It’s been a year since Ethereum completed its switch to proof of stake. Vitalik Buterin on what’s new, what’s next, and how he’s using AI

He also discusses what’s happened since Ethereum’s transition from proof of work to proof of stake and how spending time in Antarctica could help prepare humans—maybe even him?—for interplanetary travel.

(This interview has been edited for length and clarity.)

What use cases for blockchain excite you the most?

There are a few different categories of blockchain applications that I think can be valuable, and I wrote about some of these in a blog post last year, but a summary would be: One, currency applications are valuable, especially in the developing world. And two, identity, reputation, and decentralized social applications are exciting due to their potential network effects and enhanced security.

Going into the specifics of one application, I’ve been using Farcaster a lot recently. There’s a lot of interest in creating alternative stuff to Twitter now—lots of people are either unhappy with various things that Elon does or people just realize that this is a monopoly and we need alternatives.

Some of these have been on-chain. For example, my Farcaster account is connected to an Ethereum address, and if I lose my Farcaster key, I can use my Ethereum address to recover it. This is actually more secure than what centralized social media do, where they rely on phone numbers that can be hacked.

One of the big challenges in social media is how you tell who is a human and who is a bot. There’s just lots of incentives for malicious actors to just create 50,000 bots. The nice thing about Ethereum is that it already has this lively on-chain ecosystem where there are different ways of doing that.

One is you can verify just the existence of activity on an Ethereum account. To have activity, you need to pay transaction fees, so there’s a cost. You can make a few hundred fake accounts if you can pay a cost, but you can’t make a million of them.

Second, you can look at POAPs [proof-of-attendance protocols] that verify that you participated in various in-person or online events.

Third, there are proof-of-personhood protocols, and there is proof of humanity like Worldcoin. There’s a Gitcoin passport that aggregates a lot of those. The ideal goal here is to do this in a way that lets people prove that they are people and not bots, but is also compatible with the goal of preserving people’s privacy.

Last year, you experienced the milestone of transitioning Ethereum from proof of work to proof of stake. What are you focused on now, and then what’s next?

Account abstraction is definitely a big and important one. Another thing is obviously scalability. We have EIP-4844, also known as Proto-Danksharding, which is coming out very soon, which basically adds more data space on the blockchain—not transactions, but just blobs of data. The purpose of this is to enable greater scalability, specifically for layer-2 projects, our rollups.

Then after EIP-4844 there is a longer-term road map of increasing that amount of space over time and then ongoing improvements to proof of stake in identifying ways that we can improve decentralization even further.

We talked a year ago about AI, the intersection of AI and blockchain, before the release of ChatGPT and all the hype around that. What are your current views on the AI industry?

AI is obviously super important. It’s the sort of thing that’s hard to think about because it’s so hard to predict where it might go in the future. It’s massive gains in productivity, but also massive philosophical consequences just in terms of what the role of humans in this new world is going to be. The world that is controlled by AI has completely different goals from us in a way that’s difficult to detect at first—that could also lead to very bad things happening.

In terms of the intersection between AI and crypto, this is one of those things where people have tried really hard to find an answer. Unfortunately, part of the answer is just that the intersection is less than it seems. There are a few things like ZK ML [zero-knowledge machine learning], but it’s realistically a pretty small area.

There is the fact that AI’s rapid growth makes it much more important to solve certain problems, like proof of personhood, for example. Then there’s obviously the possibility of just crypto tokens being used by AIs. There is the possibility of crypto-based DAOs governing AIs and the process of training AIs.

Have you read any books in the last year that have inspired you?

In the last year, I have definitely done less books and more reading blogs and things on the internet, things like podcasts and even science fiction. Maybe 10 years ago I read a lot of Austrian economics literature and things like Human Action: A Treatise on Economics, but also things like The Moon Is a Harsh Mistress, a novel by Robert A. Heinlein about a society that creates a civilization on the moon and tries to gain independence.

