OpenAI vs. Sam Altman

Published Dec 5, 2023, 9:00 AM

OpenAI, which you may have heard a lot about lately, is the company that developed ChatGPT, a wildly popular AI bot which you most certainly have heard of. OpenAI’s board of directors recently purged the company’s CEO, Sam Altman, and various stakeholders – employees, investors, Microsoft – saw to it that Altman was reinstated. The board itself then faced a purge. This particular collision has it all: Silicon Valley innovation and Silicon Valley hubris, money, managerial snafus, ugly battles, promising outcomes, and, of course, artificial intelligence. AI is set to transform the world, we’re told. Ingenuity and upheaval at OpenAI offer a way for us to consider all of that. Parmy Olson and Dave Lee are both Bloomberg Opinion technology columnists.

Welcome to crash Course, a podcast about business, political, and social disruption and what we can learn from it. I'm Tim O'Brien. Today's crash course open Ai versus Sam Altman. Open Ai, which you may have heard a lot about lately, is the company that developed chat Gpt, a wildly popular AI bot which you most certainly have heard of. Open AI's board of directors recently purged to the company's CEO, Sam Altman. High drama ensued, and various stakeholders, employees, investors. Microsoft saw to it that Altman was reinstated. The board itself then faced a purge. Oh my, this particular collision has it all Silicon Valley innovation in Silicon Valley hubris, money, managerial snafoos, ugly battles, promising outcomes, and of course, artificial intelligence AI is set to transform the world. We're told ingenuity and upheaval and open ai offer a way for us to consider all of that. So I've invited parme Olsen and Dave Lee onto crash Course to help us outline the lessons of this particular tale. They are both Bloomberg Opinion technology columnists. Parmey It's based in London and Davi is in New York and they bring a wealth of experience and insight to today's show. Welcome my friends, Thanks him, thank you. Let's start with Sam Altman himself. Eric Schmidt, Google's former CEO, recently compared Altman to Steve Jobs and said Open Eyes Board was wrong to can him and I quote Schmidt here, these founder CEO types are unusual, they're incredibly valuable, and they change the world. Schmidt said that at a recent Axios conference. Dave, what do you think of that observation about founder CEOs and the sort of idea that there's a unicorn executive out there who is indispensable.

I think, look, that's always been Tilicon Valley view of you know, great founders and what they can do, and this idea that you should follow a founder's vision rather than say put it to committee. And you know, Steve Job's obviously the most famous example of that. Kicked out of Apple, returned again around a decade later, and of course, you know, the rest is history, and what happened with Apple surely wouldn't have happened without Steve Jobs coming back. And we've seen this since again and again, whether it's you know, Mark Zuckerberg or elor Musk or any sort of tech founder, even people like Travis Kalonick, you know, who founded Uber. There's this sense that there's this founder skill, this magic, this uniqueness that should really be sort of listened to, and although it can be questioned, you know, there's a sort of bias towards always backing the founder. Speaking of Sam Altman, of the many comments made about Altman over the past week or so, I think, you know, Paul Graham, who founded why Combinator, where Sam Moultman sort of made his name, said a quote that kind of rang out around everyone, and he said, you could parachute Sam into an island full of cannon, come back in five years, and Sam o wont would come out as king. And I think that's kind of the view of him in particular, and it does sum up, as you say, this culture of revering founders like life.

I don't know actually how to interpret the idea that you're going to parachute him onto an island full of cannibals and it'll come out the winter, because that kild speak to actually just ambition and being extremely as they say in Silicon Valley, goal oriented as opposed necessarily having innovative genius, and I think, you know, ideally you want a combination of both if things are going to get done.

But you know, if I can just jump in. It's interesting you mentioned Paul Graham, Dave, because Paul Graham, I think, is part of the reason why this mentality, this almost doctrine exists in Silicon Valley. When he started White Combinator, he was like a guru among startups, and his teaching, as it were, was that the founder is all that matters.

In a way.

The technology doesn't matter what a startups technology would it has under the hood. What matters is if the founder is a visionary hacker to hype who wants to build an empire, and if you, as an investor can find that kind of person, then you should give them free rein. The board should give them free rein, and they should be able to do whatever they need to do in order to build the empire. And that's why in most cases, someone like Mark Zuckerberg has majority voting share the board does whatever he wants pretty much, even through the most controversial times. Google's founders also had a lot of control of the company, outsized control, so I think a lot of it comes down to this kind of ideology that Paul kind of created and Sam perpetuated when he took over as y combinator head.

Steve Jobs famously had Bill Campbell as a management coach, and Bill Campbell was sort of another one of these executives who in a different era was a guru to various Silicon Valley titans. But I sort of wonder if the Bill Campbell advice managerial advice in the Steve Jobs era was qualitatively different than the advice that Paul is giving a Sam Altman right now. Is that the case Parmi has there been an evolution and the kind of advice that nascent Silicon Valley titans are hearing from their advisors.

