Vitalik Buterin holding Zuzu, the puppy rescued by people of Zuzalu. Photo: Michelle Iai
Don’t try finding Zuzalu on a map; it doesn’t exist anymore. It was a “pop-up city” conceived by the tech entrepreneur Vitalik Buterin, creator of Ethereum, and a group of like-minded people to facilitate co-living and collaboration in fields like crypto, network states, AI, and longevity. It was also, in substantial part, funded by Vitalik.
Zuzalu, located on the Adriatic coast of Montenegro, began its short history on March 25 and wound down on May 25. It was a complex and memorable phenomenon, and I’m wrapping my mind around a larger article in the works.
Usually, I don’t eat breakfast due to my intermittent fasting regimen, but in Zuzalu, breakfast, served at a particular local restaurant, was the healthiest meal of the day. Also, it was free (kudos to Vitalik, and more on that later). Most importantly, it was the place to meet new people.
This was also where, on one of my last days in Zuzalu, I sat down with Vitalik himself for a talk. Not the best setting for an interview, considering the steady hum of voices and utensils clanging in the background, but it was the only gap in Vitalik’s busy schedule.
Vitalik is 29, slender and mild-mannered, with a soft, pensive smile. When he talks, his train of thought moves fast, fueled by intelligence and curiosity. He seems to be genuinely interested in how the world works and just as genuinely disinterested in his own status – something that was characteristic of Zuzalu as a whole.
Like any Zuzalu breakfast chat, ours was a bit all over the place, and we eventually ended up discussing the possibility of an AI-driven apocalypse (everyone’s favorite topic there). Apologies to the longevity purists reading this. However, we started with Zuzalu itself.
First, what does “Zuzalu” even mean?
Zuzalu intentionally does not mean anything in any language.
I guess that explains the debates around its meaning. How was this idea born?
The idea came about six months ago. I was already thinking about many different topics at the same time. I reviewed Balaji’s book last year, so I was thinking about network states, but also about crypto, real-world applications of Ethereum, other zero-knowledge proofs, and so on.
I am also a fan of the longevity space, I read Aubrey’s book when I was a teenager, and I know how important this is. The idea came together, as an experiment, to try doing things in all those areas at the same time.
I thought we’d take 200 people, some from the Ethereum space, some from longevity, some philosophers, people just interested in building societies, and so on, bring them together for two months, and see what happens. The rationale behind the size is that it’s a large enough leap from the things people do already.
We have big conferences, but they only last a week, and we have hacker houses, but those only have ten people. So, let’s do something with two hundred people that would last for two months. It’s a big enough jump to create something new, but it’s still manageable. It’s not something crazy like going from 0 to 5,000.
Why Montenegro?
I knew a couple of locals here in Montenegro, having been introduced to the country last year. The government has been very open to becoming more crypto-friendly. On my first visit, they gave me citizenship, something that no other country has done. They did a lot, and I just happened to know people here who are very good at logistics and organization. From there, people started joining in. The team and the organization started growing very quickly.
Did it work? Zuzalu is winding down, and you probably have drawn some conclusions by now.
I think it worked. Many people reported how much they enjoyed the experience, how happy they were, how this gave them a feeling of community and family. Maybe things are different now, but when I did a poll a month ago, a third of the people here were digital nomads. One of the problems digital nomads always face is loneliness. You don’t have company, you’re going to unfamiliar places, it can be hard. Some of those people enjoy the digital nomad experience, they like to travel like that, but others are doing it out of necessity.
Like people from Ukraine and Russia? I see a lot of them here.
Yes, and also from places like China. So, that part was a success. On many other things, there were some successes and some things we can learn from. The big idea was that 200 people is already an economy of scale. It enables you to do things collectively that take too much effort to do as a person.
For instance, if you want food that’s different from what most other people eat, usually you have to go get it yourself. You go to a restaurant, and even if you order a salad or fish, you don’t know what oil they use, and so on. Here, because we represent so many people, we talked to this restaurant, and we told them what menu to use for breakfast. It’s not perfect, but we tried to follow Bryan Johnson’s Blueprint menu as much as we could, although many ingredients were very hard to get. But it’s still much better than the average breakfast [at this point, I’m nodding with my mouth full].
