their defenses was...We weren't talking about selling it.We were talking about giving it to them.Much better.So these are all hypotheticals, and they are kind of crazy-sounding hypotheticals.What emerges is a newly deep picture of Sam Altman through inside communications that haven't been out there before.
11:58What was behind his firing?Why those criticisms linger?And exactly how substantiated there.There are many, many instances.The famous memos that have never been made,public and got him fired within this company are now in there and we really go through the evidencein a very forensic way. You know, this is not a simple piece. I don't think people are going to read thisand have everyone reacting in one way. I think there are people who will look at this and they will
12:25come to the conclusion that maybe the sort of profit-minded people who think safety as the main priorityis more of a thing of the past are correct. There are people who will look at this and say,this is really scary and dangerous. The phrase that the board used at the time was notconsistently candid. Lyer, I believe, is the word use. Is it not? By many, many people and eventhings like pathological liar in great duplication across many sources. We talked to more than ahundred people that are close to Altman in one way or another. And you talk to Altman six times.So tell me what that experience was like. You went to see him. More than a dozen times in the finalfact check. No, no, no, I gather. And what's he like? Because I interviewed him here at the New Yorker
13:09I have to say, it was like interviewing a cloud. He answered in a very mild way and in a very skilledway too, I would say. But I couldn't get my arms around his answer sometimes. We talked aboutthe job loss issue, for example. And here's a bit of that interview from 2023.I don't think that most people won't work. I think for a bunch of reasons that would beunfulfilling to a lot of people. Some people won't work for sure. I think there are people in the worldwho don't want to work and get fulfillment in other ways, and that shouldn't be stigmatized either.But I think many people, let's say, want to create, want to do something that makes themfeel useful, want to somehow, like, contribute back to society.
13:54Well, let's slow down for a second. What does this imply in the much broader sense aboutwhat change is coming down the road? In concrete terms.I think it means that we all are going to have.much more powerful tools that significantly increase what a person is capable of doing,but also raise the bar on what a person needs to do to be sort of a productive member ofsociety and contribute, because these tools will do eventually.You know, I'd say that I left that interview not all that happy with myself.I didn't feel like I got any great purchase on what Sam Altman was about.
14:38but you obviously did with much greater time and much greater access, Ronan.People find it hard to wrap their arms around exactly what's going on with his motives.He is able to, and inclined to, tell different groups of people possibly conflicting thingsthat make them all feel that they have the same concerns he has.That is an extraordinarily useful skill for a business person.but over time what one person after another,a majority of the sources we talk to found,is that they just couldn't rely on any baseline of truth,
15:18that to an extraordinary extent,even on small things that are completely unnecessaryto have any deception about.We talk about an instance where he's in an office with colleaguesapparently claiming that he was like a champion competitive ping pong player.And Alman says that this was probably a joke,but people thought it was serious enoughthat they were struck by the fact
15:37that then he was one of the worst ping pong playersin the office.This is to give you an exampleof how banal it is sometimesand then as Andrew pointed outit extends to people allegethat he has concealedsome of his financial interests
15:47that he deceived some board membersand executives about the safety testing requirementsof some products.So it also then filtered into serious stuffand I will sayI've never been on a story beforewhere something this peculiarin that it's not a bright
16:04line smoking gun, but it is so prevalent that it is almost, I'm barely being hyperbolic here,all anyone can talk about after walking out of rooms with him, that is an extraordinary thing,and very difficult to grapple with in a piece like this. We try to do it with great subtlety.How is that different from Elon Musk? Well, so this is part of the presentation, the temperamentalpresentation that Altman brings is not this kind of brash, swaggering.Not at all. Not at all. And so I think that's part of the pattern. If you hear from Elon Musk,I'm not boosting my own Twitter account, and then you find out that he was.You're like, okay, not particularly shocking.
16:40Kind of Trumpian.Yes.The Altman thing is very low-key, it's very thoughtful.Part of his pitch is presenting himself as conscientious, I hear you.One person, we don't quote this in the piece, but one person once called him the Michael Jordan of listening.Part of his...Right, so he looks you in the eye and he looks you in the eye.Yeah, exactly.