I have to admit I'm not familiar with the advice that Steve Jobs got directly, but I think maybe there is a questioning more and more of whether these founder types should have all that control. That may still be the case at Facebook, but I mean, this is what made open ai so unique, right is that it had a board that didn't just have teeth, it actually used them, and it used them to fire him. So I don't know if that's necessarily changing, but it certainly was very, very different to open ai.

Talk a little bit Parmi about Sam Altman's foundational experiences. Prior to coming to open ai. He had been an entrepreneur, a young computer with and then lands essentially at y Combinator and ends up running it. Yeah, or his bona fides before he came to open ai.

Well, he paid his dues as a startup founder who created a failed startup, as so many startup founders do. He created a startup called looped, worked on it for a number of years, devoted his life to it, and then eventually sold it to another company, which you might say as a success, but in some ways it wasn't seen that way. And then he went on to become an advisor to Why Combinator, which is like basically one of the world's the world's most successful startup accelerator of all time. You've had companies like Reddit and Stripe and Airbnb all came through that, And you know, he became known for having this real ambitious view and pushing startups to be more ambitious. So he famously said to the founders of Airbnb, you should change the numbers in your pitch deck to investors from millions to billions, and then he kind of went into this stage where he got really into futuristic technology, and he was kind of gazing into the future with things like mind uploading and kind of uploading his consciousness into cloud computers and exploring the universe he's talked about up publicly. And he then started investing in very futuristic technology like nuclear fusion, life extension technology, and he pushed y Combinator in that direction, But his heart really was in AI when he had apparently he had this kind of realization at some point in his late twenties that humans aren't really all that unique and special, and our brains can be simulated by machines. And if that's the case, we can build a machine that is as unique and powerful as the human brain. And that's what he set out to do. But with this extra kind of level of idealism around spreading all the benefits of that to humanity.

Well, we could go down so many roads here. This reminds me of the Silicon Valley chatter around the Singularity that the Google guys were so fascinated with. It's interesting to me that if you get enough technological success the next step in your thinking is how do you perpetuate yourself forever inside of a machine that Maybe that's a conversation for another day, Dave. You know, the idea of highly tailed motivated people being given free reign so they can be creative isn't limited to Silicon Valley. Obviously. There's a long history of it in the business world, you know, going back to the Rockefellers and Henry Ford and in completely different industries. And it's in the arts, right. We know that film directors and painters and others are sort of seen as masters or mistresses of their own destinies and people shouldn't intrude on that. Does all of that fairly accrue to Sam Altman in this case? Is he someone presiding over a company who needs a board to support him but not get in his way.

Well, I think it's interesting, you know, thinking about Altman's upbringing, I guess through why combinates, because what that process did, I think is gained him a lot of allies and a reputation as being, you know, this real sort of network guy in tech, and I think that has allowed him to move into these future roles with a certain level of trust. I thought it was really remarkable. When we were hearing just about being fired, one of the people to immediately sort of go on to Twitter was Brian Chesky, the CEO of Airbnb, which was a Y Combinator company, immediately jumping to sam Untiner's defense and almost kind of reporting on the situation himself, which was bizarre. So I think, you know, although these founders do get this kind of reverence, it does have to be earned, and I think the way that Sam Oltman has earned that is not so much to be seen as this kind of visionary genius necessarily, but this kind of smart networker, this smart sort of bringing together of people. And the thing that made Y combinators stand out was this really fast acceleration. When companies would join y Combinator, they had to say a particular metric that they were going to grow, and every two weeks they'd checked that it was you know, two x four x and so forth. And I think that's the ethos that Sam Wontman has brought. Now the question is, you know, was that something that needs a board to rain in or is that something that a board needs to sort of sit back and say, okay, off you go and you know, I think sam Moultman must have been mindful of the reputation of founders, particularly people like Mark Zuckerberg, who you know, the board of Facebook could never fire Mark zuckerbo they don't have the power to do that. I think he was mindful of how that is seen, and I think he was trying to make a statement by saying that not only did he not have any financial stake in open ai, but he was serving as the behest of the boards. As it turned out, his reputation ended up being more influential than the votes of the board open AI. So even once they made that decision, the rallying and support around sam Oltman, thanks to his reputation within the tech industry built over the past several years, meant that was a bigger protection for him than having board votes would have been anyway.

Did you want to jump in there, Permie, Yeah.