For some things beyond that, at least for the first half of Zuzalu, there weren’t enough champions to push many of the ideas, but that has improved a lot recently. People are forming clubs for exercise, such as the cold plunge club, hiking, and others.
Things like the pop-up gym, right? There was a demand for this, so people just organized and did it.
Exactly. If you’re one person, you will not be able to have a gym, but as a group, you can make that happen. Biomarker testing that we organized also comes to mind. People enjoy doing things together.
What about introverts [I am mildly introverted myself]? Is Zuzalu a good place for them?
I feel like it’s trying to be. I think the challenge that all these co-living projects have is that if you make co-living the primary meme, you’re going to mostly attract people who want to be very close with other people, who enjoy collective cooking and stuff like this. But for many other people it’s not a good fit.
Here, it’s much more moderate in a lot of ways. People have their own apartments. If you want to retreat to your apartment and not talk to other people, you can. You are not obligated to show up for any of the events. You don’t have to eat at restaurants three times a day, don’t have to talk to people all the time. Our model gives people more choice without pushing them into a lifestyle that’s not compatible with them.
Then, there’s this interesting thing I have noticed… I have one friend here who is an extreme introvert. Normally, he goes off by himself, doesn’t really talk to people, and here he just did, he started talking to people more because those were people he wanted to talk to.
Yes, I’ve noticed that about myself too. How do you think the education part went?
On the education side, one of the big weaknesses was that we tried to organize different weeks, for each week to have a theme. There was a synthetic biology week, then public goods, then zero-knowledge proofs, then free cities and network states, now longevity. Some aspects of that work were interesting for people, but there’s a reason why college courses are in parallel and not in series. People learn better when it’s spaced over a long period of time. We didn’t do that, and that probably was a mistake.
In one of your talks here, you mentioned “cross-pollination” in the longevity field. Have you witnessed any cross-pollination between longevity and other fields?
I would say yes. I think there were two big cross-pollination events here. One is the intersection between longevity and crypto, such as the decentralized science space.
Which has been happening for some time.
Exactly, it has been happening. It has brought many different people from those groups together. I know that a lot of connections were made between science people and public goods people. I think that a lot of people realized that funding science is a natural fit for some of the work that public goods people have been doing.
The second cross-pollination event happened between the longevity people and people building new cities. There are people from Prospera here, from VitaDAO, and now, they are working much more closely together than ever before.
On the other hand, I’ve been getting this vibe from some of the tech people: “why are all those longevity people here, why do we need it?” Even those interested are sometimes dissatisfied with what we can offer them. I’ve heard things like “you, longevity people, can’t give us answers on what should we do to live longer”.
This is probably a fair question. It is true that longevity as a field has been around for many years, and we still don’t have the magic pill for immortality or anything close to that. There are very fundamental reasons why that’s true for longevity, while AI is seeing much more progress. I think we just know a lot less about the body, as it’s an incredibly complicated machine.
The way I see this question is that if you look at the difference between the first computer and what we have now, the difference is huge. By the standards of the 1950s, today’s computers feel like magic. There’s a common phrase that people always overestimate the short term and underestimate the long term, and I personally expect the longevity field to have a similar kind of progress. There are a few decades that might look useless from the outside, but they’re laying the foundations, and then the gains become faster than most people expect.
I’ve been asking people how all this – longevity, crypto, AI, network states – comes together, and some said, well, those are just Vitalik’s spheres of interest, that’s all. What do you say to that?
It’s not just my intersection. I feel like a lot of people got into those things at the same time. There’s definitely a pretty significant cluster of the crypto space that’s also interested in longevity, especially older Ethereum people.
Was one of your intentions to show them “longevity done right”?
You could say that. One of the big criticisms of the longevity space is this idea that you’re extending life, but is the life you’re extending worth living? It’s the misconception that we’re basically trying to keep 80-year-olds barely alive. I’m trying to show that this is not the case, that the longevity space is specifically about repairing damage before it develops into a pathology.
