Lenny's Podcast: Product | Career | Growth
01:14:34 · Eric Ries
And they had to defend it for two years because they didn't actually implement what's called now the long-term benefit trust until their series C, but they had the right and the intention to do it in all their legal documents from inception. That was a really important choice. And so even today, Anthropic has directors on its for profit board who are appointed by and are accountable to an outside group of trustees who are AI safety experts who do not have equity in Anthropic. So they do not have a fund- The experts who do not have equity in Anthropic. So they do not have a financial incentive. In its growth, they have an incentive to see it done properly. So whenever you see Anthropic, do the right thing. Like when they refuse to release a model because they think it's too dangerous, think about how much that's costing them. People say, "Well, they do it for the publicity," but publicity's nice, but you know what would be really nice? Is having the number one top model that everyone has to pay you to use. Okay? That's really nice. People say that they got into this fight with the Pentagon. And to be clear, I don't think that they're even the primary actor in that story. Okay? This is a story of government overreached and they were an impossible situation. Even people who say they did the wrong things, when you're an impossible situation, what are you going to do? It's the Kobayashi Maru. What are you going to do? You're in an impossible situation. There's no right answer. But even people who think they did the wrong thing, tactically speaking, admire them because we live in a time when it's so rare for companies to turn down money ever. And here they turn down... Say what you want about them, but they turned down a $200 million contract and bore the wrath of the world's largest army and government. Okay? That took a lot of courage. Part of that courage is enabled by the fact that they have this structure and investors can't just oust Dario on a moment's notice. I think the structure, frankly, is better than founder control. Dario does not have dual class shares the way that Mark Zuckerberg or Larry and Sergey had. It's more institutional in its nature. But to kind of answer your question, the critical thing we need, if we're going to really resist outside pressure, we need what I call a mission guardian.
01:16:30 · Eric Ries
It has to be somebody or some entity's job to make sure that the thing remains mission locked or mission aligned. That does not happen by accident because gravity is such a powerful force. So Anthropic's solution to that is to have a mission guardian in the form of the long-term benefit trust. OpenAI very famously had the nonprofit foundation, although now they've converted to a public benefit corp structure. Obviously Google and Facebook are protected by founder control. The founder is the mission guardian. Those are the structures. And I tell the story in the book, a very weird experience I had. I was literally at the Vatican of all places. I mean, look at me. What am I doing at the Vatican? But it was cool. [inaudible 01:17:06] had convened a conference on AI governance and they invited me to speak. So I was on this panel at the Vatican with every other AI company and me. It was weird. It was actually strange. It was like me, literally me. I go down. I was looking down the row. Me, Anthropic, OpenAI, Google, Cohere, Palantir, everyone on this panel together. And I looked on the panel and I realized not a single one of these companies has standard governance. That's how bad it is. Nobody making this technology would say, "Oh yes, standard governance, that's fine. I'm sure that..." It's just too dangerous. But they of course take different approaches to mission guardianship, some of which are better and some are worse. Like founder control I think actually has a lot of downsides to it.
01:17:45 · Lenny Rachitsky, Eric Ries
[01:17:45] S1: So and when you say standard governance, every one of these AI companies is not doing it the way every other startup was doing? They understood they needed to do something different to protect- [01:17:54] S0: Yeah. [01:17:54] S1: ... humanity essentially from AI? Okay. [01:17:56] S0: Yeah. Yeah, exactly. Because otherwise it's just this technology is so valuable. If you say that shareholders should run it, then you're saying literally whoever can borrow the most money should be able to control this technology. It's just, that's nuts. That can't be how it works. No way. I don't think any company should be governed that way, but certainly not AI companies. Certainly not. [01:18:18] S1: Okay. So just to kind of plant these things in people's heads as they, okay, we need to really think about this stuff. There's a nonprofit approach to this, there's... What are the terms that people should just think about?
01:18:30 · Eric Ries
[01:18:30] S0: So we talked about we need a mission guardian. So first question, is the mission guardian a person or a thing? That's our kind of first decision point. So yeah, some, I think for early stage companies, founder control is fine as a good bridge, a temporary bridge to a more permanent structure. And some companies get really far with founder control. That's okay. But a lot of founders who have founder control wind up really miserable, as you can see by the fact they're having a mental health breakdown right in front of us, all of us on social media basically every day. Because you become like Atlas, you can't even shrug. It's you holding back the abyss. It's a lot. So I think a better, more permanent solution is to encode the protection into the structure. Now that can be done with a single entity like Costco. Costco just has the rules written right into the structure itself. But that means that every time they lose one of the structures, every once in a while when an attack chips off a little bit of it, they don't have any way to grow them back. So it's like you've built this fortress, but you don't have a way to renew it. The better way, I think, according to the evidence, is to have some stakeholder, somebody be the steward of the mission and have a way to renew that person, that set of people. So some people accomplish that by what's called an employee ownership trust. So the employees are the mission guardians, like the John Lewis Partnership in the UK is a famous example. Obviously cooperatives have this... This can work in a cooperative at scale. Mondragon, for example, in Spain has like 80,000 employees. It's a huge company, but they're all employee cooperatives. You can do it through an employee voting trust. That's how Alibaba is protected, where the employees vote for the board members rather than the reverse. But those are much more complicated compared to me, the two simplest solutions are either a nonprofit foundation as in the Novo Nordisk example, or in the case of what's called the Perpetual Purpose Trust or PPT. And Perpetual Purpose Trust is a non-economic entity. So whereas the Novo Nordisk Foundation is the largest charitable foundation in the world because it owns a big chunk of Novo Nordisk and that's worked out pretty well. The Anthropic long-term benefit trust has no economic dimension to it at all.
01:20:30 · Eric Ries
It only has the mission, oversight, responsibility. And what's nice about a Perpetual Purpose Trust, as in the case of Patagonia is governed with a purpose trust also, is you have the trustees, but then you also have someone whose job is actually to sue the trustees if they ever deviate from the mission. So you actually have what's called the purpose protector who's an extra person who can get in there and say, "If things go wrong." So to me, it's like you have checks and balances like in a government. It's more stable. It's a more stable structure. To me though, I use the omnibus term spiritual holding company to describe this category of things. Because I said before, we have a lot of infighting. The people who advocate for each of these things thinks their version is the best. I didn't even mention the Evergreen, the Tugboat Foundation. They don't believe in having investors at all. So there's people who think the solution to this is no investors involved, that makes the problem a lot easier. So anyway, there's a lot of different ways this can be done. I don't think I even mentioned ESOPs. There's just so many. So we need to have an omnibus term. I call it the spiritual holding company. The holding company, like the Berkshire Hathaway, but rather than having everything be wholly owned, it's the holding company for the spirit, the animating essence of the whole. And it's a lot of evidence in the book, a lot that this structure is more stable, more durable, more likely to invest in quality and R&D and the things we really care about and better for shareholders than the conventional structure.
01:21:53 · Lenny Rachitsky
[01:21:53] S1: If this all sounds like a huge drag and a lot of work and really annoying, I'd say go back an hour when we talked a lot about just what is it you were trying to avoid and the pain.