And I think that's a great point about the amount of protection that he has. Silicon Valley is such a bubble, and if you're a founder, if you're successful and you've made a very valuable company, and you're very wealthy and you're a billionaire, that you have to behave so badly to fall from Silicon Valley's graces, and Elon Musk is a prime example of that. So I Meantman's return as CEO as a prime example of that. But I just wanted to go back to one point you mentioned, Tim. You were kind of casting our minds back to the Rockefellers and people in the arts who have this kind of power as founders and as leaders. But I really think that in this situation, someone like Sam Altman and the leaders of big tech companies are so different to any of that, just because of the scale and wealth and resources that these companies have, which are unprecedented in history. I mean, Google today reaches something like four billion users worldwide. That's half the global population. You know, although Chatgypt has something like two hundred million regular users. I think Microsoft services, which are going to use more and more open AI technology, touch more than a billion people around the world. So we're talking about a few people, these founders, with an incredible amount of free reign, who have like an incredible global influence at the same time, more so than any empire even in history.

And now, of course, the product they have in hand is a revolutionary and transformative product. With Google, you know, in its pre AI phase, it was a ubiquitous search engine that sort of opened the world up to its users. No one really worried about it taking the world over. You know, it was an information provider. It could be a you know, assessed pool of disinformation too. And we'll get into this morning. I think our conversations, so I think this particular product AI and technology is shaping people's views in different ways as well. But Dave, tell me what you think.

Well, I think is also worth stressing in terms of Oltman's value to open AI. You know, this is a company that people are talking about being valued at around eighty six billion dollars, and you know, a lot of this is because of the personality of Sam Oltman. The value of open II comes from the fact that for the past year, every week or so, it seemed like Sam Martman was shaking hands with a world leader. It was sitting there in front of Congress talking about AI, talking about regulation, and so there is value. There's clear value of what Sam Mortman is to open AYE. And I think it's reputation and the deal with Microsoft and the fact that it was and should still be positioned as the front runner here is down to the fact that he has this ability to communicate what the company wants to do clearly, and that doesn't necessarily come all that naturally to other people in Silicon Valley. So there is an intrinsic value in him himself well, and.

You know, all of the Steve Jobs comparison, Steve Jobs was not an engineer. Steve Jobs wasn't an inventor of technology. That was Steve Wozniak, his early partner, and Jobs was the person who was capable of verticulating the kind of Nirvana vision of a laptop and every desk and then a mobile phone in every hand. And he was a good manager, and he was good at ultimately at sort of harnessing creative people, and that is a skill as well. But what still feels very mysterious to me, you guys, is why he got pushed out to begin with. Correct me on this if I'm not up to date, But at this point my understanding is the only kind of public reason that's been given for why the board pushed him out is that board, in their own words more or less, and I'm synopsizing, felt that he wasn't being candid with them in his communications with the board, and there is now I think an Altman commissioned investigation into what led up to that. It's being led by Larry Summers and some other board members, which also kind of speaks to whether or not that actually be an independent investigation. But anyway, how did we get here? Why did the board end up feeling concerned about this wonder kid in their.

Miss Well, I think, first of all, no one still knows, apart from it seems the people on the board were, and possibly Sam because given a couple of his first public interviews, he will not answer questions about why he was fired. But I think the reason because it's all of speculation now, but it seems like the reason kind of goes back to what we were just discussing about the outsized influence and power of these tech founders. It goes unchecked in most cases, and from what people have been saying and reporting, the board uneasy essentially with Sam's behavior outside of open Ai and some of these ventures that he was pursuing. So for example, the iris scanning company world Coin, which aimed to scan the eyes of billions of people around the world in order to identify them when the Internet becomes flooded with bots, and also to help distribute the trillions of dollars of wealth that Altman believed would come about through the attainment of AGI. He was also pursuing, according to reports, creating a so called quote iPhone for AI with Johnny Ive who was the former lead designer at Apple. And he was also on top of that, reportedly talking to Middle Eastern sovereign wealth funds to raise something like ten billion dollars to build a chip company. This is a lot of fingers to have in different pies, and I think it sounds to me like the board was just overall kind of uneasy with where he was going with all these different ventures and whether he was going to use open AI's technology for any of that. So when they say he wasn't candid, it seems like they were kind of caught on the back foot with some of this stuff, and they didn't feel like they were performing their fiduciary duty to humanity, which was actually their mission as the board.

Yeah, I want to get into some of this in more depth, you know, specifically this particular board's unusual mission. But one of the other things that's been out there, Dave, is it's been reported and then dismissed by others that there was this mysterious Project Q being hatched inside open AI that maybe ran a foul of this idea of AI being harnessed ultimately for the good. It's unclear to me but that the board had concerns about this product. Others have said, actually, there were no letters written about it. So the reporting has been very murky around this, and I don't know that we should rely on it, but it gets to this issue of the board being concerned about, you know, the idea of bots controlling the world and open an AI losing its mission along the way.