17:01And reflects back what you want to hear.And I think this is an important sort of summary after spending a year immersed in this.We represent in this piece a range of perspectives.There are the defenders of Altman, of course, and we spoke to many of them, and they're represented in this piece.And amongst the critics, there's a spectrum.There's the people who are, you know, die-hard safetyists, and they say, look, this technology may kill us all.And he is Machiavellian and sociopathic.There's all these extreme terms that get used by some critics, and that therefore this is truly dangerous.
17:37There are people in between who say this is just dysfunctional management, that even if you're purely profit-motivated,an executive of a company this important can't be making conflicting representations all the time.I tend to feel inclined towards the analysis of Sam Altman through my own dealings with him in this story that holds,know, it's not that Machiavellian. I think that he is someone who, he actually grapples withthis in a new and more sincere way in this piece. He talks about having had some problems with this.He doesn't just pretend that this doesn't exist around him. He talks about changing over time.He talks about the deep roots of feeling like a people pleaser, which, you know, I understand.But I think he is reckoning now anew with the costs of that when it's taken to an extreme.
18:30I'm speaking with Ronan Farrow, along with Andrew Moran's.They've co-written a long, deep investigation of Sam Altman and the rise of Open AI.That's all on The New Yorker this week.We'll continue in just a moment.This is the New Yorker Radio Hour.This is the New Yorker Radio Hour.I'm David Remnick.Today, U.S. government agencies are starting to enforce a ban that President Trump imposed Friday
19:15barring the federal government from using AI tools made by Anthropic.In February, a feud erupted between one of the leading AI companies, Anthropic, and the U.S. government.In short, Anthropic was providing artificial intelligence capability to the Pentagon,but Anthropic wouldn't allow its clawed system to launch autonomous weapons or to be used in mass surveillance.In response, the Secretary of Defense Pete Hegset and the Pentagon called Anthropic a national security risk.Anthropic turned around and sued.And into the breach, stepped Sam Altman, the CEO of OpenAI,and he swiftly made a deal with the Pentagon and replaced Anthropic.
20:01This is the same CEO who said three years ago to Congressthat he feared what could happen if AI was deployed incorrectly.I think if this technology goes wrong, it can go quite wrong.And we want to be vocal about that.We want to work with the government.to prevent that from happening.But we try to be very clear-eyedabout what the downside case is
20:24and the work that we have to do to mitigate that.How did Sam Altman,who's now 40 years old,change his mind after all this timeon such a fundamental issue?I put that question to Andrew Moranson, Ronan Farrow,who have jointly publisheda deep investigation of Sam Altman
20:40and his history at OpenAI.We'll continue our conversation now.Andrew, what did Sam Altman sayto his own employeesabout this military contractversus what he said to the government.Altman said publicly,we stand with Anthropic,
20:58we think these are the rightbright lines for them to hold.Privately, he was negotiatingwith the Pentagon for the same contract.And this gets to the Shakespearean rivalry stuff as well.This is often personal.This is often petty.But because the stakes are presented
21:16as being so high,like literal existence,I mean, people talk about it as who will win the AGI dictatorship.People talk about it as, you know, who will get the golden ring, who will get the ring of Soron.So I think they just think, if you're Sam Altman or Dario Amadeh or Elon Musk, I think in their minds, it's like anything is worth doing to win that competition.No gold card.It is totally existential.What are the financial stakes for Sam Altman running?
21:48immense. One of the things that we talk about is Sam had both positive and negative arguments he used to buoy this company. We've talked about how he marshalled people's fear. He also really rallied people around the optimistic projections of what this technology is going to be. And there are blog posts from him in recent years where he talks about, you know, we're right on the cusp of maybe even have cleared the event horizon, as one term he uses.And God knows what that meant.Well, a trajectory that will bring us very imminently to not only artificial general intelligence,but a further development beyond artificial superintelligence.And in turn to, he itemizes, for instance, curing cancer, traveling to other planets.Essentially...Capturing the light cone of all economic value.Okay, I read that.