Much more recently, I watched some of The Three-Body Problem—I read part of the novel and I also watched part of the TV series, and I was fascinated by that because it presents its world view in a way that is totally different from things that we’re used to. There are aspects to the world view that I disagree with. There’s always this mentality that the world is super dangerous and basically the five big countries have to get together and figure things out, and no one else gets a seat at the table.

That’s not the kind of world that I want, and not the kind of mentality that I want to see get bigger. But there is this aspect of appreciating the vastness of the universe and how lucky humanity is to have been born—and how significant and how small we are at the same time.

Are there people you admire, who inspire you?

A lot of people. Akira Fuyuno has a blog called Samurai Novelist. It’s not a book, but his posts are so long, they’re basically like the equivalent of reading 10 books. He just does a very good job of making really important philosophical concepts understandable and talking about just all kinds of things in science and politics in deep ways. Other people have told me that my writing style is inspired by his.

I kind of have to say Volodymyr Zelensky. I appreciate his soft-power leadership and his ability to be a figure who rallies and inspires people in that way—like, not just giving out the orders, but just being the face and rallying people behind the cause.

Then there are Eliezer Yudkowsky and Scott Alexander—the two of them are kind of similar characters in a lot of ways. Also, within the Ethereum community, Justin Drake, Dankrad Feist, and Aya Miyaguchi, the foundation’s executive director, who’s really pushed the philosophy of subtraction—different people who have stepped up and made a piece of Ethereum their own and put their souls into making that piece as great as possible.

You’ve also mentioned Steve Wozniak. Can you tell us more about meeting him? And maybe a few other people who’ve impressed you?

Steve Wozniak, he was just a genuinely fascinating guy who had really interesting hobbies and a lot of cool things that he was able to tell me about.

From the government, Audrey Tang, Taiwan’s digital minister, she had all kinds of interesting ideas of her own. She has her own history of being involved in Taiwan’s democracy movement and then g0v [gov zero], which is trying to create a more democratic e-government infrastructure in Taiwan.

The likely new prime minister of Montenegro, Milojko Spajić, a.k.a. Mickey, I’ve met him and I’m also impressed by him. From what I can tell, it’s his dedication to making the country better and also just being a curious person who was willing to talk to me and willing to learn things about crypto.

Do you use ChatGPT or any other AI tools?

I’ve used ChatGPT. I’ve also used the Stability AI chat. I use that for a few things. One is asking specific questions that I want answers for; another one is for writing code. I find that the thing that ChatGPT is best at is writing code that uses APIs or libraries or tools that you did not really know about. It even sometimes can take natural language tasks and convert them.

Also, for a translation, ChatGPT often does better than even Google Translate for helping to learn languages. If I want to understand the grammar behind a particular sentence, it does a really good job of doing that. The things that it’s less good at are basically things where there aren’t many millions of examples to learn from.

I use either Stable Diffusion or the various other AI drawing things to draw images. I use those images and things in, like, my posts and my slides quite a bit. That has been it so far, but I’m sure as AI gets better, I’ll use it in other places as well.

It definitely does increase my productivity by quite a bit. 

What’s your biggest dream?

That’s a good question. I feel like I’m not the sort of person to have one dream. I’m the sort of person to have a lot of dreams. I dream of a world where I can be more free and more open, and where the tools that we’re building in the crypto space can be a big part of actually making that happen. I have a dream of humanity as a whole really succeeding and ending disease and becoming a multiple-energy-area civilization.

Speaking of civilization, do you think in the future we can travel between planets? And would you take that ride to space?

I think we’ll absolutely get there. I will not be the first one to go—I might be the hundredth person to go. Before we’re really ready to do space, we need to get better at doing Antarctica first. I think of Antarctica as being the tutorial ground of going into space, because it’s an environment that’s definitely somewhat harsh and you have zero infrastructure, but you still can breathe the air without wearing a helmet. Start with the simpler things. But I would definitely love to actually be a part of that myself at some point over the next few decades.