Right right, I mean the ultimate endgame concern here is that an AIO created that poses a real risk to the human race, and any step along that journey makes people pretty pretty nervous. Now, the question around whether this was a factor in Oltmand's firing, as you say, that still mrk it at this point, and I think actually it sort of stresses this issue, which is that had the board that forced Oltman out, had it had reasoning and public explanations for why it did, that we could have avoided what's now going to be, you know, this endless filling of the void with possibly conspiracies of other you know, sort of hyperboles around what open ai is developing. Now. Q Star, this sort of rumored project that supposedly can solve basic math problems. I mean, for starters, calling it Q anything seems like a mistake. It's very bonred like, I mean, it's remarkable. And just days later we Amazon named one of it's a's Q as well, which I just find staggering that these companies would be so sort of mindless in doing that. But that's another that's another matter altogether.

It's like naming something X well, right.

Yeah, I mean you might as well call killer AI. I mean, it's just going to make people sort of nervous or just conspiratorial, which you know, we have enough of that right now in the world, of course. But look, so the issue is, you know, it's very hard to sort of sometimes for people who aren't experts in AI to draw this line between something that might be able to do basic math and something that might destroy the universe. But supposedly there was some concern within open ai that this was you know, a significant break through, a significant step, and you know, speaking of x Elon Musk was in New York recently talking about the firing of Samultman, and he made what I thought was unusually lately a pretty reasonable point. He was saying, Look, either the board needs to be clear about these concerns and allow the public to know what they were, or if there wasn't a good reason, they need to you know, well they have stood down now because then you know they needed to be accountable for that error. But I think either way there needs to be a feeling of this vacuum around what exactly the concerns may have been. Because still one of the big unknowns in all this is, you know, one of open AI's co founders, Ilias skav who was the chief scientist open AI and was on the board, and he was the sort of pivotal changed mind on the board that meant the board had a majority to fire Altman. We don't quite know what it was that concerned him so much or why he changed his mind. I mean, he's subsequently come along and said he regretted the decision, and Sam Moultman, in part of his statement about returning to the company, Sam Moltman said that he harbors no ill will towards him, And you know, I've never heard that phrase said with much sincerity. So I'm kind of keen to see where he lives. Yeah, I mean it's like a football owner saying they backed the manager.

You know.

It's sort of ominous sign that things might not be so well. And so this is one of these details that we need to sort of hear more about. And I think open AYE, if it wants to be at the front of talking to governments and being this hugely an entry company, the trade off has to be that it is more public, perhaps than tech companies would typically be in the past. It doesn't get to hide behind this super secrecy that Silicon Valley is more comfortable with. It has to tell us what was going on here.

We should also know. And Elon Musk says he was an early board member at opmen AI along with Sam Altman. Before Sam became CEO, he was on the board and as Parmi noted earlier, Elon has run rough shot over his own board. We're going to take a break in a sec But this idea that AI is going to take over the world and that open AI is sitting on top of a Pandora's box, is that warranted. Parmi.

First of all, I would say an answer to that question that I think open ai secretly loves that people think that because it makes their technology. However, but they keep it under wraps, it makes their technology seem that much more powerful. You know, for people to fear it doesn't mean they don't want to pay for it necessarily. Companies kind of see that almost as a kind of nice attractive feature to software. But I think I would say that with something like this new development with q starr, for example, that open ai is working on this ability to do great school math. You so this means that AI can reason, because when you do math, you have to figure out steps to solve a problem, and it's a little bit like solving problems in real life. And so right now, chatgybt is amazing because it can generate language, it can generate text, and a lot of that is just statistical based predictions. But if a model can do things much more strategically, if it can plan, if it can solve problems, many people say in the AI field that this is a step towards more general intelligence. And the reason why I'm saying that in response to your question is because although I don't think AGI is necessarily going to take over the world. I personally don't subscribe to the rogue AI theory, certainly not anytime soon. But this idea of humans outsourcing more than just tasks, but actual responsibility to AI is certainly going to be happening in the next couple of years with these kinds of developments like q Star, and also Google's new model called Gemini, which hasn't released yet, also has the sexpertise and strategic planning, so it's kind of similar to what open ai is working on, and I think that's going to be really interesting to see how business leaders, how anyone uses these kinds of systems not just to generate text or to ask advice, but to actually carry out some of our day to day work tasks and responsibilities.

On that note, we're going to take a quick break to hear from a sponsor and then we'll come right back. We're back with parme Olsen and Dave Lee, and we're discussing all of the recent upheaval surrounding open ai and its impact on its product chat GPT. Dave sam Altman was the CEO and on the board at OpenAI. He's now just the CEO. How do we interpret that? Did he win the power struggle that ensued after he was forced out and brought back. Does he have more influence now? Does he as less influence? Can we even know?