22:40Time out.We put it in quotes.What the hell did that mean?It's a sci-fi-ish thing that basically.means capturing all the economic value in the solar system.So it involves space colonization usually.So we're talking, you know, comfortable living after that.Super abundance.
22:58We'll all be chilling.So some of it is the, you know, the Keynesian thing of like in the future, we'll all havelives of leisure and we won't have to work.But you can see how this stuff.Too late for me.And he does, for instance, he appears to believe it.And he advocates for, you know, universal basis.income as a part of this future. But when we asked, for instance, you know, what do you think
23:25about all of these economic projections that hold that so many jobs are going to be disrupted bythis? The reflection and the grappling with it was not, in my view, terribly deep. You know,he does believe or seem to believe, as he says them, all of his optimistic projections.But then he also says it's a bubble at the same time. And so it's like, howdo you? He basically on the joblessness and the risk of the bubble, you know, he says,well, actually, it's just going to make everything better. Everyone will have access to chat GPT.That's going to allow people who are unemployed to, I mean, I'm paraphrasing broadly, but essentiallyit's going to allow for more startup creation, and that's going to help everyone, and we'll have a big
24:07old foundation, and we're going to do some charitable activity, and that'll help.And also, you see, I mean, Ronan, you brought up his blog posts, just to talk about the shiftingpitch over time. You know, people callhim this great pitch man.The blog posts now are verybullish, optimistic, we're going tocure cancer. You go backtwo or three years and the blog posts say
24:27we need to solve alignment or we will have arogue AI that stamps out humanity.The alignment problem. So the alignment problem is supposed to beif we build a superintelligencethat is not aligned with our interest,it might, so it doesn't have tocome alive and becomeHal 9,000 or whatever, to kill us.
24:44We quote a blog postin the piece of someone saying, all it has to do is, is, you know, be misaligned with our interestsand accidentally kill us. And the person who wrote that blog post was Sam Altman.One of the things that you're kind of getting at here is a politics, kind of...Diplomacy.Ethereal politics. And, you know, some of his rhetoric is kind of utopian lefty.And yet he's made his accommodation with and friendship with the Trumps.Tell me about that.
25:14Yeah.So for a long time, he donated to Democrats, and he said Trump is this unacceptable threat to America.Compared him to Hitler.As did his vice president and several other people.And then he, you know, in 2024, he starts to, you know, shift.You dialed back to Hitler.Dowled back the Hitler.We almost went with that as the headline of the piece.
25:39And then he starts to say, you know, I think this country will be okay no matter what happens.and it seems very clear, actually, according to a bunch of Biden administration national security officials, who we spoke to.He used to go to the Biden White House all the time.All the time.And encourage them to regulate more heavily and say, this executive order doesn't go far enough.We need to restrict and regulate this technology more.Then Trump comes in, literal day one, literal first day of the Trump administration, they announced massive new data infrastructure projects.And then Trump and his administration start blessing this acceleration off.
26:15the races. And the rhetoric of the Trump White House's safety is a, is a false concern. We heard that from, I think J.D. Van says this in the piece. And David Sachs and, yeah. Safety has fallen out of favor in Silicon Valley and Washington to a great extent. And one of the things we document in this piece is Sam Altman's various transformations and his conflicting stances at various times also represent a wider sea change. The moment of the blip when people in this industry were still.The coup attempt, the firing.When people in the industry were still uncertain about whether you should treat executiveswho shape this transformative technology as just other executives and hold them to those normal standards,or whether this requires people with an elevated level of integrity because they hold our future in their hands,that was unsettled at the time in a way that really led to these events,where you had a company that started as a nonprofit, was still to some extent a nonprofit,a bunch of people who joined signing up for that mission,
27:18and they said, this guy is lying too much, and he has to go.And just for the record, now OpenAI is solidly a...Now it is a for-profit.No less than the others.And it's what was once...Yeah, they call themselves a PBC, but we were told...What's a PBC?A public benefit corporation.