But then people see someone like Bryan Johnson. He is a multimillionaire who literally puts his life into being as healthy as possible. He takes this extremely customized menu, a huge number of supplements, spends his entire days doing exercises and so on. People look at that and they think, first, that it is only accessible to rich people, and, second, this is something you’d only do if you don’t care about actually living your life. Neither of those things are necessarily true.
To me, a part of the motivation was to show people a different model. It’s also a personal struggle for me. I can’t dedicate my entire life to being healthy. I have Ethereum stuff, I need to travel everywhere, I’m a nomad, all my supplies are in a 40-liter backpack, so I have to compromise between a lot of things.
What we tried to show here is that if we do things in groups with economies of scale, it can really help the average person to maintain a reasonable lifestyle routine, including things like exercise and diet.
There are people here who are pretty intense about health stuff – as we said, cold plunges, sauna, gym. I know someone who runs for two and a half hours every day. Still, they don’t look like they’re willing to sacrifice their life to extend their lifespan.
My take on Bryan is that he’s an elite “rejuvenation athlete”, and like every elite athlete, he’s going that extra mile, putting, say, two times more effort into getting 10% more result. That’s his choice but we don’t have to follow it, there are other ways.
I totally agree, and that’s an argument that not enough people are making. Bryan’s example creates an impression that you have to go out of your way to stay healthy, but I think the extent to which it’s true is exaggerated. If you look at Aubrey, he is pretty “normie” in his personal lifestyle, but the people who make news are usually on the extreme ends of things. I think it’s good that they exist, and we’ve learned a lot from Bryan, but someone has to make a different case.
Are you going to continue the Zuzalu experiment, maybe on a longer time scale?
I would say, absolutely. We did a poll about one and a half weeks into the experiment, and one of the questions was, if there was another Zuzalu, would you show up? Zero people voted “no”.
I think it’s going to be renewed anyway, with or without us. When we asked who was thinking of making their own Zuzalu, about 15 people raised their hands. It’s going to happen, and the question is, what role are we taking in this experiment?
Scaling is a big challenge. There’s a difference between doing this for two hundred people and doing something that includes thousands or tens of thousands of people. Once you have this number of people, it’s not one village anymore, you will have interactions between villages, you will have conflicts.
There’s also the question of, what’s the long-term goal of this. If you want to create a biotech-friendly network state, you can’t jump locations every two months. The equipment is not going to move, and you can’t convince a new country to install favorable regulations every two months. Convincing even one is hard.
On the other hand, if your goal is to, say, create a new type of university, then moving every two months would be great. Giving people new experiences would make learning even more enjoyable.
So, different groups have different needs. Figuring out what makes sense for people is a learning process. That’s true for cities too. You have big cities and small cities, cities focused around particular industries, university towns, natural resources gathering cities, trade towns. All these look different. For any new category of institutions that are based on co-living in person, you will have to account for this diversity.
Overall, it feels like the basic format has been validated; it turned out to be something that a lot of people like and enjoy more than their usual life. People are willing to spend a lot of time here rather than in big cities. In the future, with better choice of location, with better preparation, this can be much cheaper than big cities, more enjoyable and more useful professionally for many people. So, many things were proven, but there was also probably a huge number of small mistakes.
Another interesting longevity-related thing I’ve heard here more than once, uttered quite seriously, is that there’s no point in extending lifespan because humanity will soon perish at the hands of an almighty AI. What’s your take on it? Are you personally an AI “doomer”?
I think there’s some chance that the arguments that AI doomers make are correct, but that chance is far from 100%. I think it’s good to worry about those things. I’m happy that people are taking the problem of AI alignment seriously. It’s a small amount of work that could make a big difference, so it’s obviously worth doing.
Do you agree with the idea of stopping or slowing down AI development until we figure it out?
It’s harder for me to be convinced that taking that step is a good idea because it has its own risks. The very first question is “How do you even enforce it?” We have all those different countries that are going to have their own ideas. If some countries try to enforce a slowdown when others do not want to go along, that could itself lead to serious conflicts.