Yeah? I mean, look, I think a lot depends on a couple of things. I mean, one of them being how does the rest of this board look like in the coming weeks and months. You know, the interim board of three is going to expand possibly through as many as nine or they haven't quite confirmed the shape up of that, and there's going to be lots of sort of questions that's about what the makeup of that board's going to be. I mean, already the board has lost its only women. It's now an all male interim boards, So there's going to be pressure to reflect a diversity there. But then also there's going to have to be people who are incredibly versed in safety round AI, around policy and AI, people that are going to be at least seen as potentially being able to hold someone like Sam Altman to account or at least sort of reign in perhaps some of those commercial interests that might sort of provide conflicts or just kind of have him going full steam ahead. So I think it we'll know more about sam Altman's power. Once we know more about that board.

Party you were mentioning earlier, how sam Alton was allowed to pursue all these other ventures outside of open Ai, including companies that possibly would have ended up as competitors getting funding for those rather than possibly getting funding for open Ai, just all of the myriad financial and professional conflicts. One of the other things that I find very strange in the story is that open ai is a nonprofit company housing a for profit subsidiary that offers a product Chat GPT for free unless you want extra Chat GPT features, and then as a consumer, you have to pay extra to get those. This confuses me, even in that structure at one point, the company at evaluation of about eighty six billion dollars, isn't this structure in it of itself bound to cause problems.

One thing I underestimated about this whole story when it happened was how much power this board had, And I underestimated and didn't fully appreciate the fact that they really could fire sam Altman. But the idea of trying to combine for profit the nonprofit is not completely unusual In Silicon Valley. Mozilla is connected with the Mozilla Foundation that is a nonprofit, Signal, which is the encrypted messaging app that is also a nonprofit, and even just recently, two of the leading AI firms in Silicon Valley. One is called Inflection, the other one is Anthropic, the latter of which split away from open Ai a couple of years ago. Those are also trying to thread this needle between being businesses that make money but also building AI that is safe and beneficial to humanity, and so they are also tinkering with these kind of unusual corporate structures, like being a public benefit corporation. I think in the case of Inflection, they're structured in such a way where they have to prioritize the environment and consumers on the exact same level as their investors. So profit is not the number one priority for them. It has to be upheld with these other things now, and this is clearly very difficult to pull off open Ai. It doesn't seem to have worked for them. Deep Mind, which is the big AI lab that's part of Google, also tried to do this. They tried to spin out of Google for several years. They tried to create a governance structure called a Global Interest Company. They wanted to be like a nonprofit style company. It totally failed. Google next it. So there's this history of AI builders trying to do this because they know this technology is so transformative and they don't necessarily feel comfortable even with it being controlled by monopolistic corporations essentially, but it's been very difficult to figure out how to make it work.

I think I was sop on. Isn't it the case that this model of a nonprofit running a technology product, it comes under a lot more strain given the sheer amounts of money necessary to pull it off. With aim absolutely eating power Microsoft place. I think it was up to thirteen billion dollars to have open ai use its servers to do the crunching that makes AI possible. So, you know, Mozilla is a great example there of an organization that runs a browser, but creating and running a browser isn't in the same league as having to build cutting edge AI. And even in the case of Mozilla, you know, much of their funding comes from companies like Google and others. So it's always been a bit uneasy this tech model with nonprofits, and it's been particularly strained when it comes to AI just because of the magnitude of what needs to be done well.

And you also have, you know, a chief executive whose mission is to grow the company and increase profitability and expand the share of the product that's selling. And then a board which is invested with responsibilities for helping to achieve that, but also being kind of a self regulatory device that is trying to stand in the way of abuses. All of it, Parmi was saying, And you know, I think this is the historical tension anyway between boards and executives. It happens in lots of industries. Boards came into being to sort of make sure, especially publicly traded companies, that investors interests were being looked after and that you didn't have rogue CEOs. And there's a lot of situations even outside of tech, where you can have a rogue CEO. But of course, and again, as Party mentioned earlier, we're talking about a product and a technology at play here that scares people. You know, the board, Parmi continues to interest me because, as you've noted in one of your columns, open Ai had a chance to get more women on its board and it just drove right past that, which now has an all male board. One of the people on the board that was reconstituted, Helen Turner, interested me because Helen Toner is an academic. She has done a lot of work herself on AI and the uses of technology, and it has been reported that she had worked on a paper that raised some questions about even whether or not open AI's own products could be problematic, and that Sam Altman confronted her about that and they had a dispute. She was a pivotal person in his departure based on the report we've seen. Tell me about those two things, about the fact that we have an all male board there now and what happened to Helen Toner.