27:35But functionally, it is a for-profit institution,and what was once the main event,the nonprofit now has a 27%?26%.It has a minority state.Yeah. The inflection point of the firing and the reason we look at it as an important one in this piece is a moment where that argument that we should have this elevated standard, the rubber met the road. It was tested. And what we see afterwards, him clawing his way back in this way, partly on the strength of going to a bunch of investors, as we document, we get inside of those rooms and we see how those conversations went and saying like, hey, these people just fired me for this, you know, vaporous thing. Now, you know, you.he was able to do that in part because I think that old board that fired him did notacquit themselves strategically in many ways, shall we say. There was not a lot of transparency,
28:24but he was able to, in making the argument for himself, also I think assert and get into thebloodstream a broader argument that that kind of safety-mindedness no longer has a place in a raceto achieve AGI with massive economic stakes. And today you see that reality that there is a veryfair argument that these firms are on safety engaged in something of a race to the bottom.Elon Musk isn't the only person that's suing OpenAI. How many court cases are there currentlyagainst the company that are related to suicides and murders allegedly prompted by chat GPT?Yeah. Many ongoing suits for various different things. So some of this has to do with liability.you don't have to believe in the sci-fi existential harms
29:14to see the very real harms that are already occurring.Psychosis that is furthered by addiction to these chatbots.How would that happen?Well, this is not just like a Google searchwhere this is an ongoing kind of relationshipthat people have with a chatbot,and the chatbots can be sycophantic.ChatGBTGBT is now able to convince
29:38human observers that it is human more often than humans are able to.So we're in...Please be more specific before I throw myself out of the way.This is research published last year that when you look at this test of can a chatbotdeceive a human observer into thinking that the chatbot is a human,human beings pass that threshold less often than chat GPT?And people are doing these quizzes all the time of, you know,can you tell what's AI writing and what's Kormick-McCorp?
30:08Carthy and, you know, people are no better than 50% and stuff. So we don't...That promises well for our profession. Yeah, well, so this is why I was going to say,we word people don't like thinking about this. But it turns out that it kind of is the casethat if you take all the words in the universe, crunch them onto a chip, it can kind of createa gallum of new words that can kind of infinitely spit out. And we should point out that Kanye Nas,The company that owns the New Yorker and Vogue and Wired and all the rest has a deal like many, many other publication companies with OpenAI and other such companies for that very purpose.Yeah, and like all these.They're limited and, you know, I think the fear for a lot of publishers is that they're going to take it anyway and we'll get nothing for it.
30:58Like a lot of these dual-use technologies, you know, even the most dire critics of this stuff can't deny that it's useful and fun and engaging.And, you know, if it weren't so useful, it wouldn't pose such an economic threat.And worth pointing out useful in a very sincere and deep way.I mean, when you look at the medical applications, lives are being saved.For diagnostics and research.It is a game changer for things like, you know, severe weather warnings.which may sound banal, but that is truly a lifesaver.There are all kinds of applications where this is the real deal already.
31:39I'm speaking with Ronan Farrow and Andrew Morance,who have just published a long and very thorough account in The New Yorkerof Sam Altman's tenure as CEO of OpenA.I.Our conversation continues in a moment.This is the New Yorker Radio Hour.Stick around.This is the New Yorker Radio Hour.I'm David Remnick,
32:17and I've been speaking today with two of our writers,Ronan Farrow and Andrew Morantz.They spent over a year reporting on Sam Altman,the CEO of OpenAI,for a piece that's just been published in the New Yorker this week.They wanted to understand Altman and his vision of the future,but they also wanted to understand where he comes from,including parts of his life that Altman himself seemed wary of analyzing.