Also, slowing down AI obviously slows down longevity research. Many people think longevity is fundamentally hard, and we will need strong AI to make this problem solvable.
It’s easier for me to be convinced that we need some medium level of more carefulness and slowing down of some specific things than to be convinced of more drastic attempts to slow AI progress greatly or stop it outright.
It’s easy to dismiss the “doomers” until you realize how much thought they have put into their position – something being in Zuzalu has helped me with.
I agree with that, and that’s a big part of why I do take them seriously. They have powerful arguments, and many people who argue against the “doomers” have only very basic counterarguments that the “doomers” already thought of and responded to ten years ago. I’m definitely not going to just dismiss their arguments. If people do suggest pragmatic ways to either slow down AI research or put a lot of resources into solving this problem, I’ll be very open to that.
I guess it’s hard for me to accept either of the extreme positions – either that we’re clearly going to be totally fine, or that there’s a greater than 50% chance we’ll all die – because there’s just so many unknowns. For example, five years ago when the best AI was AlphaZero, I don’t think it was even within many people’s space of possibilities that we’re going to switch away from goal-directed reinforcement learning and toward this really weird paradigm of managing to solve thousands of problems by, like, predicting text on the internet. So, I expect similar things that are outside of our current imagination to happen another few times before we get to the “singularity”.
If I had to predict a concrete place the AI doomer story is wrong, if it had to be wrong, I would say it’s in the idea of a fast take-off: that AI capabilities will pile on so fast that we won’t be able to adapt to problems as they come. We may well have a surprisingly long period of approximately human-level AI. But then again, these are only speculations, and you should not take me for a specialist.
I guess you have followed the AI safety discussion here in Zuzalu. Do you think it was fruitful?
I think yes, but also kind of chaotic. Many people have not been exposed to deep AI issues at all, and then Nate [Soares, head of MIRI] is coming in with those very deep radical arguments on why AI is going to destroy the world. There’s this big disconnect between what one side believes and the other side believes, something you can’t resolve in a three-day conference.
I am not an expert, but I’ve had several chats on this topic here, including with Nate. I think the main question is why and how would AI develop its own goals, and many people’s answer is that AI will misinterpret the goals we give to it, but then the whole thing can be distilled into “how do we get better at formulating those goals?”
I think Nate would say that this is the entire problem they’re trying to solve.
That doesn’t seem impossible.
As I understand his argument, it’s basically that even if we make a definition that works really well from our point of view, and if we had it trained on ten million examples, and it makes sense to us, the AI will be much more computationally powerful than we are, and it will find some really weird way to satisfy its model of those values in a way that totally goes against what the original intention was. Just how tractable or intractable that problem is, is one of the things that are very hard for me to judge, because it’s so abstract.
I still think this makes the problem more manageable. If we believe that at some point, AI will develop its own goals, there’s not a lot we can do about it, but if we think that there’s a big distance between a very smart AI and an independent entity with its own goals, then I’m a bit less scared.
Yes, I think there’s a big chance that the alignment will turn out to be much simpler than we expected, and the time period during which a combination of human and AI will continue to be smarter than AI alone will be much longer than we expected.
I also think there’s a big chance that there are no easy strategies for destroying the entire world. The few counterexamples like biolabs can be dealt with individually instead of dealing with them on the AI side. There’s also some chance that humans are much closer to the ceiling of what kind of intelligence is possible to have from AI.
Still, I think there are many different totally unknown things that could happen, and our prediction power is limited. People generally did not predict that we would go that fast from a more goal-directed AI like AlphaZero to a less goal-directed AI like ChatGPT. It shows you how easy it is to have all kinds of surprises.
I also don’t want all that I’m saying here to be misinterpreted as my definite statement when in reality, my thoughts on this are going in all kinds of different directions and I could easily disagree with myself a year from now.
Do you think that at some point, we will merge with AI to stay relevant?
I’d say probably. I don’t know how such a merger would look like though.
Are you open to it or afraid of it?
[Long pause] I’m curious about it.