Well, I'll start with Toner, and absolutely right. She did write that research paper. You can see it. It's public and has written in association with Georgetown University, and she's scathing about open Ai. She referred to frantic corner cutting in the progress of building chat Ept, and then she compared open Ai with its rival Anthropic and said that Anthropic, which was the group of people at open Ai that broke away to make Safer AI, had done a better job of more slowly deploying a product, making sure that it was safe before they put it out into the wild, and sam Altmand apparently was livid about that. Didn't like the fact that she was putting that into the public domain when you know the open ai is being investigated by the Federal Trade Commission for potentially infringing on people's privacy. He told her, according to reports, that it was compromising the company, and so, yes, there was a lot of tension there, but I think tim, I mean, you could look at it both ways. Yes, what she was doing was compromising open AI's reputation, but she was an academic. They knew that when they brought her on the board. And as you also mentioned, this is kind of what boards are supposed to do, push back a little bit, and boards typically have a fiduciary duty to shareholders. This was not the case with open Ai. The board had a fiduciary duty to the open ai mission to benefit humanity. It's kind of, if you think about it, quite a bizarre phrase. Feels weird saying that, but that is literally how it was written and signed and agreed by everyone. But Toner was the one that lost her seat, and so was Tasha McCauley.

It's smacks of Sir Gey Brand and Larry Page do no harm in their early Google literation.

Absolutely yeah. And Google also famously said in the very beginning, we don't want advertising, we hate advertising. And now they're the world's biggest ad giant.

So there you go, Ah, the children, but we need them among us. Whatever the merits are of how the board intervened here, and I have sympathy for a lot of them, it also had quite the clown show aspects to the whole thing. You sort of wondered why this could have been choreographed or sorted in a more private way, but maybe that was impossible. Here is there another way this could have played out at open AI?

I mean yes, And the way it could have played out was the phrase that was bringing your receipts. If the board had this feeling that there was issues with how Sam Wontman had been communicating with them, it needed examples. It needed a way to sort of explain its thinking in a way it just hasn't done. It kind of reminds me of when you have a sort of argument with a spouse and you kind of well, give me one example, and you go, well, I can't think of one right now, but you never quite get anywhere, do you, And nobody gives any concession on either side, and that's kind of where we're at with Open AI. It was being described as a coup, and I think that's kind of accurate, and that there was a sort of swing for the king and you've got to have everything in a row and if you swing, you better not miss, and that's what they did. They swung and they and they missed because they didn't have the reason to back it up. And you know, as we've now discovered as well, they were fighting the forces of the commercial interest and open AI, so you could argue they perhaps never stood a chance. But I spoke to one the business professor recently who made the point that talking about this sort of nonprofit management and so forth, we shouldn't necessarily write off this model because what this could have just been was in competency in this board in particular when it came to carrying out this coup or whatever we want to describe it.

I completely agree with that, and I think that's what's such a shame about the whole thing, because it does sound like they did come across as quite incompetent, and if you look at them they didn't have a lot of experience on other boards. So if they had just executed it a little bit more professionally, then I think people could maybe consider this kind of model as potentially working. It's not a terrible model. I mean the idea behind it is quite noble when you think about it. It just couldn't work.

There's the thing, you know, talking about the paper from Helen Toner. I mean, I think it's worth remembering that that paper, like you say, it was written, it was public and Helen Tona was still on the border Open AI. So that was an indication that that dynamic did work right. I mean, it created tension, but ultimately the governance was still in place. It was just this sort of follow up move that caused all this chaos.

Let's take another quick break, my friends, and then we'll come right back and get into our last act here. Today, we're back and we're chatting about Sam Altman, Open Ai, chat, GPT and the AI Revolution with Parmie Olsen and Dave Lee David. It's not clear to me that open AI's management problems are solved and they're not operating in a protected market as it were. They have a head start they have a product that is early out of the gates and much beloved, but they have to deal with a lot of competitors who are in this market too, And do you think that their management problems are in rain enough that this episode won't affect their longevity.

Well, it's interesting. I was the other night and they just happened to be a fairly senior person open aiye there, and they said to me, you know that prior week when all this was up in the air, Yes, it was disruptive. They didn't know whether they were going to be working for Microsoft in a week or any of these other sort of different outcomes that we could have seen. But as of sort of Monday and Tuesday, the week after, things were basically back to normal and people were coding again, people were planning again, and it's almost as if nothing is that had ever happened. Now, of course, a big thing had happened, and we've yet to see how that fully shakes out. But I think given where they were at on that Friday evening when this announcement was made and the weekend of complete madness, I actually think it's been a pretty impressive sort of pulling back into just sort of normality, So we'll see whether that has any long standing impact. I think you know those other companies, Google in particular, part Me mentioned Gemini earlier one of their AI projects that's delayed for various reasons. So a company like Google's probably thinking, oh, wouldn't it be good if open Eye was slowed down a little bit because we could do with catching up here. So I think competitors were mildly delighted to see that, But I still think open eye is going to be considered to be the leader here still, and that perhaps wouldn't have been the case had they successfully asked sam Oltman Harmy.

It's been almost exactly a year since chatchept was released publicly. What has open ai done since then to protect and expand its franchise. I think chatchipt can accept more queries. It's not only techt base. Its database is more up to date. But as you've noted, also, mistakes have bound still when you use the product. So how do you see the evolution of the company over the last year?