32:46You wanted to get a sense of who this guy was,and I think he told you that he was the victim of a really serious,homophobic attack when he was a teenager,although he was reluctant to go into it in much detail.How does he think, Ronan,that that moment shaped him, if at all,and how do you see it?If I'm being honest,
33:07through this reporting, I feel oddly somewhat connected to Sam.He may have a different view of the rapport.I do think I understand him on a certain level.You know, I had these moments with himwhere I would ask, like,what is your personal human experience of so many people around you saying these things about yourhonesty and integrity? Because for me, it would be such a terribly devastating thing to hear.And I will say on this point, he is somewhat resistant to self-reflection, is my impression,
33:41as he is on a range of these questions. When we talked about his roots and some of theseperhaps formative factors, his sexual identity, his Jewishness, you know, his socioeconomicbackground. Which was prosperous. Prosperous, right, a kind of fancy suburban family's mother's,a dermatologist, very well connected, his father is a kind of housing activist. He often, and particularlyon the sexuality question, you know, he'll even cop to the lack of reflection. He's in thepiece saying, well, yes, you know, I got beat up this time. I don't want to talk about it becauseI think people will see it as me playing for sympathy and being manipulative. So he's internalizedsome of these critiques. And also he sincerely, in that moment and later when we were fact-checking
34:33this piece, just really emphasized, I want to be dismissive of that. You know, this did not defineme. And when I would ask him some of these questions, you know, what is your experience of this?Have you grappled with this in therapy, for instance?Do you do therapy?You know, he recites a lot of the kind of conventional West Coast wisdom about, like, breathwork,and said he has tried and liked therapy vaguely.But it was very clear that amidst this stratospheric rise and a lot of these difficultiesand this would be for most of us, I think, painful criticism,
35:09there has not been a process of really looking in.word. That is my take having, you know, asked him these questions. He sort of is breezily dismissive.And I think that that runs through a lot of the professional situations that we narrate, that there are other people who really have a relationship that matters to them internally with the truth of a situation.And I think he is uniquely suited in some ways to this job because he really can effortlessly shift between one version of reality and another as he is marshalling people to his cause.You spent a lot of time in the last year looking into some very lurid allegations about Sam Altman's personal life, sexual life, what he may or may not have done.What did you conclude? What can you say about that?So this is one area where it's important to note we didn't set out looking for that.Part of what makes this circumstance extraordinary is the prevalence of this allegation that Sam Allman lies all the time.
36:11another thing that makes it extraordinaryis that in addition to those, in my view,more substantive critiques that are so widespread,there are incredibly widespreadclaims about his personal lifethat don't stand up to as much scrutiny,again, in my view, having looked at this for months and months.We capture the way in which actually, ironically,
36:31the existence of these falsifiedor thin claims trumped up by rivalskind of obfuscates the real criticism.the atmosphere of conflict in this field is, we quote one executive saying, Shakespearean.The most dangerous...Worse than the rise of the Internet and other businesses, the railroads or whatever it might be.Historians could debate, but certainly we talk to many people in this field who say absolutely yes.And, you know, one of the things we encountered, and I'm not talking about like a little bit here and there from arrival.
37:04I mean, I got, you know, the better part of a dozen incoming calls from government.officials, from people at investment firms, from rivals. You talk to anyone in this industry,and they will cite this, in many cases, as common knowledge, claims that, you know, Sam pursuesminors. That's a very persistent one. And let's quickly stipulate that there's no... And we foundno evidence of this. Sam and I had direct conversations about it, and while obviously, you know,people have been telling us to take things Sam Altman says with a grain of salt, I did feel there was adegree of sincerity in some of those conversations where we would talk, you know, in addition to ouron-the-record conversations, we had, you know, frank personal conversations where I think I got a
37:45picture of his relationship with these allegations. And we put what the facts we uncover can sustainin the piece, which is we found absolutely nothing. This appears to be untrue. And it's pushed byhis opponents. I mean, we have dossiers from Elon Musk's intermediaries, in some casespaid by Elon Musk.And if Elon Musk were sitting here, he'd say what?Well, we certainly reached out to him for an interview about it, and he declined.He was busy.But we did fact-check with other intermediaries of his, and he has responses to some of the things that we say.
38:22On this matter, it is incontrovertible that Altman's rivals are pushing this and hard.Microsoft has been a huge funder of OpenAI with a lot of,of exclusive access to their own products.And just recently it was reported that Microsoft is considering whether to sue OpenAI and Amazonfor a deal that seems to go around Microsoft.Explain what this is all about because it seems like a mess to me.Well, it's one of several examples where business partners or business rivals accuse Altmanand OpenAI of making conflicting announcements about deals that they say are incompatible.
39:03Now, I will not bore you with the technical particulars of this,but essentially Microsoft is an exclusive provider of a certain kind of foundational models,and they announced a deal on top of that.Actually, on the same day, they reaffirmed that Microsoft exclusivity with Amazon,and Microsoft says this new deal, which is to do with enterprise productsthat allow businesses to build agents, they say that that depends on the very thingthat Microsoft is supposed to control exclusively.And, you know, look, they've since then released sort of mutual statements, sort of mutual.
39:38Microsoft saying, like, we gather that Open AI understands their legal obligations here.So it's tense.But you talk to Microsoft executives behind the scenes, and they say, first of all, this was absolutely in conflict.That he announced two conflicting things, that there is no way for him to achieve the thing that is promised in the Amazon announcement,which is essentially we're going to build a new solution that de-conflicts these things.Well, this also gets to the circularity of a lot of these deals.Like a lot of times a meme that you'll see when describing one of...Open AI has a deal with Nvidia, and Nvidia has a deal with Amazon, and Amazon has a deal with Open AI.
40:13People will just put a picture of an extension cord plugged into itself.Like, there are a lot of these deals that are like, I buy your stuff, you buy my stuff.Andrew, how big is it open AI?Does it make a profit?They certainly had lots and lots and lots of investing.Well, how big and do they make?make a profit or different questions.They are burning through cash at an enormous rate.
40:34I mean, so while this piece was in production, we kept having to change.They've just sustained the biggest fundraising round in history to a higher number becausethey kept closing more fundraising rounds.The latest, I think, was $122 billion.And at the same time, they're still, as far as we can tell, losing money.Where does the money go?So a lot of it is going into data centers.So they're building one.
40:58in the UAE that is going to be seven times as big as Central Parkand use about as much electricity as Miami.So this is great for the environment.Yeah, perfect for the environment.But don't worry because AGI is going to fix the environment.I feel better already.And geopolitically, it's worth noting a big thread in this pieceis, in addition to talking about some of those early ideas that were raised
41:19about, you know, pitting powers against each other,there is still a present-day reality of this computational power becoming, as Sam has said,the new currency of the world,and the factor that may shape the balance of power between nations.And so there are people within this industryand within the national security establishmentwho feel that this unrestrained and, in Altman's case,often very nakedly transactional approach
41:45to getting as much money as possible,which inevitably in the current realitymeans doubling down on Middle Eastern money,there are those critics who saythat is concentrating a new,kind of power that is incredibly geopolitically sensitive under autocracies.That may eventually be beyond our control, that you could wind up with a situation where a dictatorhas a disproportionate ownership of the most powerful technology on Earth, and that that could pose
42:16a real national security threat.But that's where the money is.And that is where the money is.And we should say it's outcome standard in Silicon Valley to fundraise from the Middle East.Hollywood, everywhere.Everywhere.But I will say even against that backdrop, there was a lot more grumbling about Altman's sweeping vision of really getting such a massive amount of money and in exchange promising such a massive amount of infrastructure in UAE particularly.Where you get your funding for a movie is different than where you put, as some people call it, a country of geniuses in a data center.
42:49In other words, if you really think that you're growing a new form of superintelligence.You want to keep it in the family.Exactly.I'm going to ask you both, as first, Ronan.I get the sense that we all know what Elon Musk at this point means and what Elon Musk wants.Sam Altman is more mysterious, it seems to me, Ronan.What does Sam Altman want?What's the grand ambition here, to be the richest man in the world, to, at some point he says that he was more interested in power than he is in money?
43:23What's your sense of it?don't think it's just a caricature or hyperbole to say,Sam Altman, very often,wants what you want in this moment.And, you know, that is the...What does that mean?That is the crux of the matter.Sam Altman, more often than not,
43:42it seems, across all of this documentation and sourcing,reads what the person on the other side of the table needs and wants,and he tells them those are the things he wants.Give me an example.well, the entire founding story of the company, that at a time when safety fears had run wild and were kind of the animating engine of discourse, he was the safety guy. And he told everyone that was his deepest belief. Now, he says, and not totally unfairly, you know, evolution over time is a real thing. And I've had to adapt to changing circumstances. But even allowing for that in both that big picture where he'sgone from being the safety guy to being the, he would dispute this, but I would say fairlyaccurate to say no regulation guy.And also in the micro picture, in all of these small interactions.