Oh, over the last year? I think really the success of Chatchipt, first of all, took them by surprise. They were taking bets internally about how many people would use it within the first week, and the top bet was one hundred thousand users and it ended up being more like a million. So I think, really the past year has just been open ai scrambling to keep up with viral success that it absolutely did not anticipate. Having said that, it has from a business perspective, done incredibly well to keep up with and satisfy the demand for people who want to use this, whether you're a consumer who wants to spend twenty dollars a month on chat GPT plus, or a company who wants to get access to open AI's software and technology, or a company that wants to access it through Microsoft. And I think that's really where open ai has real stability is through that partnership with Microsoft. Because the thing about Microsoft is that it's big product through which it dispenses open AI's technology is cloud. It's known as the Azure platform, and it has something like eighteen thousand customers and their big names, you know, like Rolls, Royce, Adobe, the Seattle Seahawks, just all sorts of random big companies, And the great thing for Microsoft is that these customers are locked in to this product because it's actually really expensive to extricate all your systems from what one cloud provider like Microsoft and move on to another cloud provider like Amazon. The customers and IT professionals hate this. They really don't like the fact that they can't shop around. But this is really good for Microsoft and also very very good for open Ai. So this past year kind of integrating themselves more and more and serving their top investor, Microsoft, who owns forty nine percent of open Ai, has really stood them in good stead, and it certainly did over this past weekend because part of me wonders would Altman have been able to be reinstated if he didn't have such an Adella pushing so hard for.

Him well, and that brings us, Dave to the question I wanted to ask you about Microsoft. Given that Sacha Nadella, Microsoft's CEO, was very enmeshed in the open Ai upheaval, he almost got Sam Altman and many of his employees for a song. It probably would have been one of the most epic, sort of cheap takeovers in corporate history if that had panned out. Nadella has referred to Microsoft as being the co pilot company now, like the company that makes products that sort of assist you in having a better corporate liver a better personal life. And obviously chat GPT fits right into that. Microsoft is a player here, and it's a heavily influential player both in this drama and where I think open ai is going to go correct.

Yes, I mean it's not just a players, you know, as the player externally. I mean there are other investors in open ai, but it's only Microsoft that really has the power here. I think, you know, Palme was touching on a really important point there, and one that I think's worth remembering is that the winner in AI isn't necessarily going to be the company that has the best AI or at least the most sophisticated AI. It's going to be the company that can put that AI into applications that consumers are either already using or that they want to use in future. And so when you have Microsoft on board, that is incredibly powerful. Because people have been using Word for years, people have been frustrated at PowerPoint for years, Excel, you know all these things that people do, you know, Search changes, and people laugh about Microsoft and Clippy, the little the little mascot that used to have and that sense you you know where we sort of heading again forgotten all about Clipping. I love Clip was he was underrated. But this is where we're going again. And I think that's why this deal is so valuable to both open I and to Microsoft, and it's why Satching Thedella leapt to action to step in to stop his investment in ape and ai going up in flames. That when Microsoft said, you know, we will take on and we will salary match as many open ai employees are interesting in joining us. I mean, that's a huge thing that was still I saw one estimate that was going to cost around a billion dollars to do that. Now that's arguably getting a eighty billion dollar company a huge sort of black fried a deal. But that's a huge amount to add to your Headcunt. They weren't even sure where they were going to put them. They were clearing out desks at an old office that used to be used for LinkedIn, which one of the companies that Microsoft owns, and so this would have all been very very very very difficult for Microsoft. What they've come to instead, I think is their sort of preferred option. Sam Oltman back in the job, and also on this new board that's being made up. They have an observer seat, so they can't vote, can't have a say in the decisions, but can observe the board now, I asked, at the time we're recording this, we're still learning about what exactly that means. I asked Microsoft. They weren't sharing any more about what exactly they'd be able to observe and what imput they could have or indeed who at Microsoft might take that position. But Microsoft has gone from a situation where it had a sort of very much outsider perspective on open Ai, even though it invested more than ten billion dollars, to having a situation now where it's got a much closer relationship and we'll get a heads up on any crazy decisions like this future. One of the things that was gobsmacking about how the Sam Altman firing went down was that Microsoft learned just about the same time I did when the blog post went up. There it is, and that given the reliance that Microsoft has on open Ai and its importance to Microsoft's strategy.

I mean, it's a major investorable thing.

Yes, absolutely, and typically your huge investors, part of that investment would mean getting a board seat, getting two board seats maybe, but such as the popularity of open ai and also were stressing. Is one of the reasons why Microsoft was something of an outside is because it worried that being too close to OpenAI would raise some level of scrutiny from competition regulators, sort of wondering, you know, it is Microsoft going to flex its power to stop others getting access to good AI. So there's a lot of factors at play, but in sure, I think Microsoft has found a scenario that suits it very very nicely.

Indeed, Yeah, And I would just add that if it had been a subsidiary of Microsoft, if Sam Altman had been part of this advanced AI group, that's not what Inodella wanted at all, because that would mean Microsoft takes on all of the reputational risk, all of the legal risk for all the crazy stuff that open ai regularly does, like putting chatchypt out into the world. Microsoft would never do that because copywriter issues for a start, and crazy things that this bot could say. That's why big tech companies don't do the kinds of things that open ai does. So it's much better to be an investor take on all the glow of open ai and none of the liability.

Which saw a sign off that didn't we part me a few years ago, which I'm sure you remember, when Microsoft had an AI bot that it kind of released onto Twitter. I believe it was called Tay the Box. Yeah, and almost immediately people realized that if they said certain things to it, it would repeat them back to them, and it became sweary, it became racist, and it was just this is just a tiny, tiny experiment, but it caused Microsoft no end of reputational issues and things to be accountable for. So you can totally see why they want to be outside to some degree.

Saw people still cringe at that memory.

And I think Microsoft still licking its federal antitrust investigation wounds of the nineteen nineties. That makes it a little as as I put it earlier, absolutely has it meant to appear to be the titan in this nascent.

And then these who were dealing with that in the nineties, like Brad Smith the Legal Council are still there, like they all remember this stuff. So yeah, and satchin Adella two, I.

Want to ask you each about lessons that you've learned from this dramatic little moment in Silicon Valley. I like to ask people that at the end of the show. So let's start with you, Parmie. What is something you've learned watching this recent debacle that you didn't know before it happened.

Hmmm, it's a good question. I guess I would just go back to my answer previously about the fact that a board like this could actually topple someone like Sam Altman. But I suppose the other thing is that this isn't so much something that I've learned, but just a suspicion. The cynic me has been confirmed that Silicon Valley will Silicon Valley, And even if you set up a board and a governance structure that looks good to everyone, that looks like it's there to help humanity, it will still ultimately serve the purpose of its investors. And that is precisely what happened after this dramatic weekend where Sam was ousted and brought right back in. So a little bit of hypocrisy there, and also just Silicon Valley being Silicon Valley.

Dave, I don't know, can you still learn things? Both of you know so much that I sometimes wonder.

Learning every day I've learned Listeners to parme me on this podcast and I'm glad you let her go first. I can have a moment to think about this, because it's interesting because on some levels it feels utterly predictable this happened. And I think one thing I've learned or was surprised by, was I think we always knew that at some point there would be some friction around the direction of one of these companies. I'm surprised it's come so soon for open ai I thought we'd be more sort of tangible threats or fall out or crisis that we'd all sort of know about that would lead to a moment like this. I'm surprised it's come sort of from what seems to be more just sort of human personality clashes than anything else. So that surprised me. The frigidity is surprising. But I think that the lesson, if we can take one so far, and I think it's yeah, we're still learning the lessons. I think the lesson is that open aiy is a tech company. It's a tech company like Facebook, to tech company like Google's the tech company. And I think I'm going to constantly remind myself of this sort of this week and remember that ultimately it was the needs of the tech industry and the money of the tech industry that essentially had the final say. And so I think open AI. Although we can talk about governance and nonprofit status, I think ultimately it's still a tech company and we should cover it as such and think about it as such. So that's my lesson from this.

I think we are out of time. Prime and Dave, thank you so much for coming on today.

Thank you.

Thanks.

Tim parme Olsen and Dave Lee are both columnists with Bloomberg Opinion. You can find their work on the Bloomberg Opinion website and on the Bloomberg terminal. Here at crash Course, we believe that collisions can be messy, impressive, challenging, surprising, and always instructive. In today's crash Course, I learned that no matter how talented or how smart you are, no matter what industry you're working in, even if it's in Silicon Valley, humans can be humans and they can do crazy, crazy things. What did you learn? We'd love to hear from you. You can tweet at the Bloomberg Opinion handle at Opinion or me at Tim O'Brien using the hashtag Bloomberg Crash Course. You can also subscribe to our show wherever you're listening right now, and please leave us a review. It helps more people find the show. This episode was produced by the indispensable and decidedly non crazy Adam Azarakus and me. Our supervising producer is Magnus Hendrickson, and we had editing help from the Sage Bauman, Jeff Grocock, Mike Nitze and Christine Vanden Bilart. Blake Maples does our sound engineering and our original theme song was composed by Luis Gara. I'm Tim O'Brien. We'll be back next week with another crash course

In 1 playlist(s)

  1. Crash Course

    66 clip(s)

Crash Course

Hosted by Bloomberg Opinion senior executive editor Tim O'Brien, Crash Course will bring listeners d 
Social links
Follow podcast
Recent clips
Browse 66 clip(s)