44:36You know, we tell the story of an interaction where he summons Dario Amadei, this now competitorfrom Anthropic.From Anthropic, when prior to the founding of Anthropic, Amade was a senior person at OpenAI,into a room along with Amaday's sister, who was also at the company.and accused them both of being overly political and working against him.And then they called in another executive who Altman had said was the source of this rumor.And that executive said, I never said that.And Altman said, well, I never said that either.
45:06According to an account from someone there, and we talked about all the people involved,you can see the exact reporting.But the gist is, you know, he will often in the same moment reflect the different views and desires.He is a profoundly by his own telling, conflict-diverse person.And I think the piece holds a lot of sympathy for the reality of that emotionally and the understandability of that, the relatability of that.But it also doesn't explain it away.This is a trait that is present in him to a truly extraordinary extent that has ramifications for his businesses and for the world.Andrew.
45:43Yeah, I think it's important to grasp how much of the initial pitch for this was based on this view of the world that has completely.completely flipped over time because we think of it through the lens of, sure, businesses put out nice press releases and they say, we're not going to be evil. And then they redefine what they think is evil over time. This is not that. This is, this is, Altman went to great lengths to write blog posts, to convince people, to take people out to dinner, to say, I am the person you can trust to usher this technology into existence without literally killing everyone on aearth. That was literally the pitch. So then for him now to say, we don't need to worry about all thatdoomer stuff. Like, it's kind of one of those either you were lying then or you're lying now.And I personally... Or he might argue that something happened in the research on AI that made himless alarmist. He might, but he didn't. What he says is he redefines the problem. He redefines thealignment problem away from thiscivilizational existential thing. Now
46:53he defines the alignment problem as somethingthat's annoying like, you know,Instagram algorithms that tempt you towaste your time. So he just kind of shifts the pitch.I think it'simportant to say sometimes peoplesee, you know, heavy-dutyinvestigative reporting from the New Yorker,
47:09maybe particularly my work, and they thinklike this is a hit pieceabout a villain. I would of coursedispute that characterization. We look at thesecomplex problems fairly.In this case in particular, though, this is something different than cases where we're looking at a single clear-cut criminal allocation.There are people in Silicon Valley who think of Sam Alman as a villain.For what it's worth, I emerge from this reporting, not thinking of Sam Alman as a villain.
47:34I think he is a complicated character.I think he often believes what he is saying in the moment.I think what he says about this being rooted in conflict diversion is very likely real.and as one person close to him told us in the piece,he really seems to lack any self-doubt.So that is a superpower.He believes it, I think, when he says it.And I think he's grappling with the consequences.
47:59The industry needs to grapple with the consequences, too,I think is the main case I make.I really agree.And I think one way of putting this, David,is those memos that were the initial thing that got him fired,if something simple enough,if those memos had contained a simple enough smoking gun,we would have known about it long before this.
48:19They don't contain a single smoking gun.What they contain is a pattern of behaviorthat you need a 16,000 word New Yorker pieceto elucidate, genuinely.Roden Farrow, Andrew Morantz.Thanks so much.Thanks, David.You can read that piece at New Yorker.com.
48:38It's called Sam Altman May Control the Future.Can He be trusted?And you can subscribe to the New Yorker there as well.New Yorker.com.This is The New Yorker Radio Hour.Thanks so much for listening.See you next time.The New Yorker Radio Hour
48:55is a co-production of WNYC Studiosand The New Yorker.Our theme music was composed and performedby Merrill Garbus of Tune Yardswith additional music by Louis Mitchell.This episode was produced by Max Bolton,Adam Howard, David Krasnow,Jeffrey Masters,
49:11Louis Mitchell,Jared Paul, and Ursula Summer.With guidance from Emily Boutinand assistance from Michael May,David Gable, Alex Parrish, Victor Gwan, and Alejandra Deccan.The New Yorker Radio Hour is supported in part by the Cherina Endowment Fund.
There aren't comments yet for this episode. Click any sentence in the transcript to leave one.